Journal of Personality and Social Psychology, 2006, Vol. 91-04

Volume 91 Number 4 Published monthly by the American Psychological Association October 2006 ISSN 0022-3514 Journal of...

Author: JPSP

323 downloads 2111 Views 8MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Volume 91 Number 4 Published monthly by the American Psychological Association

October 2006

ISSN 0022-3514

Journal of

Personality and Social Psychology ATTITUDES AND SOCIAL COGNITION

Charles M. Judd, Editor Dacher Keltner, Associate Editor Anne Maass, Associate Editor Bernd Wittenbrink, Associate Editor Vincent Yzerbyt, Associate Editor INTERPERSONAL RELATIONS AND GROUP PROCESSES

John F. Dovidio, Editor Daphne Blunt Bugental, Associate Editor Jacques-Philippe Leyens, Associate Editor Antony Manstead, Associate Editor Cynthia L. Pickett, Associate Editor Jeffry A. Simpson, Associate Editor Scott Tindale, Associate Editor Jacquie D. Vorauer, Associate Editor PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES

www.apa.org/journals/psp

Charles S. Carver, Editor Tim Kasser, Associate Editor Mario Mikulincer, Associate Editor Eva M. Pomerantz, Associate Editor Richard W. Robins, Associate Editor Gerard Saucier, Associate Editor Thomas A. Widiger, Associate Editor

The Journal of Personality and Social Psychology publishes original papers in all areas of personality and social psychology. It emphasizes empirical reports but may include specialized theoretical, methodological, and review papers. The journal is divided into three independently edited sections: f ATTITUDES AND SOCIAL COGNITION addresses those domains of social behavior in which cognition plays a major role, including the interface of cognition with overt behavior, affect, and motivation. Among topics covered are the formation, change, and utilization of attitudes, attributions, and stereotypes, person memory, self-regulation, and the origins and consequences of moods and emotions insofar as these interact with cognition. Of interest also is the influence of cognition and its various interfaces on significant social phenomena such as persuasion, communication, prejudice, social development, and cultural trends. f INTERPERSONAL RELATIONS AND GROUP PROCESSES focuses on psychological and structural features of interaction in dyads and groups. Appropriate to this section are papers on the nature and dynamics of interactions and social relationships, including interpersonal attraction, communication, emotion, and relationship development, and on group and organizational processes such as social influence, group decision making and task performance, intergroup relations, and aggression, prosocial behavior and other types of social behavior. f PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES publishes research on all aspects of personality psychology. It includes studies of individual differences and basic processes in behavior, emotions, coping, health, motivation, and other phenomena that reflect personality. Articles in areas such as personality structure, personality development, and personality assessment are also appropriate to this section of the journal, as are studies of the interplay of culture and personality and manifestations of personality in everyday behavior. Manuscripts: Submit manuscripts to the appropriate section editor according to the above definitions and according to the Instructions to Authors. Section editors reserve the right to redirect papers among themselves as appropriate unless an author specifically requests otherwise. Rejection by one section editor is considered rejection by all; therefore a manuscript rejected by one section editor should not be submitted to another. The opinions and statements published are the responsibility of the authors, and such opinions and statements do not necessarily represent the policies of APA or the views of the editors. Section editors’ addresses appear below:

ATTITUDES AND SOCIAL COGNITION Charles M. Judd, Editor c/o Laurie Hawkins Department of Psychology University of Colorado UCB 345 Boulder, CO 80309

INTERPERSONAL RELATIONS AND GROUP PROCESSES John F. Dovidio, Editor Department of Psychology University of Connecticut 406 Babbidge Road Storrs, CT 06269-1020

PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES Charles S. Carver, Editor ATTN: JPSP: PPID Department of Psychology University of Miami P.O. Box 248185 Coral Gables, FL 33124-0751 Change of Address: Send change of address notice and a recent mailing label to the attention of the Subscriptions Department, American Psychological Association, 30 days prior to the actual change of address. APA will not replace undelivered copies resulting from address changes;

journals will be forwarded only if subscribers notify the local post office in writing that they will guarantee periodicals forwarding postage. Electronic access: APA members who subscribe to this journal have automatic access to a 3-year file of the journal in the PsycARTICLES姞 full-text database. See http://members.apa.org/access. Reprints: Authors may order reprints of their articles from the printer when they receive proofs. Single Issues, Back Issues, and Back Volumes: For information regarding back issues or back volumes write to Order Department, American Psychological Association, 750 First Street, NE, Washington, DC 20002-4242. Microform Editions: For information regarding microform editions, write to University Microfilms, Ann Arbor, MI 48106. Copyright and Permission: Those who wish to reuse APAcopyrighted material in a non-APA publication must secure from APA and the author of reproduced material written permission to reproduce a journal article in full or journal text of more than 500 words. APA normally grants permission contingent upon like permission of the author, inclusion of the APA copyright notice on the first page of reproduced material, and payment of a fee of $20 per page. Permission from APA and fees are waived for those who wish to reproduce a single table or figure from a journal for use in a print product, provided the author’s permission is obtained and full credit is given to APA as copyright holder and to the author through a complete citation. (Requesters requiring written permission for commercial use of a single table or figure will be assessed a $25 service fee.) Permission and fees are waived for authors who wish to reproduce their own material for personal use; fees only are waived for authors who wish to use more than a single table or figure of their own material commercially (but for use in edited books, fees are waived for the author only if serving as the book editor). Permission and fees are waived for the photocopying of isolated journal articles for nonprofit classroom or library reserve use by instructors and educational institutions. A permission fee may be charged to the requester if students are charged for the material, multiple articles are copied, or large-scale copying is involved (e.g., for course packs). Access services may use unedited abstracts without the permission of APA or the author. Libraries are permitted to photocopy beyond the limits of U.S. copyright law: (1) post-1977 articles, provided the per-copy fee in the code for this journal (0022-3514/06/ $12.00) is paid through the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923; (2) pre-1978 articles, provided that the per-copy fee stated in the Publishers’ Fee List is paid through the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923. Address requests for reprint permission to the Permissions Office, American Psychological Association, 750 First Street, NE, Washington, DC 20002-4242. APA Journal Staff: Susan J. A. Harris, Senior Director, Journals Program; Skip Maier, Director, Journal Services; Paige W. Jackson, Director, Editorial Services; Becky Shaw, Account Manager; Jodi Ashcraft, Advertising Sales Manager.

Journal of Personality and Social Psychology (ISSN 0022-3514) is published monthly in two volumes per year by the American Psychological Association, 750 First Street, NE, Washington, DC 20002-4242. Subscriptions are available on a calendar year basis only (January through December). The 2007 rates follow: Nonmember Individual: $442 Domestic, $488 Foreign, $515 Air Mail. Institutional: $1,349 Domestic, $1,443 Foreign, $1,470 Air Mail. APA Member: $208. Write to Subscriptions Department, American Psychological Association, 750 First Street, NE, Washington, DC 20002-4242. Printed in the U.S.A. Periodicals postage paid at Washington, DC, and at additional mailing offices. POSTMASTER: Send address changes to Journal of Personality and Social Psychology, 750 First Street, NE, Washington, DC 20002-4242.

The paper in this journal meets or exceeds EPA guidelines for recycled paper. Since 1986, this journal has been printed on acid-free paper.

NEW RELEASES from the American Psychological Association

Second-Order Change in Psychotherapy The Golden Thread That Uniﬁes Effective Treatments

J. Scott Fraser and Andrew D. Solovey 2007. 312 pages. Hardcover. List: $69.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-436-1 ISBN-13: 978-1-59147-436-4 Item # 4317117

Rumor Psychology Social and Organizational Approaches

Nicholas DiFonzo and Prashant Bordia 2007. 392 pages. Hardcover. List: $69.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-426-4 ISBN-13: 978-1-59147-426-5 Item # 4316079

Educating the Human Brain Michael I. Posner and Mary K. Rothbart

Dialogues on Difference Studies of Diversity in the Therapeutic Relationship

Edited by J. Christopher Muran 2007. 336 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-451-5 ISBN-13: 978-1-59147-451-7 Item # 4316083

Insight in Psychotherapy Edited by Louis G. Castonguay and Clara E. Hill 2007. 488 pages. Hardcover List: $69.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-477-9 ISBN-13: 978-1-59147-477-7 Item # 4317122

Contributions Toward Evidence-Based Psychocardiology A Systematic Review of the Literature

Edited by Jochen Jordan, Benjamin Bardé and Andreas Michael Zeiher 2007. 552 pages. Hardcover. List: $99.95 APA Member/Afﬁliate: $59.95 ISBN 1-59147-358-6 ISBN-13: 978-1-59147-358-9 Item # 4318028

Shy Children, Phobic Adults Nature and Treatment of Social Anxiety Disorder Second Edition

2007. 272 pages. Hardcover. List: $79.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-381-0 ISBN-13: 978-1-59147-381-7 Item # 4318029

Psychology and Economic Injustice Personal, Professional, and Political Intersections

Bernice Lott and Heather E. Bullock 2007. 192 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-429-9 ISBN-13: 978-1-59147-429-6 Item # 4318029

Bereavement in Late Life Coping, Adaptation, and Developmental Inﬂuences

Robert O. Hansson and Margaret S. Stroebe 2007. 232 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-472-8 ISBN-13: 978-1-59147-472-2 Item # 4317119

APA Dictionary of Psychology Editor-in-Chief: Gary R. VandenBos 2006. 1,008 pages. Hardcover. List: $49.95 APA Member/Afﬁliate: $39.95 ISBN 1-59147-380-2 ISBN-13: 978-1-59147-380-0 Item # 4311007

Deborah C. Beidel and Samuel M. Turner 2007. 408 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-452-3 ISBN-13: 978-1-59147-452-4 Item # 4317118 AD0472

To O r d e r : 8 0 0 - 374 -2721 • w w w. a p a .or g / b o o ks

apa books

Available July 2006

A Landmark Reference That Deﬁnes the Lexicon of Psychology

APA Dictionary of Psychology Editor-in-Chief: Gary R. VandenBos, PhD The American Psychological Association is proud to announce the publication of an invaluable addition to your reference shelf, one that represents a major scholarly and editorial undertaking. With over 25,000 terms and deﬁnitions, the APA Dictionary of Psychology encompasses all areas of research and application, and includes coverage of concepts, processes, and therapies across all the major subdisciplines of psychology. Ten years in the making and edited by a distinguished editorial board of nearly 100 psychological scholars, researchers and practitioners, the Dictionary is destined to become the most authoritative reference of its kind. Academicians, researchers, clinicians, undergraduates and graduate students, and professionals in allied mental health, education, medicine, and law, as well as academic and public libraries, will ﬁnd the Dictionary essential. 2006. Hardcover. 1,008 pages. List: $49.95 | APA Member/Afﬁliate: $39.95 ISBN 1-59147-380-2 | Item # 4311007

The APA Dictionary of Psychology includes •

25,000 entries offering clear and authoritative deﬁnitions

•

Thousands of incisive cross-references directing the user to synonyms and antonyms, acronyms and abbreviations, and related terms and concepts that deepen the user’s understanding of related topics

•

Balanced coverage of over 100 subject areas across the ﬁeld of psychology including clinical, experimental, neuropsychology, cognitive, personality and social, developmental, health, psychopharmacology, methodology and statistics, and many others

• 800-374-2721 www.apa.org/books

Entries include nearly 8,000 terms from the APA’s Thesaurus of Psychological Index Terms® which helps students and researchers reﬁne their APA database searches (such as the ﬂagship PsycINFO® bibliographic database’s 2+ million records)

• •

“A Guide to Use” and “Quick Guide to Format” that together explain important stylistic and format features to help readers most effectively use the Dictionary Each of four appendices gathers terms into a thematic summary listing, covering (1) biographies; (2) institutions, associations and organizations; (3) psychological therapies and interventions; and (4) psychological tests and assessment instruments

AD0470

Journal of

Personality Social Psychology and

www.apa.org/journals/psp October 2006 VOLUME 91 NUMBER 4

Copyright © 2006 by the American Psychological Association

Attitudes and Social Cognition 583

Accuracy, Error, and Bias in Predictions for Real Versus Hypothetical Events David A. Armor and Aaron M. Sackett

601

Regulatory Fit as Input for Stop Rules Leigh Ann Vaughn, Jill Malik, Sandra Schwartz, Zhivka Petkova, and Lindsay Trudeau

612

See What You Want to See: Motivational Influences on Visual Perception Emily Balcetis and David Dunning

626

Jealousy and the Threatened Self: Getting to the Heart of the Green-Eyed Monster David DeSteno, Piercarlo Valdesolo, and Monica Y. Bartlett

Interpersonal Relations and Group Processes 642

Romantic Involvement Often Reduces Men’s Testosterone Levels—But Not Always: The Moderating Role of Extrapair Sexual Interest Matthew McIntyre, Steven W. Gangestad, Peter B. Gray, Judith Flynn Chapman, Terence C. Burnham, Mary T. O’Rourke, and Randy Thornhill

652

Stereotyping and Evaluation in Implicit Race Bias: Evidence for Independent Constructs and Unique Effects on Behavior David M. Amodio and Patricia G. Devine

662

Regulation Processes in Intimate Relationships: The Role of Ideal Standards Nickola C. Overall, Garth J.O. Fletcher, and Jeffry A. Simpson

686

Procedural Justice and the Hedonic Principle: How Approach Versus Avoidance Motivation Influences the Psychology of Voice Jan-Willem van Prooijen, Johan C. Karremans, and Ilja van Beest

698

The Paradox of Group-Based Guilt: Modes of National Identification, Conflict Vehemence, and Reactions to the In-Group’s Moral Violations Sonia Roccas, Yechiel Klar, and Ido Liviatan

712

Negotiation From a Near and Distant Time Perspective Marlone D. Henderson, Yaacov Trope, and Peter J. Carnevale

(contents continue)

Personality Processes and Individual Differences 730

Psychological Resilience, Positive Emotions, and Successful Adaptation to Stress in Later Life Anthony D. Ong, C. S. Bergeman, Toni L. Bisconti, and Kimberly A. Wallace

750

The Differential Effects of Intrinsic and Identified Motivation on Well-Being and Performance: Prospective, Experimental, and Implicit Approaches to Self-Determination Theory Kimberly D. Burton, John E. Lydon, David U. D’Alessandro, and Richard Koestner

763

A First Large Cohort Study of Personality Trait Stability Over the 40 Years Between Elementary School and Midlife Sarah E. Hampson and Lewis R. Goldberg

780

Relating Emotional Abilities to Social Functioning: A Comparison of Self-Report and Performance Measures of Emotional Intelligence Marc A. Brackett, Susan E. Rivers, Sara Shiffman, Nicole Lerner, and Peter Salovey

Other 697 600 779 749

American Psychological Association Subscription Claims Information E-Mail Notification of Your Latest Issue Online! Instructions to Authors Low Publication Prices for APA Members and Affiliates

ii

ATTITUDES AND SOCIAL COGNITION CHARLES M. JUDD, Editor University of Colorado at Boulder ASSOCIATE EDITORS DACHER KELTNER University of California, Berkeley ANNE MAASS Universita` di Padova, Padova, Italy BERND WITTENBRINK University of Chicago VINCENT YZERBYT Catholic University of Louvain, Louvain-la-Neuve, Belgium CONSULTING EDITORS ICEK AJZEN University of Massachusetts

ALICE H. EAGLY Northwestern University

NIRA LIBERMAN Tel Aviv University, Tel Aviv, Israel

LINDA SKITKA University of Illinois at Chicago

NICHOLAS EPLEY University of Chicago

DIANE M. MACKIE University of California, Santa Barbara

JOHN SKOWRONSKI Northern Illinois University

RUSSELL H. FAZIO Ohio State University

NEIL MACRAE Dartmouth College

ELIOT R. SMITH Indiana University Bloomington

LISA FELDMAN BARRETT Boston College

TONY MANSTEAD Cardiff University, Cardiff, Wales

SUSAN T. FISKE Princeton University

THOMAS MUSSWEILER Universita¨t Ko¨ln, Cologne, Germany

DIEDERIK STAPEL University of Groningen, Groningen, the Netherlands

BARBARA L. FREDRICKSON University of Michigan

JAMES M. OLSON University of Western Ontario, London, Ontario, Canada

WENDI GARDNER Northwestern University

MAHZARIN BANAJI Harvard University

BERNADETTE M. PARK University of Colorado at Boulder

DANIEL GILBERT Harvard University

MONICA BIERNAT University of Kansas

RICHARD E. PETTY Ohio State University

THOMAS GILOVICH Cornell University

IRENE V. BLAIR University of Colorado at Boulder

NEAL J. ROESE University of Illinois at Urbana– Champaign

ANTHONY G. GREENWALD University of Washington

GALEN V. BODENHAUSEN Northwestern University

DAVID L. HAMILTON University of California, Santa Barbara

MARKUS BRAUER LAPSCO, Universite´ Blaise Pascal Clermont-Ferrand, France

EDWARD R. HIRT Indiana University Bloomington

MARILYNN B. BREWER Ohio State University

TIFFANY ITO University of Colorado at Boulder

JOHN T. CACIOPPO University of Chicago

YOSHIHISA KASHIMA University of Melbourne, Victoria, Australia

OLIVIER CORNEILLE Catholic University of Louvain, Louvain-la-Neuve, Belgium

KARLE CHRISTOPHE KLAUER Albrecht-Ludwigs-Universita¨t Freiburg, Freiburg, Germany

PATRICIA DEVINE University of Wisconsin—Madison AP DIJKSTERHUIS University of Amsterdam, Amsterdam, the Netherlands DAVID DUNNING Cornell University

MYRON ROTHBART University of Oregon LAURIE RUDMAN Rutgers, The State University of New Jersey MARK SCHALLER University of British Columbia, Vancouver, British Columbia, Canada TONI SCHMADER University of Arizona NORBERT SCHWARZ University of Michigan

ARIE W. KRUGLANSKI University of Maryland

GU¨N R. SEMIN Free University, Amsterdam, the Netherlands

ALAN LAMBERT Washington University in St. Louis

JEFFREY W. SHERMAN University of California, Davis

JENNIFER LERNER Carnegie Mellon University

STEVEN J. SHERMAN Indiana University Bloomington

FRITZ STRACK Universita¨t Wu¨rzburg, Wu¨rzburg, Germany ABRAHAM TESSER University of Georgia YAACOV TROPE New York University THERESA K. VESCIO Pennsylvania State University WILLIAM VON HIPPEL University of New South Wales, Sydney, Australia DUANE T. WEGENER Purdue University DANIEL M. WEGNER Harvard University DIRK WENTURA Saarland University, Saarbru¨cken, Germany DANIEL WIGBOLDUS Radboud University Nijmegen, Nijmegen, the Netherlands TIMOTHY D. WILSON University of Virginia PIOTR WINKIELMEN University of California, San Diego MARK P. ZANNA University of Waterloo, Waterloo, Ontario, Canada

ASSISTANT TO THE EDITOR—LAURIE HAWKINS

INTERPERSONAL RELATIONS AND GROUP PROCESSES JOHN F. DOVIDIO, Editor University of Connecticut ASSOCIATE EDITORS DAPHNE BLUNT BUGENTAL University of California, Santa Barbara BEVERLEY FEHR University of Winnipeg, Winnipeg, Manitoba, Canada JACQUES-PHILIPPE LEYENS Catholic University of Louvain, Louvain-la-Neuve, Belgium ANTONY MANSTEAD Cardiff University, Cardiff, United Kingdom JEFFRY A. SIMPSON University of Minnesota, Twin Cities Campus

ARTHUR ARON State University of New York at Stony Brook

RUPERT BROWN The University of Kent at Canterbury, Canterbury, England

XIMENA ARRIAGA Purdue University

LORNE CAMPBELL University of Western Ontario, London, Ontario, Canada

WINTON W. T. AU The Chinese University of Hong Kong, Shatin, Hong Kong MARK BALDWIN McGill University, Montreal, Quebec, Canada KIM BARTHOLOMEW Simon Fraser University, Burnaby, British Columbia, Canada C. DANIEL BATSON University of Kansas

SCOTT TINDALE Loyola University Chicago

B. ANNE BETTENCOURT University of Missouri—Columbia

JACQUIE D. VORAUER University of Manitoba, Winnipeg, Manitoba, Canada

GERD BOHNER Universita¨t Bielefeld, Bielefeld, Germany

CONSULTING EDITORS DOMINIC ABRAMS University of Kent at Canterbury, Canterbury, England

NIALL BOLGER Columbia University

CHRIS AGNEW Purdue University

JONATHON D. BROWN University of Washington

NYLA R. BRANSCOMBE University of Kansas

SERENA CHEN University of California, Berkeley MARGARET CLARK Yale University CARSTEN DE DREU University of Amsterdam, Amsterdam, the Netherlands STE´PHANIE DEMOULIN Catholic University of Louvain Louvain-la-Neuve, Belgium, and Belgan National Fund for Scientific Research, Brussels, Belgium

KLAUS FIEDLER University of Heidelberg, Heidelberg, Germany GARTH FLETCHER University of Canterbury, Christchurch, New Zealand SHELLY GABLE University of California, Los Angeles LOWELL GAERTNER University of Tennessee, Knoxville SAMUEL L. GAERTNER University of Delaware ADAM GALINSKY Northwestern University PETER GLICK Lawrence University STEPHANIE A. GOODWIN Purdue University

DAVID DESTENO Northeastern University

MARTIE G. HASSELTON University of California, Los Angeles

STEVE DRIGOTAS Johns Hopkins University

S. ALEXANDER HASLAM University of Exeter, Exeter, United Kingdom

ELISSA S. EPEL University of California, San Francisco VICTORIA ESSES University of Western Ontario, London, Ontario, Canada

(editors continue)

VERLIN HINSZ North Dakota State University GORDON HODSON Brock University, St. Catherine’s, Ontario, Canada

MICHAEL A. HOGG University of Queensland, Brisbane, Australia

LAURA J. KRAY University of California, Berkeley

ANDREA B. HOLLINGSHEAD University of Southern California JOHN G. HOLMES University of Waterloo, Waterloo, Ontario, Canada RICK H. HOYLE University of Kentucky

JAMES R. LARSON JR. University of Illinois at Chicago COLIN WAYNE LEACH University of Sussex, Sussex, United Kingdom JOHN LEVINE University of Pittsburgh JOHN E. LYDON McGill University, Montreal, Quebec, Canada

JOLANDA JETTEN University of Exeter, Exeter, United Kingdom

JON K. MANER Florida State University

JAMES D. JOHNSON University of North Carolina at Wilmington TATSUYA KAMEDA Hokkaido University, Sapporo, Japan BENJAMIN R. KARNEY RAND Corporation, Santa Monica, California YOSHI KASHIMA University of Melbourne, Victoria, Australia

BRENDA MAJOR University of California, Santa Barbara CRAIG MCGARTY Australian National University, Canberra, Australia WENDY BERRY MENDES Harvard University RICHARD MORELAND University of Pittsburgh

DEBORAH A. KASHY Michigan State University

SABINE OTTEN University of Gro¨ningen, Gro¨ningen, the Netherlands CRAIG D. PARKS Washington State University LOUIS A. PENNER Wayne State University PAULA PIETROMONACO University of Massachusetts at Amherst

CHRISTINE SMITH Grand Valley State University HEATHER J. SMITH Sonoma State University RUSSELL SPEARS Cardiff University, Cardiff, Wales CHARLES STANGOR University of Maryland GARY L. STASSER Miami University—Ohio

TOM POSTMES University of Exeter, Exeter, United Kingdom

WALTER STEPHAN New Mexico State University

FELICIA PRATTO University of Connecticut

WILLIAM B. SWANN JR. University of Texas at Austin

HARRY T. REIS University of Rochester

JANET SWIM Pennsylvania State University

W. STEVEN RHOLES Texas A&M University

LEIGH L. THOMPSON Northwestern University

JENNIFER A. RICHESON Northwestern University

TOM TYLER New York University

MARK SCHALLER University of British Columbia, Vancouver, British Columbia, Canada

JEROEN VAES University of Padova, Padova, Italy

BRIAN MULLEN KERRY KAWAKAMI University of Kent at Canterbury, York University, Toronto, Ontario, Canada Canterbury, England JANICE R. KELLY AME´LIE MUMMENDEY Purdue University Friedrich-Schiller-Universita¨t, Jena, DACHER KELTNER Jena, Germany University of California, Berkeley MARK MURAVEN DAVID A. KENNY University at Albany, State University University of Connecticut of New York

DAVID A. SCHROEDER University of Arkansas

KEES VAN DEN BOS University of Utrecht, Utrecht, the Netherlands

CONSTANTINE SEDIKIDES University of Southampton, Southampton, England

PAUL A. M. VAN LANGE Free University, Amsterdam, Amsterdam, the Netherlands

PHILLIP R. SHAVER University of California, Davis

LAURIE R. WEINGART Carnegie Mellon University

J. NICOLE SHELTON Princeton University

GWEN M. WITTENBAUM Michigan State University

DOUGLAS T. KENRICK Arizona State University

SANDRA L. MURRAY State University of New York at Buffalo

MARGARET SHIH University of Michigan

NORBERT L. KERR Michigan State University

STACEY SINCLAIR LISA A. NEFF University of Virginia University of Toledo ASSISTANT TO THE EDITOR—CHRISTINE KELLY

WENDY L. WOOD Texas A&M University MICHAEL ZA´RATE University of Texas at El Paso

PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES CHARLES S. CARVER, Editor University of Miami ASSOCIATE EDITORS TIM KASSER Knox College

GEORGE A. BONANNO Teachers College, Columbia University

AVSHALOM CASPI MARIO MIKULINCER Bar-Ilan University, Ramat-Gan, Israel King’s College, London EDWARD C. CHANG EVA M. POMERANTZ University of Michigan University of Illinois at Urbana– Champaign RICHARD W. ROBINS University of California, Davis GERARD SAUCIER University of Oregon THOMAS A. WIDIGER University of Kentucky

SERENA CHEN University of California, Berkeley A. TIMOTHY CHURCH Washington State University JAMES COAN University of Wisconsin—Madison M. LYNNE COOPER University of Missouri—Columbia

EDDIE HARMON-JONES Texas A&M University

DANIEL W. RUSSELL Iowa State University

TODD HEATHERTON Dartmouth College

OLIVER C. SCHULTHEISS University of Michigan

JUTTA HECKHAUSEN University of California, Irvine

SUZANNE C. SEGERSTROM University of Kentucky

STEVEN J. HEINE University of British Columbia, Vancouver, British Columbia, Canada

KENNON M. SHELDON University of Missouri—Columbia

RICHARD KOESTNER McGill University Montreal, Quebec, Canada

C. R. SNYDER University of Kansas SANJAY SRIVASTAVA University of Oregon

DAVID LUBINSKI Vanderbilt University

TIMOTHY STRAUMAN Duke University

MICHAEL EID University of Geneva, Geneva, Switzerland

RICHARD E. LUCAS Michigan State University

MICHAEL J. STRUBE Washington University

ROBERT R. MCCRAE National Institute on Aging, Baltimore

JERRY SULS University of Iowa

ANDREW J. ELLIOT University of Rochester

WENDY BERRY MENDES Harvard University

WILLIAM B. SWANN JR. University of Texas at Austin

LISA FELDMAN BARRETT Boston College

RODOLFO MENDOZA-DENTON University of California, Berkeley

HOWARD TENNEN University of Connecticut Health Center

WILLIAM FLEESON Wake Forest University

DANIEL K. MROCZEK Fordham University

MICHAEL C. ASHTON Brock University, St. Catherines, Ontario, Canada

SUZANNE THOMPSON Pomona College

R. CHRIS FRALEY University of Illinois at Chicago

STEPHEN A. PETRILL Pennsylvania State University

OZLEM AYDUK University of California, Berkeley

ANTONIO L. FREITAS State University of New York at Stony Brook

RALPH L. PIEDMONT Loyola College in Maryland

ROBERT J. VALLERAND Universite´ du Que´bec a` Montre´al Montreal, Quebec, Canada

CONSULTING EDITORS STEPHAN A. AHADI American Institutes for Research, Washington, DC JAMIE ARNDT University of Missouri—Columbia JENS B. ASENDORPF Humboldt-Universita¨t Berlin Berlin, Germany

E. ASHBY PLANT Florida State University

ROY F. BAUMEISTER Florida State University VERO´NICA BENET-MARTI´NEZ University of California, Riverside

DAVID C. FUNDER University of California, Riverside STEVEN W. GANGESTAD University of New Mexico

BRENT ROBERTS University of Illinois at Urbana–Champaign

APRIL L. BLESKE-RECHEK University of Wisconsin—Eau Claire

CAROL L. GOHM University of Mississippi

MICHAEL D. ROBINSON North Dakota State University

ASSISTANT TO THE EDITOR—JESSICA LILLESAND

KATHLEEN D. VOHS University of Minnesota DAVID WATSON University of Iowa BARBARA WOIKE Columbia University REX A. WRIGHT University of Alabama at Birmingham

ATTITUDES AND SOCIAL COGNITION

Accuracy, Error, and Bias in Predictions for Real Versus Hypothetical Events David A. Armor and Aaron M. Sackett Yale University Participants made predictions about performance on tasks that they did or did not expect to complete. In three experiments, participants in task-unexpected conditions were unrealistically optimistic: They overestimated how well they would perform, often by a large margin, and their predictions were not correlated with their performance. By contrast, participants assigned to task-expected conditions made predictions that were not only less optimistic but strikingly accurate. Consistent with predictions from construal level theory, data from a fourth experiment suggest that it is the uncertainty associated with hypothetical tasks, and not a lack of cognitive processing, that frees people to make optimistic prediction errors. Unrealistic optimism, when it occurs, may be truly unrealistic; however, it may be less ubiquitous than has been previously suggested. Keywords: optimism, optimistic bias, prediction, accuracy, construal level theory

overestimate how well they will do on exams (Shepperd, Ouellette, & Fernandez, 1996) and to underestimate how long it will take for them to complete their assignments (Buehler, Griffin, & Ross, 1994). Undergraduate and professional (Master of Business Administration) students tend to overestimate their prospects for success on the job market (Hoch, 1985; Shepperd et al., 1996), and even gainfully employed professional financial analysts tend to overestimate corporate earnings (Calderon, 1993; Lim, 2001). A great deal of research has also shown that people of all ages and backgrounds tend to overestimate how likely they are to experience a wide variety of positive outcomes, and to underestimate how likely they are to experience an even wider variety of negative outcomes relative to other people (e.g., Weinstein, 1980, 1987; see also Perloff & Fetzer, 1986; Quadrel, Fischhoff, & Davis, 1993; cf. Heine & Lehman, 1995; for a review, see Helweg-Larsen & Shepperd, 2001). Despite the prevalence of these demonstrations, which give the impression that optimistic biases may be nearly unavoidable (see, e.g., Weinstein & Klein, 1995), there is a growing body of research suggesting that this portrait of unwavering optimism may be overly simplified (see Armor & Taylor, 1998, 2002; Shepperd, Sweeny, & Carroll, 2006). There are at least two reasons for this. First, research that provides a clear criterion for evaluating prediction accuracy (e.g., by comparing predictions to attained outcomes) has shown that people’s predictions often may be optimis-

How accurate are people’s visions of the future? The general consensus, which may be gleaned from even a brief perusal of research on personal forecasts, is that people are not very accurate at all. To date, several hundred studies have shown that people’s predictions tend to be excessively and unrealistically optimistic (Weinstein, 1998; for reviews, see Armor & Taylor, 1998, 2002; Buehler, Griffin, & Ross, 2002; Helweg-Larsen & Shepperd, 2001). To be sure, some degree of error in personal prediction is to be expected. As Yogi Berra famously, if mythically, aphorized, “It’s tough to make predictions, especially about the future.”1 However, people’s predictions appear to be prone not only to error—random deviations and distortions that would be expected to balance out over the long run— but to consistent and pervasive bias. The common conclusion from studies of personal forecasts is that people expect that their futures will be more pleasant and less painful than they have any right to expect them to be. Examples of optimistic biases in personal predictions are not difficult to find, whether in real life or in the relatively artificial confines of the research laboratory. Students, for example, tend to

David A. Armor and Aaron M. Sackett, Department of Psychology, Yale University. Aaron M. Sackett is now at the Graduate School of Business, University of Chicago. We thank Melissa Blakeley, Aimee Keen, and Leslie Ono for assistance with data collection, Clayton Critcher and Han-Ya Hsu for help with data coding, and Gregory Miller for suggesting the lyric with which we close the article. Correspondence concerning this article should be addressed to David A. Armor, who is now at the Department of Psychology, San Diego State University, San Diego, CA 92182-4611. E-mail: [email protected]

1 Although commonly attributed to Berra, this statement may be more a part of the Hall of Fame catcher’s legend than of his loquacious history; Berra himself admitted “I really didn’t say everything I said” (Berra, 1998). Near paraphrases have also been attributed, with comparable frequency, to Nobel Laureate Niels Bohr.

Journal of Personality and Social Psychology, 2006, Vol. 91, No. 4, 583– 600 Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.583

583

584

ARMOR AND SACKETT

tically biased without being entirely unrealistic. Research on the planning fallacy, for example, has shown that whereas people’s task completion estimates tend to be overly optimistic (in that people expect to be able to do more in less time than they are typically able to do), these same estimates tend to be quite highly correlated with actual task completion times (correlations between predicted and actual completion times in the initial studies of the planning fallacy ranged from .36 to .81; see Buehler et al., 1994). Thus, although biased, these optimistic predictions do not appear to be completely divorced from reality. One of the goals of the present research, then, is to provide a clear criterion against which the accuracy of individual predictions can be assessed. Second, people do not appear to be indiscriminately optimistic but to be sensitive to the context in which they make their predictions (Armor & Taylor, 1998, 2002; Shepperd et al., 2006). There is an accumulating body of research suggesting that people may be more likely to be overly optimistic in situations in which it is less likely that the accuracy of their predictions will be tested or challenged or in situations in which the consequences of being inaccurate are expected to be less severe. For example, whereas students tend to overestimate how well they will do on their exams, this optimism is most prominent (and most unrealistic) when those exams are some time away; optimistic biases appear to be much less pronounced as the “moment of truth” draws near (Gilovich, Kerr, & Medvec, 1993; Shepperd et al., 1996, Study 2). Similarly, whereas several studies have found that students overestimate how much they will earn in their first job, students tend to see their job prospects in more modest terms—and, indeed, accurately estimate their starting salary—as they approach the market for real (Shepperd et al., 1996, Study 1). Even Calderon’s (1993) analysis of professional analysts’ financial predictions revealed a systematic decline in optimistic bias as a function of the proximity of the forecast date to the realization date (although optimistic bias was evident even among the most proximal predictions). On the basis of results such as these, Armor and Taylor (1998, 2002) have argued that people are not indiscriminately optimistic, but are “situated optimists” and that the expression of optimistic biases is not invariant within persons but rather is largely dependent on the social and psychological context in which people find themselves.2 The notion that optimistic biases are not inherent within persons but situated within contexts raises a number of important theoretical questions—namely, when and why do people tend to be unrealistically optimistic?— but also raises the question of whether the prevalence of unrealistic optimism may be over- or underestimated as a function of the experimental setting in which unrealistic optimism is typically studied (for similar arguments, see Gilovich et al., 1993, p. 559). In other words, it is not clear whether the seemingly common expressions of unrealistic optimism reflect a general tendency for people to be optimistic or an equally general tendency for psychologists to measure optimism in settings in which unrealistic optimism may be especially likely to be expressed.

The Social Psychology of the Unrealistic Optimism Experiment In the early 1960s, Orne (1962) published an influential critique suggesting that psychologists—social psychologists in particular—should pay careful attention to the psychology experiment as

a “social situation” and that psychologists should be mindful of the effects that these situations have, above and beyond any effects of the experimental manipulations, on the thoughts, feelings, and observable behaviors of research participants. Tetlock’s critique of decontextualized research on judgmental errors and biases makes similar arguments (e.g., Tetlock, 1992; Tetlock & Lerner, 1999; see also Schwarz, 1994). Although a tremendous number of studies have shown that people tend to be excessively optimistic, a close look at the contexts in which these data have been collected reveal that the overwhelming majority of these studies have been conducted in a manner that may be especially conducive to the expression of unrealistic optimism. In most studies, accuracy incentives are either minimal or absent altogether, as participants are often asked to make predictions anonymously and without concern about possible consequences of inaccurate predictions. In many cases the outcome of participants’ predictions simply cannot be known— by either the researcher or the person making the prediction—and, as a consequence, prediction accuracy itself is often impossible to verify. In the modal study, for example, young, healthy college students might be asked of their chances, relative to the chances of other young, healthy college students, of one day succumbing to cancer, of needing dentures, or of being fired from a job for which they have yet to be hired. Although these may all be meaningful questions, and although arguments can be made about their relevance to respondents’ current behavior (e.g., decisions to smoke, eat sweets, or to develop the “seven coveted habits of highly effective employees”), these outcomes are also likely to be quite far removed from most participants’ current concerns. To the extent that research participants are asked to estimate the probability of hypothetical outcomes or outcomes that are unimaginably distant, it seems possible that they may be aware of the fundamentally unverifiable nature of their predictions and that they may, knowingly or unknowingly, alter their predictions accordingly. Of course, not all studies of unrealistic optimism have involved predictions for entirely hypothetical outcomes. Research on the planning fallacy, for example, has asked participants to make predictions about tasks they knew they would be completing (see Buehler et al., 2002, for review), and at least one study (Camerer & Lovallo, 1999) has found evidence of unrealistic optimism even when participants could (and did) lose money as a function of their overly optimistic self-assessments. It is not entirely clear, however, to what extent participants in these studies may have been even more optimistic—and potentially more unrealistically optimistic— had predictions been made without the expectation that predictions would be verified.

2 The notion that optimism is situated does not deny that there are meaningful individual differences in people’s tendency to be optimistic (see, e.g., Norem & Cantor, 1986; Scheier, Carver, & Bridges, 1994). However, studies examining the relationship between optimism as a general trait and specific instantiations of optimism in individual predictions reveal little correspondence between the two (Armor & Taylor, 1998; Buehler & Griffin, 2003). Our focus, therefore, is on outcome-specific expectations and not more generalized expectancies captured by measures of dispositional optimism.

REAL VS. HYPOTHETICAL

Predictions for Real Versus Hypothetical Events The present studies examine the accuracy of predictions people make for tasks that they do or do not expect to complete. Several lines of research seemed relevant, a priori, as sources of hypotheses, including Gilovich et al.’s (1993) research on the effects of temporal distance on subjective confidence; Trope and Liberman’s (e.g., 2003; Liberman & Trope, 1998) research on construal level theory; Sedikides, Herbst, Hardin, and Dardis’s (2002) work on accountability and self-evaluation; and Shepperd et al.’s (1996) research on the effects of expecting proximate feedback on pessimistic “bracing for the worst.” In Gilovich et al.’s (1993) experiments, people were found to predict, optimistically, that they would perform well above average on tasks that were described as being temporally distant (i.e., when the opportunity to complete these tasks was weeks or months away), but the extent to which people showed this effect of predicting “better than average” performance was markedly reduced when the tasks were described as being temporally proximate (i.e., when the opportunity to complete these tasks was only moments away). Although Gilovich et al. (1993) emphasized that their hypotheses and data pertained to temporal distance, and not to the real versus hypothetical status of the event in question (see, e.g., Gilovich et al., 1993, p. 552), there is reason to suspect that the two variables may have similar effects on people’s predictions. According to Trope and Liberman’s (2003) construal level theory, a variety of manipulations—including temporal distance and the real versus hypothetical status of events—may be functionally interchangeable as manipulations of “psychological distance.” Insofar as hypothetical tasks may be seen as being infinitely temporally distant, we hypothesize that people will have considerably greater leeway to make overly optimistic predictions when the task in question is hypothetical as opposed to real. Two other lines of work suggest that, regardless of psychological distance, the possibility of receiving evaluative feedback may be the critical independent variable (Sedikides et al., 2002; Shepperd et al., 1996; see also Shepperd, Grace, Cole, & Klein, 2005; Taylor & Shepperd, 1998). The potential for feedback—whether real or imagined, public or private, externally provided or internally derived—is one of the characteristics that differentiate real tasks from hypothetical ones. When tasks are real, one can (and often will) learn how well one has performed on that task, but when a task is hypothetical, opportunities for these kinds of reality checks are largely absent. However, studies demonstrating the effects of feedback expectations, which generally show a reduction of optimism, have focused largely on assessments made after participants had completed a critical task (e.g., estimates of how well one has performed on a task that one has just completed; Sedikides et al., 2002; Shepperd et al., 1996, 2005; see also Sniezek, Paese, & Switzer, 1990). Studies that have examined the effect of manipulated feedback expectations on pretask predictions have yielded inconsistent results, sometimes showing no effect (e.g., Study 2 of Buehler et al., 1994) and sometimes showing a reduction in optimism (e.g., Taylor & Shepperd, 1998). It is thus not clear that the expectation of feedback is a necessary precondition for a reduction in pretask optimism. In the studies reported here, we examine the degree of accuracy, error, and bias in predictions people make about their performance on tasks that they believe are either real (i.e., tasks that they will

585

soon be completing) or hypothetical (i.e., tasks that they do not think they will complete). We hypothesize that optimistic bias will be more pronounced in conditions in which participants do not expect that their predictions will be put to the test and that their predictions will be less biased—and more accurate—in conditions in which they do. Studies 1 and 2 set the empirical groundwork for subsequent theoretical development and evaluation, establishing whether, in essence, the hypothesized differences in predictions and in prediction accuracy obtain and, thus, need to be explained. In Study 3, we next evaluated the necessity of critical feedback in the real versus hypothetical effects (cf. Sedikides et al., 2002; Shepperd et al., 1996). In Studies 3 and 4, we also tested a number of explanatory hypotheses suggested by work on temporal proximity (e.g., preemptive self-criticism, Gilovich et al., 1993; mood as information, Savitsky, Medvec, Charlton, & Gilovich, 1998; Shepperd et al., 2005) as well as a number of previously untested hypotheses derived from Trope and Liberman’s (2003) construal level theory. We provide further detail about these hypotheses in the context of the studies designed to test them.

A Note on Accuracy The present studies share a concern with evaluating the accuracy of individual predictions. Optimistic biases have been most commonly investigated with relative appraisals, that is, assessments of the extent to which people see themselves as more likely than others to experience positive events and less likely than others to experience negative events (e.g., Perloff & Fetzer, 1986; Weinstein, 1980). There are, however, several drawbacks to using relative appraisals as indicators of prediction accuracy. First, these measures reveal information about bias only in the aggregate and cannot, on their own, reveal whether any one prediction is accurate (e.g., if a majority within a group expects to perform better than the average performing member of that group, many of them will be right). Second, several recent lines of research suggest that the appearance of optimistic bias in relative appraisals may reveal more about respondents’ egocentrism than about optimism or pessimism per se. For example, whereas people appear to be overly optimistic when stating relative chances for success on easy tasks, they appear to be overly pessimistic when stating relative chances for success on difficult tasks (Chambers & Windschitl, 2004; Kruger, 1999; Kruger & Burrus, 2004). Other anomalies— such as the observation that optimistic biases appear to be reduced in Eastern cultures when assessed by relative appraisal methods (e.g., Heine & Lehman, 1995) but robust when predictions are compared with obtained outcomes (e.g., Buehler, Griffin, Otsubu, Lehman, & Heine, 2000; see Buehler et al., 2002, for discussion)—further suggest that measures of relative appraisals may be assessing something more (or something less) than people’s tendency to be optimistic. In the context of evaluating potential moderators of optimistic biases, the ability to assess individual prediction accuracy is critical. In Gilovich et al.’s (1993) studies of the effects of temporal proximity, for example, it is certainly possible that the observed reduction in the better-than-average effect reflected greater accuracy among individual participants in the temporally proximate conditions (an interpretation we are inclined to favor), but it is also possible that these participants could have become mindlessly middling, cautiously responding near the midpoint of the available

ARMOR AND SACKETT

586

response scale without becoming any more accurate in their predictions, or that some participants may have become overly pessimistic while others remained excessively optimistic (thus appearing average only in aggregate). A reduction in bias, therefore, does not necessarily imply a reduction in error (see, e.g., Buehler et al., 1994). In the studies presented here, we assessed predictions in a variety of ways, including relative appraisals, but in Studies 1 through 3 we also obtained an unambiguous criterion (e.g., test performance) against which prediction accuracy could be assessed. With this criterion, deviations from accuracy could be evaluated in a number of useful but different ways. Evidence of bias was revealed by signed differences between predictions and performance. Evidence of prediction error was revealed by the extent of absolute deviation between prediction and performance. As an additional measure of prediction accuracy, we also assessed the correlation between predictions and performance.

Study 1: The Accidental Scavenger In Study 1, participants were asked to estimate how well they would perform on a novel and ultimately highly involving activity: a scavenger hunt. Half of the participants were informed at the outset that they would be asked to complete the hunt; the remaining participants were not informed that they would complete the hunt until after they had made performance predictions. To provide a basis for comparing predictions with performance (i.e., a criterion for accuracy), all participants were next asked to complete the scavenger hunt, and their performance was recorded. We hypothesized that participants who did not expect to complete the scavenger hunt would exhibit excessive optimism, as is so commonly demonstrated in studies of people’s personal forecasts, but that participants who were led to expect that they would complete the hunt would make more modest—and more accurate—predictions.

Method Participants Participants were 38 university undergraduates (58% male) who participated in exchange for research participation credit. One additional participant failed to respond to a majority of our dependent measures and was not included in the final sample.

found, with varying degrees of effort and ingenuity, on the university campus. A number of simple rules (e.g., one must hunt alone; no item may be stolen or purchased for the purpose of the hunt, etc.) were included to give all participants a common frame of reference. Predictions and task evaluations. In order to bolster the taskexpectation manipulation, the phrasing of the task assessment and prediction measures was altered to correspond to the experimental condition: Participants who were led to expect that they would complete the scavenger hunt were asked to evaluate the hunt and to make performance predictions as if it were an impending reality (e.g., “How well will you do on the scavenger hunt?”), whereas participants who were not led to expect that they would complete the scavenger hunt were asked to evaluate the hunt and to make predictions as if it were hypothetical (e.g., “how well would you do, if you were asked to complete the scavenger hunt?”). Performance predictions were obtained both in general terms, by asking participants how well they thought they would [will] do on the scavenger hunt, on a scale ranging from 1 (extremely poorly) to 7 (extremely well), and in very specific terms by asking participants how many of the 32 items listed they thought they would [will] find in the allotted time. In order to assess relative appraisals, participants were also asked to estimate how well the “typical student” at their university would perform on the scavenger hunt (as with the personal predictions, predictions of others were requested in both general and specific terms); these estimates could then be compared with self-estimates in order to derive relative evaluations. An additional specific assessment asked participants to estimate how long it would take them to find half of the items on the list (in minutes) if they did not have a time limit. For exploratory purposes, we asked participants to make several additional general assessments, including: how much fun participants thought the scavenger hunt would [will] be, how much they thought they would [will] enjoy participating in it, how difficult they thought the scavenger hunt task would [will] be, and how good they were at this sort of game. Responses to all general assessment questions were made on 7-point scales. Performance. Once participants completed the dependent measures, they were asked to complete the scavenger hunt. Participants were equipped with a stopwatch so that they could keep track of elapsed time and a shopping bag in which to carry scavenged items. They were reminded of the rules and of the 30-min time limit and were then prompted to begin at the experimenter’s cue (the experimenter also started a stopwatch at this time in order to have an independent assessment of how long participants took to complete the hunt). At the end of the scavenger hunt, participants returned to a different experiment room and were greeted by a different experimenter who was blind to the participant’s task-expectancy condition and performance predictions (but not to elapsed time). This experimenter administered a brief performance assessment questionnaire and tallied the number of items collected.

Results

Procedure In order to protect against social facilitation effects and outright cooperation on the scavenger hunt task, participants were tested individually. Manipulating task-completion expectations. We manipulated participants’ expectations of whether they would be asked to complete the scavenger hunt by altering the manner in which the scavenger hunt was introduced. Half of our participants were told, from the start, that their task as participants would be to evaluate a scavenger hunt and then to complete the hunt that they evaluated. Remaining participants were told that they would be asked to evaluate the scavenger hunt, but no mention of actual performance was made. The scavenger hunt. The scavenger hunt task was similar to one used in prior research (for details, see Armor & Taylor, 2003). The hunt was described as a 30-min activity in which participating “scavengers” would be required to find as many items as they could from a list of 32 objects (e.g., a safety pin, a party invitation, a library book). All items could be

Preliminary analyses revealed no interaction effects with participant gender in this or any subsequent study. Thus, this variable is not discussed further.

Performance Predictions As hypothesized, the task-expectation manipulation had a considerable impact on performance predictions, with participants in the task-expected condition being consistently less optimistic than participants in the task-unexpected condition. As can be seen in Table 1, in which means and corresponding inferential statistics are presented, the effects of the task-expectancy manipulation tended to be more pronounced when the performance predictions were assessed with the use of specific, unambiguous (and therefore

REAL VS. HYPOTHETICAL

587

Table 1 Performance Predictions as a Function of Whether Participants Expected to Complete the Scavenger Hunt (Study 1) Task completion expectation Measure General/global predictions Predicted performance (self) Predicted performance (others) Self–other difference Specific predictions Predicted number of items found (self) Predicted number of items found (others) Self–other difference Time to find half of items on list Other measures Task difficulty Personal efficacy Anticipated enjoymenta

Not expected

Expected

F(1, 36)

4.21 4.32 ⫺0.11

3.58 4.16 ⫺0.58

1.40 0.18 1.16

16.68 16.42 0.26 40.74

10.32 12.74 ⫺2.42 76.73

12.58** 4.37* 4.17* 7.12*

4.63 4.68 3.97

5.05 3.84 4.00

1.06 2.87† 0.00

a

Responses to questions assessing anticipated fun and anticipated enjoyment were almost perfectly correlated, r(38) ⫽ .97, and therefore were averaged prior to analysis. † p ⬍ .10. * p ⬍ .05. ** p ⬍ .01.

potentially verifiable) response options. Although responses to the more global, ambiguously defined prediction and task-assessment questions were in the hypothesized direction, these differences rarely approached statistical significance. By contrast, the effect of task expectancy on responses to the more specific, objectively scaled prediction measures were clear-cut: Participants who were led to expect that they would complete the hunt thought that they would find considerably fewer items ( p ⫽ .001), that other scavengers would also find fewer items ( p ⫽ .044), and that they would require more time to find half of the items on the list ( p ⫽ .011) than did participants who did not expect to complete the scavenger hunt. It is worth noting that neither group displayed the often demonstrated better-than-average effect. Indeed, participants in the task-expected condition thought that they would perform worse, on average, than the average participant. This apparent overpessimism effect was nearly significant when comparing participants’ predictions for self and other on the general measures of predicted performance, t(18) ⫽ 2.00, p ⫽ .061, and clearly significant when comparing the number of items participants expected that they and others would find, t(18) ⫽ 2.60, p ⫽ .018. Even those who did not expect to complete the hunt, who generally appeared to be more optimistic, did not expect to outperform the average participant, both ts(18) ⬍ 1.

Performance Performance on the scavenger hunt was assessed by tallying the number of items each participant returned. Despite the fact that participants came to the task with markedly different performance predictions, participants in the two experimental conditions did not differ in terms of performance: Those in the task-unexpected condition returned with 10.8 items, on average, whereas those in the task-expected condition returned with 11.2 items, t(36) ⬍ 1.0, ns.3

Accuracy, Error, and Bias The principal analyses of these data involved comparing predictions with performance. As hypothesized, evidence of prediction accuracy, error, and bias differed markedly as a function of participants’ experimentally induced expectations of whether they would complete the scavenger hunt (see Table 2). Participants who did not expect to complete the scavenger hunt were considerably, and unrealistically, optimistic, expecting to find, on average, over 50% more items (M ⫽ 16.7) than they actually found (M ⫽ 10.8), t(18) ⫽ 4.25, p ⬍ .001. The magnitude of this optimistic bias was significantly reduced in the task-expected condition, t(36) ⫽ 4.24, p ⬍ .001. In fact, participants who did expect to complete the hunt expected to perform somewhat worse (M ⫽ 10.3) than they actually performed (M ⫽ 11.2), though this apparent over-pessimism effect was not significant, t(18) ⫽ 1.09, ns. Thus, participants in this study did overestimate how well they would perform, displaying excessive optimism, but only if they did not expect to complete the scavenger hunt. Results from several additional analyses suggest that participants in the task-expected conditions were not only less biased but were also more accurate (see Table 2). For example, the standard deviation of the prediction–performance discrepancies was significantly smaller in the task-expected condition than in the task3 A substantial proportion of participants (53%) failed to complete the scavenger hunt within the 30-min time limit, though over three quarters (76%) returned within 1 min of the deadline. Only three scavengers (8%) returned more than 5 min late (two from the task-unexpected condition, one from the task-expected condition). Although neither the tendency to return late nor the overall time spent on the scavenger hunt differed significantly between conditions, we did conduct additional analyses that adjusted the number of items returned to correct for advantages of coming in late (cf. Armor & Taylor, 2003). These analyses yielded results that were substantively similar to the uncorrected return numbers. Thus, we report analyses of raw performance scores; one exception is noted in the text.

ARMOR AND SACKETT

588

Table 2 Accuracy, Error, and Bias in Performance Predictions as a Function of Whether Participants Expected to Complete the Scavenger Hunt (Study 1) Task completion expectation Comparison of predictions to performance

Not expected

Prediction–performance discrepancy (bias) M (SD) Absolute deviation (error) M (SD) Prediction–performance correlation Pearson’s r

5.84*** (5.99) 6.47*** (5.26) .44

Expected ⫺0.84 (3.37) 2.42*** (2.43) .76***

*** p ⬍ .001 (significantly different from zero).

unexpected condition, as determined by a Levene test, F ⫽ 7.61, p ⫽ .009, indicating smaller average discrepancies in this condition. A more direct assessment of absolute error, calculated simply as the absolute value of the difference between predictions and performance, was significantly smaller for the task-expected group than it was for the task-unexpected group, t(36) ⫽ 3.05, p ⫽ .005. Another way of looking at the accuracy of participants’ predictions is to examine the correlation between predictions and performance. Among those who expected to complete the scavenger hunt, this correlation was substantial and significant (r ⫽ .76, p ⬍ .001). Among those who did not expect to complete the scavenger hunt, the correlation between predictions and performance was more modest (r ⫽ .44; p ⫽ .059), although even this seemingly respectable correlation appears to have been inflated by a few participants who took considerably longer than the time allowed to complete their hunts (see Footnote 3). If one controls for the amount of time participants took to complete the hunt, the resulting partial correlations between predictions and performance were .79 ( p ⬍ .001) in the task-expected condition and .33 ( p ⫽ .19) in the task-unexpected condition. These partial correlations are significantly different from one another, z ⫽ 2.06, p ⫽ .039.4 It appears, then, that the predictions of participants in the task-unexpected condition were not only optimistically biased but also truly unrealistic, insofar as these predictions bore very little relation to subsequent performance, whereas predictions made by participants in the task-expected condition were considerably more accurate.

Discussion The results of Study 1 are consistent with the hypothesis that optimistic biases are not invariant within persons, but situated within contexts, and that these biases are more strongly expressed in situations in which they are less likely to be challenged or tested (Armor & Taylor, 1998, 2002). The results of Study 1 are also broadly consistent with the results of Gilovich et al.’s (1993) studies of temporal proximity but show that (a) the reduction of optimistic biases can occur for real (as opposed to hypothetical) tasks just as for proximate (as opposed to distant) ones and that (b) this reduction in bias in not an artifact of the method of measurement (i.e., relative appraisals). In fact, the measure of relative appraisals used in Study 1 revealed a seemingly anomalous finding: Had relative appraisals been used as a sole indicator of our

participants’ prediction accuracy, participants who were not expecting to complete the task would have appeared to have been more accurate, not less (they expected, on average, to perform about as well as the average participant), whereas participants who did expect to complete the task appeared to have been excessively pessimistic (having expected to perform worse than average) despite having quite accurately predicted their own performance. Although these results were not expected, they are not entirely unprecedented: Savitsky et al. (1998) obtained similar worse-thanaverage effects in conditions of temporal proximity, and expectations of average performance in conditions of temporal distance, in a study designed to closely replicate the procedures of Gilovich et al. (1993). The relationship between different indicators of prediction accuracy were examined further in Study 2. The scavenger hunt task used in Study 1 has a number of advantages (e.g., participants found it to be novel and engrossing), but it also has a number of disadvantages that may limit the generalizability of the results that we obtained. First, although participants appeared to be quite engaged with the task (many returned at full sprint, sweaty and out of breath), the scavenger hunt was ultimately just a game, and some may have found it to be trivial or unimportant. Second, participants in the task-expected condition may have achieved greater accuracy by applying a stop rule, that is, by finishing their hunt once they found the number of items they predicted they would find. (Several pieces of data argue against this possibility, however: Participants in the task-expected condition did not finish their hunts earlier than participants in the task-unexpected condition, and they did not find fewer items despite making more modest predictions.) Third, not all participants finished the scavenger hunt on time, which may have influenced the results (although corrections described in Footnote 3 suggest that this was not the case). To address these potential limitations in Study 2, we used a performance task that was familiar, important, and less vulnerable to a stop-rule strategy, in 4

If one simply excludes data from participants who returned more than 5 min late, the prediction–performance correlation in the task-unexpected condition drops to .35 ( p ⫽ .18); excluding data from participants who returned more than 1 min late drops the correlation to .19 ( p ⫽ .53). Removal of similarly tardy scavengers from the task-expected group leaves the correlations virtually unchanged (rs ⫽ .76 and .80, ps ⬍ .001).

REAL VS. HYPOTHETICAL

addition to being a task for which the time limit could be more strictly enforced.

Study 2: Pop Quiz In Study 2, participants were asked to evaluate (and then to complete) a specially crafted test of questions that were described as being drawn from future versions of the Graduate Record Examination (GRE; Educational Testing Service, 2005). To ensure that this test would be important and familiar to our study participants, we recruited a sample of individuals who were both expert in taking standardized academic tests and invested in doing well on them (Yale University undergraduates).

Method Participants Seventy-five undergraduates (49% women) who were enrolled in an introductory psychology course at Yale University participated in the study in exchange for research participation credit.

Procedure Participants were tested individually and in small groups of no more than three. In order to heighten the perceived importance of the GRE test, participants were told that the study was part of a joint project between the university psychology department and the Educational Testing Service (ETS) and that the experimenters would be asking the participants to evaluate a set of questions that the ETS was considering for use in future versions of the GRE. Manipulating task-completion expectations. As in Study 1, half of the participants were informed upfront that they would be asked to complete the set of GRE questions after answering an initial set of questions about it; remaining participants were told that they would be asked to respond to questions about the test, but no mention was made that anyone would actually be completing the test until after they had responded to the prediction measures. All test materials and prediction measures were phrased in condition-appropriate ways (e.g., “you will be asked. . .” vs. “test takers would be asked. . .”). The test. The test itself was a time-limited, 20-item test consisting of moderate to difficult questions taken from verbal sections of past GREs. Participants in each test-expectancy condition were randomly assigned to evaluate (and then to complete) one of three different tests, each consisting of

589

only one type of GRE test question: analogies, antonyms, or sentence completions. All participants were provided with a thorough description of their particular test as well as some factual information that could, in principle, be used as a basis for making predictions. For example, participants were told that students nationwide have, on average, answered approximately 8 of the 20 questions correctly when answering these questions as part of the full GRE general test. Participants were also provided with a representative sample question of approximately the same average difficulty as the items that would appear on their test, as well as an official ETS-supplied explanation of the correct answer (these explanations were obtained from the ETS website, at http://www.gre.org/practice_test/takesc.html). Prediction measures. As in Study 1, participants were asked to make a general prediction (“How well do you think you will [would] do on the sample test?”), to which they responded on a 7-point scale, and a specific prediction (“How many out of 20 questions do you think you will [would] answer correctly?”). In order to assess relative appraisals, participants were asked to make a single prediction about how well they would perform relative to other students on an 11-point percentile scale ranging from 0% (not better than any other Yale student) through 50% (better than half) to 100% (better than all other Yale students). For exploratory purposes, we also asked participants to evaluate their confidence in the accuracy of their prediction, their belief in how well they generally do on these kinds of tests, and how diagnostic they thought the test would be in terms of determining their academic potential. Responses to these questions were assessed on 7-point scales, with higher numbers reflecting greater confidence, efficacy, and diagnosticity. Taking the test. After providing their responses, all participants were asked to complete the 10-minute GRE test. Participants were then asked a number of follow-up questions, probed for suspicion, and debriefed.

Results and Discussion Preliminary analyses revealed unexpected differences in participants’ Scholastic Aptitude Test (SAT) Verbal scores across conditions (test-unexpected M ⫽ 719, test-expected M ⫽ 743), t(73) ⫽ 1.90, p ⫽ .06. Accordingly, all analyses used these scores as a covariate. Additional analyses did not reveal any interactions with gender or test type; all subsequent analyses therefore collapse across these variables.

Performance Predictions As in Study 1, and as can be seen in Table 3, the test-expectancy manipulation consistently influenced the predictions participants

Table 3 Performance Predictions as a Function of Whether Participants Expected to Complete the Graduate Record Examination (Study 2) Test completion expectation Measure Prediction measures General prediction (how well?) Specific prediction (how many?) Relative appraisal (better than %) Other measures Personal efficacy (testing ability in general) Confidence in prediction accuracy Diagnosticity of test

Not expected

Expected

F(1, 72)

4.57 11.91 50.52

3.90 10.19 42.13

4.71* 4.09* 4.23*

5.53 4.06 3.19

5.38 3.92 3.01

0.46 0.66 0.54

Note. Means and inferential statistics have been adjusted for Scholastic Aptitude Test scores. * p ⬍ .05.

ARMOR AND SACKETT

590

made about their performance on the GRE test. Participants in the test-unexpected condition thought they would perform better ( p ⫽ .033), that they would answer more questions correctly ( p ⫽ .047), and that they would outperform a larger percentage of students at their university ( p ⫽ .043), than did participants in the testexpected condition. The manipulation did not, however, influence people’s beliefs about their general test-taking ability, their confidence in the accuracy of their predictions, or the extent to which they saw the test as diagnostic of their academic potential. Again, as in Study 1, neither group of participants believed that they would perform better than average. Although Study 2 used a more direct measure of relative appraisals, the results were similar: On this measure, participants in the test-unexpected condition give the appearance of being reasonably accurate, expecting on average to perform about average (M ⫽ 51st percentile), while participants in the test-expected condition expected to perform significantly worse than average (M ⫽ 42nd percentile), t(37) ⫽ 2.78, p ⫽ .004.

Performance As in Study 1, performance did not differ as a function of the test-expectancy manipulation. Although there was a trend for participants who expected to complete the test to perform a bit better (M ⫽ 10.2) than participants who did not expect to complete the test (M ⫽ 9.5), this difference was not significant, F(1, 72) ⫽ 1.15, ns.

predictions and performance, was also significantly greater among those who were not expecting to complete the test than among those who were expecting to complete the test, F(1, 72) ⫽ 8.03, p ⫽ .006. The difference in the correlation between predictions and performance was similarly substantial: For those not expecting to take the test, the correlation between predictions and performance was negligible (r ⫽ .05, p ⫽ .76), but for participants who did expect to complete the test, this correlation was positive and strong (r ⫽ .52, p ⬍ .001). These correlations are significantly different from one another (z ⫽ 2.18, p ⫽ .029).

Study 3 Although Studies 1 and 2 used very different performance tasks, the results were the same: In both, predictions were unrealistically optimistic in the task-unexpected conditions but impressively accurate in the task-expected conditions. Notably, in both studies, the tasks under consideration were quite specific, and the amount of information provided to participants was held constant across conditions. Therefore, between-condition differences in predictions and prediction accuracy in these studies cannot be attributed to differences in the amount of information available about the task. The explanatory hypotheses evaluated in Study 3 suggest other possibilities.

Three Explanatory Hypotheses

Accuracy, Error, and Bias

Feedback

As in Study 1, participants who expected to complete the test not only made more modest predictions than did participants who did not expect to complete the test, but they made more accurate predictions as well (see Table 4). Participants who did not expect to complete the test significantly overestimated how many questions they would answer correctly, t(37) ⫽ 3.55, p ⫽ .001, showing a clear optimistic bias, whereas participants in the testexpected condition did not, t(38) ⬍ 1, ns. This reduction in bias is significant, F(1, 72) ⫽ 6.09, p ⫽ .016. The average absolute error, again taken as the mean absolute value of the difference between

One important distinction that has been left unresolved in Studies 1 and 2 is whether the effects of the real versus hypothetical nature of the task were due to the effects of expecting to complete the task, as we have suggested, or whether our effects have been the consequence of differences in expectations about the possibility of critical feedback. In Studies 1 and 2, expectations about completing the test were potentially confounded with expectations about the availability of evaluative feedback. Although no mention of feedback was made in either study, participants in the taskexpected conditions may have nonetheless worried about the possibility of receiving feedback and made less optimistic predictions either (a) in an effort to avoid disappointment of unmet expectations (self-protection; e.g., Shepperd et al., 1996; Taylor & Shepperd, 1998; see also Josephs, Larrick, Steele, & Nisbett, 1992; Larrick, 1993) or (b) in an effort to avoid looking foolish in the eyes of others (public accountability; e.g., Study 2 of Regan, Gosselink, Hubsch, & Ulsh, 1975; Sedikides et al., 2002; see also Lerner & Tetlock, 1999; Tetlock, 1992; Tetlock & Kim, 1987). In Study 3, we attempted to disambiguate the effects of test completion expectations from feedback expectations by manipulating these expectations independently.

Table 4 Accuracy, Error, and Bias in Performance Predictions as a Function of Whether Participants Expected to Complete the Graduate Record Examination Test (Study 2) Test completion expectation Comparison of predictions to performance Prediction–performance discrepancy (bias) M (SD) Absolute deviation (error) M (SD) Prediction–performance correlation Pearson’s r

Not expected

Expected

2.45** (4.20)

0.03 (4.20)

4.46*** (2.64)

2.71*** (2.64)

.05

.52***

Note. Means and inferential statistics have been adjusted for Scholastic Aptitude Test scores; adjusted standard deviations are listed in parentheses. ** p ⬍ .01. *** p ⬍ .001 (significantly different from zero).

Mood Mood has been found to mediate the effects of temporal proximity (Savitsky et al., 1998) and feedback proximity (Shepperd et al., 1996, 2005) on performance predictions and post-task performance evaluations. In these studies, it is generally assumed that the proximity of performance (or of feedback) induces feelings of nervousness, which is then interpreted as a physiological signal that one will not, or did not, perform so well (see also Schwarz &

REAL VS. HYPOTHETICAL

Clore, 1983, 2003). Although it was not clear why this hypothesis would predict greater accuracy rather than greater pessimism, we tested for possible effects of mood in Study 3 by assessing mood at the time participants made predictions.

Construal Level According to construal level theory (Trope & Liberman, 2003), various manipulations of “psychological distance” will meaningfully influence the level of abstraction at which actions, events, and persons are mentally represented. Although most studies of construal level theory have focused on manipulations of temporal distance (e.g., Liberman, Sagristano, & Trope, 2002; Liberman & Trope, 1998; Nussbaum, Trope, & Liberman, 2003), Trope and Liberman (2003) have speculated that other manipulations, including the real versus hypothetical nature of events and actions, may effectively manipulate psychological distance as well. In the context of the present studies, construal level theory suggests that hypothetical events will be represented in more abstract, high-level terms and that real, proximal events will be represented in more concrete, low-level terms. As Trope and Liberman (2003) described the distinction, “high-level construals are relatively simple, decontextualized representations that extract the gist from available information,” whereas “low-level construals tend to be more concrete and include subordinate, contextual, and incidental features of events. . . low-level construals are thus richer and more detailed but less structured and parsimonious than high-level construals” (p. 405). At a higher, more abstract level, the GRE test may be thought of as a meaningful academic exercise on which participants will want to perform well. Such a representation could thus motivate people to be optimistic, as successful performance could be seen as symbolic of the test taker’s intellectual potential. At a lower, more concrete level, by contrast, the GRE test may be thought of in terms of broken pencils, smudgy erasers, and impending time limits—not the kind of high-level terms that would inspire optimistic predictions. When thinking of the test at this lower level, then, participants in the test-expected conditions may be more mindful of a host of contextual factors (e.g., testing conditions, their own level of alertness, etc.) that may be peripheral to the “gist” of the GRE test but relevant to making an accurate prediction. In an effort to assess these different levels of construal in Study 3, we asked participants to do two things: to recall very specific details about the GRE test, and to indicate, in essence, how meaningful they thought the test would be. To the extent that participants in the test-unexpected condition were representing the GRE test in abstract, high-level terms, they should be expected to have greater difficulty remembering specific details of the test; participants in the test-expected conditions, by contrast, were expected to form more concrete, lower-level representations of the test and thus were expected to remember more test details. The higher-level representation of the GRE test was also expected to be a more meaningful representation. We hypothesized that participants in the test-unexpected conditions would thus see more value in the test, and attach greater significance to their imagined performance on the test, in comparison with participants in the testexpected conditions who were expected to “dismiss the forest for the trees” and to deny the significance of the test and their test performance.

591 Method

Participants Seventy-five undergraduates (37 women and 38 men) from an introductory psychology course were recruited to participate in exchange for research participation credit. Four additional participants did not attend to experimental instructions and thus were not included in the final sample.

Procedure The test and cover story were identical to those used in Study 2: Participants were informed that they would be asked to evaluate (and, in some conditions, to complete) a test containing GRE questions that the ETS was considering for future examinations. Manipulating performance and feedback expectancies. Two thirds of participants were assigned to one of two test-expected conditions in which they were told that they would be taking the test. Of these participants, half were assigned to a “feedback-expected” condition: These participants were told that they would be scoring their own test at the end of the study and thus would know how they performed on the test. The remaining participants in the test-expected conditions were assigned to a “no-feedback” condition: they were told that the ETS would not allow participants to review the answers to test questions, that even the experimenter was not informed of the test answers, and that the experimenter would not be allowed to see the participants’ responses. To make this manipulation believable, participants in the no-feedback condition were told that they would be asked to seal their materials from the session, including their tests and answer sheets, in a stamped envelope addressed to the Educational Testing Service and to deposit these envelopes into a U.S. mail bin that had been placed in the test room. The remaining third of the participants was assigned to a test-unexpected condition identical to that of Study 2: They were told that they would be asked to answer questions about the test, but no mention was made of participants actually having to complete the test. Predictions and other measures. The prediction questions were the same as in Study 2, except that the efficacy question was not included. In addition, participants were asked to estimate their best and worst possible scores on the test. In order to assess participants’ representation of the test (i.e., as an indicator of construal level), participants were asked a series of questions about how meaningful they thought the test would be: They were asked how important it was for them to perform well on the test, how pleased they would be if they performed better than they expected, and how disappointed they would be if they performed worse than expected. The participants were also asked how diagnostic they thought the test would be and how important it was to make an accurate prediction. Participants were next asked to indicate how they felt “at that moment” by completing a brief measure of current mood (consisting of 10 items taken from the Positive and Negative Affect Schedule [PANAS]; Watson, Clark, & Tellegen, 1988; sample items included enthusiastic, happy, distressed [reverse scored] and nervous [reverse scored]; ␣ ⫽ .71). In order to assess the extent to which participants had attended to test details (i.e., as a second measure of construal level), all participants were then asked to recall specific pieces of information from the test description. In a freerecall task, administered first, participants were asked to simply recall as much information as they could about the test (i.e., from the one-page information summary that they had read earlier in the study). A subsequent cued recall test asked participants to recall four specific details from the information page (e.g., the amount of time given for the test, the number of questions answered correctly in a nationwide sample of GRE test takers, etc.). As a manipulation check, participants in the two feedback expectancy conditions were asked to recall whether they would receive their scores on the test (all remembered correctly). After providing their responses, all

ARMOR AND SACKETT

592

participants completed the 10-min GRE test and then were asked to complete several additional measures not central to the present investigation.

Results and Discussion Primary hypotheses were tested with pairs of orthogonal contrasts conducted in the context of a series of univariate analyses of covariance (ANCOVAs) in which participants’ SAT Verbal scores were entered as the covariate. The first of these contrasts analyzed the test-expectancy effect by testing for differences between the test-unexpected group and the two test-expected groups. The second contrast tested the effects of the feedback expectancy manipulation within the two task-expected conditions.

Performance Predictions As was the case in Studies 1 and 2, performance predictions were influenced by the real versus hypothetical nature of the GRE test. As can be seen in Table 5, in comparison with participants in the two test-expected conditions, participants who did not expect to complete the test tended to make more optimistic general predictions ( p ⫽ .094) and made significantly more optimistic specific predictions ( p ⫽ .010), expecting to answer over 15% more questions correctly than did participants in the test-expected conditions. Participants who did not expect to complete the test

also tended to imagine a better best-case score ( p ⫽ .086) and made significantly less pessimistic worst-case score predictions ( p ⫽ .003) than participants in the test-expected conditions. In contrast to Studies 1 and 2, participants’ predictions about their relative standing on the GRE test were not influenced by the test-expectancy manipulation. As can also be seen in Table 5, the feedback expectancy manipulation did not significantly affect any of the prediction measures, suggesting that it is the knowledge that one actually has to complete the task, and not the (normally associated) opportunity for feedback, that influences people’s predictions.

Performance Performance was once again not influenced by the testexpectancy manipulation, t(71) ⫽ 1.16, p ⫽ .25, and it was also not influenced by the feedback expectancy manipulation, t(71) ⫽ 1.06, p ⫽ .29. On average, participants answered 10.2 questions correctly.

Accuracy, Error, and Bias Our indices of prediction accuracy were strongly and significantly influenced by the test-expectancy manipulation but not by the feedback-expectancy manipulation (see Table 5). Although

Table 5 Effects of Test and Feedback Expectancy on Predictions, Prediction Accuracy, and Other Measures (Study 3) Experimental condition Measure Prediction Measures General prediction Specific prediction Best possible score Worst possible score Relative appraisal Accuracy, error, and bias Prediction–performance discrepancy (bias) M (SD) Absolute deviation (error) M (SD) Correlation Pearson’s r Potential Mechanisms Mood Meaningfulness Importance of doing well Pleased if score is better Disappointed if worse Diagnosticity of test Importance of accuracy Memory Free recall Cued recall

Test unexpected

No feedback expected

Planned contrasts t(71) Feedback expected

Test expectancy

Feedback expectancy

5.19 13.99 17.76 8.73 51.99%

4.87 12.21 16.79 6.86 51.59%

4.60 11.93 16.54 6.09 46.83%

1.70† 2.66** 1.74† 3.02** 0.68

0.86 0.32 0.33 0.86 1.00

4.39*** (4.37)

2.22** (2.98)

1.03 (4.29)

2.83**

0.35

5.31*** (3.22)

2.96*** (2.55)

3.61*** (2.54)

2.96**

0.79

.57**

.45*

z ⫽ 2.66**

z ⫽ 0.53

⫺.14 3.32

3.26

3.46

0.30

1.42

4.62 6.01 4.67 4.61 3.77

3.92 5.40 3.74 3.39 3.52

3.62 5.55 4.07 4.29 3.35

2.01* 2.24* 2.33* 2.51* 0.82

0.58 0.52 0.84 2.50* 0.36

3.91 2.46

5.01 3.05

5.64 3.41

2.53* 2.99**

0.96 1.17

Note. The prediction measure and potential mechanism values are means. Scholastic Aptitude Test Verbal scores have been covaried out of the above analyses. † p ⬍ .10. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.

REAL VS. HYPOTHETICAL

participants in both of the test-expected conditions tended to overestimate how well they would perform, neither group was as optimistically biased (combined M ⫽ 1.63) as participants in the test-unexpected condition (M ⫽ 4.39), p ⫽ .019. Viewed another way, among participants in the test-expected conditions, 1 in 3 (36%) met or exceeded their predicted score, and almost all (86%) met or exceeded their worst-case prediction; by contrast, among participants in the test-unexpected condition, only one in five (20%) met or exceeded their predicted score, and just over half (52%) met or exceeded their worst-case prediction. As in Studies 1 and 2, participants in the test-expected conditions were not only less biased but were also better calibrated than participants in the test-unexpected condition. The average absolute error between predictions and performance was significantly lower among those who were expecting to complete the test (combined M ⫽ 3.29) than among those who were not expecting to complete the test (M ⫽ 5.31), p ⫽ .004. Moreover, participants who expected to complete the test made predictions that were significantly correlated with their actual scores (r ⫽ .50, p ⬍ .001), but participants who did not expect to take the test did not (r ⫽ ⫺.14, ns), and this difference in correlations was significant (z ⫽ 2.66, p ⬍ .01). As with the other indicators of prediction accuracy, neither absolute error nor the correlations between predicted and actual scores were influenced by the feedback expectancy manipulation.5

Possible Mechanisms The lack of effects of the feedback manipulation renders the two feedback hypotheses (self-protection and public accountability) less plausible: If participants had been motivated to make more accurate predictions out of a concern that their predictions might be invalidated or out of a concern that others might view them critically, the feedback expectancy manipulation would have been expected to have had a significant effect on prediction and prediction accuracy, and it did not. Although it remains possible that a more powerful manipulation of feedback—perhaps with the addition of an evaluative component, or a clear public audience—may have had a more discernible effect (see, e.g., Regan et al., 1975; Sedikides et al., 2002), participants’ responses to the post-performance questions described in Footnote 5 (and to the task diagnosticity question, described below) suggest that they had attended to the feedback manipulation and were affected by it. Another mechanism that does not appear to account for the real versus hypothetical effect is mood. As can be seen in Table 5, neither the test-expectancy manipulation nor the feedback expectancy manipulation had effects on participants’ reported mood (this was true regardless of whether we looked at the composite mood score, as reported in Table 5, separate indices of positive and negative affect, or responses to individual items). Differences in mood, therefore, do not appear to account for the observed differences in predictions and prediction accuracy.6 Participants did differ, however, in how meaningful they thought the test would be (see Table 5). As predicted by construal level theory, participants in the test-unexpected condition reported that it would be more important for them to do well on the test ( p ⫽ .049), that they would be more emotionally affected if they over- or underperformed in relation to their prediction ( ps ⫽ .028 and .023) and that the test would be more diagnostic of their true

593

ability ( p ⫽ .014) than did participants in the test-expected conditions. The feedback expectancy manipulation, by contrast, only influenced the perceived diagnosticity of the test, revealing the unsurprising finding that participants who did not expect to receive feedback believed the test would be less diagnostic than those who did expect feedback ( p ⫽ .015). This effect—the sole significant effect of the feedback manipulation on pre-performance measures—is important because it demonstrates that participants were mindful of the manipulation and responded to it in a reasonable way (see also Footnote 5) and suggests that the lack of effects of this manipulation on the prediction and accuracy measures were not the consequence of a failed manipulation. As can be seen in Table 5, the test-expectancy manipulation also influenced how much participants remembered about the test before they took it. Again in keeping with the construal-level hypothesis, participants in the test-expected conditions were more successful in recalling significant test details than were participants in the test-unexpected condition, and this was true both for free recall ( p ⫽ .014) and cued recall ( p ⫽ .004). As was the case with the prediction and prediction accuracy measures, memory for test details was not influenced by the feedback expectancy manipulations. To determine whether the measures of meaningfulness and memory for details mediated the effects of the test-expectancy manipulation on prediction accuracy, we entered composite indices of meaningfulness (␣ ⫽ .77) and memory (␣ ⫽ .70) into a pair of multiple mediator models (see Kenny, Kashy, & Bolger, 1998; Preacher & Hayes, 2005), one testing mediation of the testexpectancy effect on prediction error, the other testing mediation of the effect on prediction bias. We evaluated the mediation hypotheses using a bootstrap approach advocated by Preacher and Hayes (2005; for additional discussion of the value of bootstrapping over other techniques for assessing mediation, such as the Sobel test, see Shrout & Bolger, 2002). Results of these analyses 5

We do not wish to claim that the feedback manipulation had no effects on optimism. The null effects of the feedback manipulation reported in the text were observed on prediction measures that had been obtained before the test was completed and thus well before the threat of feedback was imminent. Additional measures obtained after participants had completed the test revealed that participants in all conditions became pessimistic after they completed the test (reporting that they had answered, on average, 0.59 fewer questions correctly than they actually did answer correctly, F(1, 71) ⫽ 4.27, p ⫽ .042; cf. Shepperd et al., 1996; Heath & Jourden, 1997) but that this pessimism effect was more pronounced among those expecting feedback (M ⫽ ⫺1.10) than among those who did not (M ⫽ ⫺0.46) and those originally assigned to the test-unexpected condition (M ⫽ ⫺0.21). In fact, when looked at within condition, this post-performance pessimism only approached significance in the feedback-expected condition, t(24) ⫽ 1.93, p ⫽ .066. Because these post-performance procedures and results are beyond the scope of this article, they are not discussed further (for additional details, see Sackett, 2002). 6 Notably, mood was assessed after predictions. Therefore, it is possible that participants in the test-expected conditions may have (a) experienced a relatively negative mood prior to making predictions, (b) made more modest predictions in response to that mood, and then (c) experienced relief from that negative mood as a function of having made more modest predictions. In this way, our lack of mood effects does not necessarily rule out mood as a mechanism linking task completion expectations to prediction accuracy. We thank James Shepperd for suggesting this interpretation.

ARMOR AND SACKETT

594

reveal that reported meaningfulness, but not memory for test details, mediated the effects of the test-expectancy manipulation on our index of prediction bias. Specifically, the direct effect of test-expectancy on bias was rendered nonsignificant when the mediators were included in the model, t(70) ⫽ 1.58, p ⫽ .12. Moreover, the indirect effect through meaningfulness was statistically significant: Using Preacher and Hayes’s bootstrapping procedure, we obtained a point estimate for this effect of ⫺.18, with a bias-adjusted and accelerated 95% confidence interval of ⫺.44 to ⫺.01 (we can conclude that meaningfulness is a statistically significant mediator because this confidence interval does not contain zero). The indirect effect through memory, however, was not significant: The point estimate for this effect was ⫺.16, with a bias-adjusted and accelerated 95% confidence interval of ⫺.52 to .03. Neither meaningfulness nor memory for test details were found to mediate the effects of the test-expectancy manipulation on our index of prediction error.

Alternate Explanations? Although the meaningfulness and memory results are consistent with a construal-level interpretation, alternative explanations are possible. For example, the finding that participants in the testexpected conditions appeared to downplay the meaningfulness of the GRE test could be interpreted as a self-protective strategy (dismissing the significance of the test could be an effective strategy for minimizing the impact of receiving the low scores that they expected to receive) rather than a result of lower-level construals posited by temporal construal theory; however, the lack of effects of the feedback expectancy manipulation would seem to render this interpretation less plausible. A different explanation could be invoked to explain the memory data: It is possible that participants in the task-unexpected conditions may have remembered fewer details about the test because they simply did not think very carefully about tasks that they believed would remain hypothetical. However, neither self-protection nor this “insufficient processing” account can parsimoniously explain why participants in the test-unexpected condition saw the test as more meaningful and remembered fewer details about it than did participants in the test-expected conditions.

knew they would not be asked to complete. Half of the participants were assigned to a “context-unspecified” condition in which they were asked to imagine our hypothetical test in the same way we had asked participants to imagine the test in our prior testunexpected conditions, that is, without any specificity about when or where they would complete the test. Remaining participants were assigned to a “here-and-now” condition in which they were asked to imagine not only that they would be taking the test but also that they would be taking the test at that very moment (i.e., at that exact time and in that location). The simple addition of these temporal–local constraints on participants’ conceptualization of the test was expected to make the hypothetical seem real—and thereby to influence participants’ construal level of the test and their predictions of how well they would perform on it—without requiring them to actually complete the test. As a second goal, we sought to further disentangle the construallevel hypothesis (which focuses on the content of processing) from the insufficient processing alternative (which focuses on quantity of processing). We did this, in part, by asking participants to elaborate on the factors that might influence their performance on the GRE test either before or after they were asked to make predictions. This procedure was intended both as a manipulation of the extent to which participants elaborated on the test before making predictions (allowing a test of the causal effect of more thorough cognitive processing on optimistic predictions) as well as a method for assessing cognitive content both before and after participants made their predictions. If participants in our previous task-unexpected conditions had been unrealistically optimistic because they had simply thought less thoroughly about the tasks in question, then the following should hold true: (a) Participants in the context-unspecified condition should generate fewer factors as relevant to their performance (insofar as the context-unspecified conditions are intended to replicate the task-unexpected conditions of Studies 1 to 3) and (b) Encouraging participants to think more thoroughly about the factors influencing their performance before they make their predictions should lead these participants to be less excessively optimistic.

Method Participants

Study 4: Making the Hypothetical Seem Real Our final experiment was designed with two goals in mind. First, we sought to determine whether the kinds of thought processes invoked by the real prospect of completing a task could be induced without requiring participants to actually complete the task. In other words, we wanted to know whether personal predictions could be made to be less optimistic even when the task under consideration is understood to be purely hypothetical. If real tasks are somehow more motivating or have more implications for self-evaluation or self-protection, and if these features serve as incentives to be accurate, then removing these features by making the task hypothetical in all conditions should eliminate betweencondition differences in participants’ predictions. If, on the other hand, the processes involved are largely imaginative, then one should be able to mimic the conditions of taking a real test without actually asking participants to complete the test. Therefore, in Study 4, we asked all participants to imagine taking a test that they

One hundred and two university students (64% women) were recruited to participate in a brief study in exchange for a small incentive (a candy bar or a cold drink).

Procedure Participants were asked to make predictions about how well they thought they would perform on a test of sample GRE questions. The test description provided to participants was identical to the ones used in Studies 2 and 3. The experimenter made it clear to all participants that they would not actually be taking the test, and the experiment itself was presented as an exercise in imagination. Half of our participants were randomly assigned to a “contextunspecified” condition—a condition directly analogous to the testunexpected conditions of Studies 2 and 3—in which they were asked to imagine taking the GRE test without any constraints on when or where they should imagine themselves taking the test. Remaining participants were assigned to a “here-and-now” condition in which they were asked to

REAL VS. HYPOTHETICAL imagine that they really would be taking the test and that they would be doing so in the immediate present (i.e., at that moment and in that location). To provide an experimental test of the insufficient processing hypothesis, participants in each of the imagination conditions were randomly assigned to either a “prior elaboration” condition (which was intended to encourage effortful processing when making predictions) or to a “delayed elaboration” condition (which was not). In the prior elaboration conditions, participants were asked to “list as many factors as you can think of that might influence your performance on this verbal test” before they were asked to make their performance predictions. In the delayed elaboration condition, participants were not asked to list influential factors until after they had made their performance predictions. As in the previous studies, participants were asked to make both general (“how well?”) and specific (“how many?”) predictions, as well as a number of other assessments (how difficult they thought the test would be, how capable they would be of concentrating, how much they would enjoy the test, and how motivated they would be to perform well).

Results and Discussion Effects on Predictions As hypothesized, participants in the “here-and-now” conditions, who had been asked merely to imagine taking the test in the immediate present, made more modest predictions and task assessments than did participants in the context-unspecified (or “pure hypothetical”) conditions. In comparison, participants in the context-unspecified conditions expected to perform better on the test (Ms ⫽ 5.54 vs. 4.78), F(1, 98) ⫽ 11.4, p ⬍ .001, and to answer more questions correctly (Ms ⫽ 15.0 vs. 13.6), F(1, 98) ⫽ 4.82, p ⫽ .03. Notably, participants in both imagination conditions expected to answer more questions correctly than participants actually had answered correctly when taking the same test in Studies 2 (M ⫽ 9.8) and 3 (M ⫽ 10.2), suggesting that both groups were optimistically biased. Nonetheless, the significant betweencondition difference in predicted scores reveals that participants in the here-and-now conditions were significantly closer to these expected values. In comparison with participants in the context-unspecified conditions, participants in the here-and-now conditions also indicated that they would enjoy the test less (Ms ⫽ 3.06 vs. 3.88), F(1, 98) ⫽ 6.22, p ⫽ .014, that they would be less capable of concentrating (Ms ⫽ 3.60 vs. 5.25), F(1, 97) ⫽ 32.7, p ⬍ .001, and that they would be less motivated to perform well (Ms ⫽ 4.36 vs. 5.62), F(1, 98) ⫽ 15.5, p ⬍ .001. The only responses that were unaffected by the imagination manipulation were the assessments of task difficulty (F ⬍ 1, ns). In contrast to the effects of the imagination manipulation, which significantly influenced responses on five of our six dependent measures, the cognitive elaboration manipulation did not consistently influence predictions or task assessments. The only trend to approach significance was the question about anticipated motivation, with participants in the prior elaboration condition stating that they would be less motivated (M ⫽ 4.64) than participants in the delayed elaboration condition (M ⫽ 5.29), F(1, 98) ⫽ 3.71, p ⫽ .057 (all other Fs ⬍ 1). The order manipulation also did not interact with the imagination condition to influence either predictions (Fs ⬍ 1) or the other assessment measures (Fs ⬍ 2.9, ps ⬎ .09). These results are thus inconsistent with the insufficient processing hypothesis: Inducing participants to think more thoroughly about the factors that might influence their performance did not

595

have a significant effect on their predictions, either as a main effect or through selectively making seemingly “neglectful” participants in the context-unspecified condition more mindful.

Factors Listed: Quantitative Differences The lack of significant effects of the cognitive elaboration manipulation does not appear to have been the consequence of an unsuccessful manipulation: Participants in the prior elaboration conditions listed more factors (M ⫽ 8.46) than did participants in the delayed elaboration conditions (M ⫽ 6.90), F(1, 98) ⫽ 4.47, p ⫽ .037, suggesting that participants in the prior elaboration conditions had thought more thoroughly about these factors than had participants in the delayed elaboration conditions. Thinking thoroughly, then, does not appear to be a sufficient remedy for overly optimistic predictions. A second strike against the insufficient processing hypothesis is that participants in the context unspecified conditions, who did make more optimistic predictions, did not appear to reason less thoroughly (and thus generate a shorter list of causal factors) than participants in here-and-now conditions; if anything, there was a trend in the reverse direction (Ms ⫽ 7.06 and 8.25, respectively), F(1, 98) ⫽ 2.37, p ⫽ .13.

Factors Listed: Qualitative Differences Coding and analysis of the content of participants’ factor lists provided additional insight into the thought processes involved in the here-and-now and context-unspecified conditions. To facilitate coding, each participant’s factor list was first partitioned into single-factor units, and the resulting factors were then coded along three independent dimensions (described below). In order to assess rater reliability, a second rater coded factor lists from a random sample of 20 participants drawn equally from all conditions (24.6% of all factors generated). Interrater reliability was acceptable for all three categories (agreement ⬎ 83%; kappas ⬎ .69). Attribution. Drawing from conventional distinctions within attribution theory (e.g., Kelley, 1967), causes were coded as making reference to the self (person attributions), to the test (stimulus attributions), or to the testing environment (circumstance attributions). In order to test predictions from construal-level theory, self-attributions were further subdivided into stable self characteristics (such as traits, general abilities, and test-taking experience) and unstable self characteristics (such as current mood, alertness, and level of concentration). Overall, more than two thirds of the factors listed identified aspects of the self, and the majority of these aspects referred to unstable characteristics (including mood states, feelings of confidence, alertness, health, and ability to concentrate). However, in keeping with the predictions of construal level theory, participants in the context-unspecified conditions were more than twice as likely to make reference to stable aspects of themselves, such as to their vocabulary, intelligence, and affinity for standardized tests, than were participants in the here-and-now conditions (Ms ⫽ 18% and 8%, respectively), F(1, 97) ⫽ 5.49, p ⫽ .021. Participants in the context-unspecified conditions were also twice as likely to make reference to the only other stable factor (i.e., the test), though this effect (7% vs. 3%) was only marginally significant, F(1, 97) ⫽ 3.45, p ⫽ .066. Participants in the here-and-now conditions, by contrast, made correspondingly greater reference to the two less

596

ARMOR AND SACKETT

stable factors: unstable self characteristics (59% vs. 53%) and the environment (29% vs. 21%). None of these effects were moderated by the order of the factor listing procedure (all interaction Fs ⬍ 2.0, all ps ⬎ .15). Causal certainty. Some of the causal factors that participants listed were described in such a way as to indicate that participants were certain of their relevance to predicting test performance (e.g., “I generally do well on these kinds of tests”; “timed tests are always bad”), and other factors were described so as to indicate that participants were not sure of their relevance (e.g., “the weather, maybe”; “I might get bored”). To capture this causal certainty distinction, factors were coded as either definite or as possible/uncertain. Participants in the context-unspecified conditions expressed greater uncertainty about whether the factors they listed would be relevant to their performance on the GRE test, at least in the delayed elaboration conditions. Analysis of the proportion of factors participants described as being of uncertain relevance to their performance revealed a main effect of the imagination manipulation, F(1, 97) ⫽ 6.03, p ⫽ .016, that was driven by a significant interaction between the imagination and elaboration manipulations, F(1, 97) ⫽ 7.12, p ⫽ .009. In the prior-elaboration conditions, participants in the context-unspecified condition were uncertain about roughly the same proportion of factors (M ⫽ 30%) as were participants in the here-and-now conditions (M ⫽ 32%). In the delayed elaboration conditions, by contrast, participants in the context-unspecified condition were uncertain about a much larger proportion of factors (M ⫽ 47%) than were participants in the here-and-now condition (M ⫽ 15%). Although the greater uncertainty in the context-unspecified conditions is not surprising—this uncertainty may explain why participants in these psychologically distant conditions need to rely on abstract, gist-based representations of the target event (i.e., because so many of the specifics remain unknown)—the reason behind the interaction effect is unclear. Causal valence. Coders also determined the expected effect of each causal factor (or of the participants’ standing on these factors) on test performance. Causal factors were coded as implying either a positive effect on test performance (e.g., “I know a lot of vocabulary”) or a negative effect on test performance (e.g., “lack of preparation”). In some cases, participants listed causal factors without specifying the nature of the effect it may have (e.g., “my mood might influence my performance”) or explicitly indicated uncertainty about the likely effect (e.g., “the test format is odd. . . I’m not sure whether that will be a good thing or a bad thing”); these factors were coded as unspecified/unknown, and reflected a different kind of uncertainty. Participants in the here-and-now conditions identified proportionally more factors that would have negative implications for their performance (M ⫽ 41%) than did participants in the contextunspecified conditions (M ⫽ 21%), suggesting that these participants may have been thinking more self-critically (see Gilovich et al., 1993, Study 4; see also Tetlock, 1992; Tetlock & Kim, 1987; cf. Eyal, Liberman, Trope, & Walther, 2004). Neither group was particularly inclined to identify factors that would have clearly positive effects on their performance (Ms ⫽ 13% and 15%, respectively). However, in keeping with the results of the causal certainty codes, participants in the context-unspecified conditions listed factors that were more frequently coded as having unspec-

ified or unknown effects (M ⫽ 64%) than factors listed by participants in the here-and-now conditions (M ⫽ 45%), again suggesting that participants in the context-unspecified conditions were facing greater uncertainty when formulating their predictions. In contrast to the results with the causal certainty codes, however, these causal valence results were not moderated by the elaboration order manipulation. Summary. The coded data help explain both the logic of construal level theory and the general implausibility of the insufficient processing hypothesis. Participants in the context-unspecified condition were asked to confront a prediction task facing many unknowns—what the testing conditions would be like, what their own mental and physical state would be like at the time of the test, and so on—whereas participants in the here-and-now condition were confronted with the same prediction task with many fewer unknowns. It thus does not appear that participants in the testunexpected conditions were simply thinking less thoroughly about the determinants of their performance (though they do appear to have been thinking about them less self-critically; see also Gilovich et al., 1993). The added uncertainty inherent in hypothetical events—“when?” “where?” “what kind of state will I be in?”— thus creates a more complex prediction task, one that may not be solvable through the simple application of cognitive effort.

General Discussion As early as LaPiere’s (1934) analysis of attitude– behavior inconsistencies, psychologists have been aware that people’s claims about their beliefs and behaviors will differ depending on the social context in which they find themselves. Although LaPiere’s classic study may be more familiarly discussed in the context of attitudes and behavior (e.g., Kraus, 1995; Wicker, 1969), the questions LaPiere asked of his respondents were, in fact, behavioral predictions, and those predictions were asked anonymously and without the expectation that they would subsequently be put to test. Results of LaPiere’s inquiries, and many others since, have consistently shown that under hypothetical, task-unexpected conditions, people tend to make predictions that are largely in line with societal expectations and aspirations of the day (see also Linn, 1965; Sherman, 1980; Weinstein, 1980; Woodzicka & LaFrance, 2001). In short, they tend to make predictions that are unrealistically optimistic. Results from the four studies presented here revealed that people’s claims about their own expected outcomes will also differ depending on the context in which their predictions are made. Participants who had been asked to make predictions about performance on tasks they thought were hypothetical made predictions that were unrealistically optimistic in the truest sense of the term: They overestimated how they would perform, often by large margins, and their predictions were, at best, only weakly correlated with their actual performance. By contrast, participants who had been asked to make predictions about their performance on tasks that they knew they would complete made predictions that were impressively accurate: Their predictions, on average, did not deviate from their average performance, their overall rate of error was reduced, and their predictions were strongly correlated with their performance. These results are consistent with Armor and Taylor’s (1998, 2002) suggestion that optimistic biases are not invariant within persons but rather are situated within contexts:

REAL VS. HYPOTHETICAL

These biases do not appear to be an inevitable part of the prediction process, individual forecasters do not appear to be indiscriminately optimistic, and the expression of these biases appears to be largely dependent on the psychological context in which predictions are made (for similar conclusions, see also Armor & Taylor, 2003; Sackett & Armor, 2005a; Shepperd et al., 2006). The results of these studies are also broadly consistent with Gilovich et al.’s (1993; Savitsky et al., 1998) work on the effects of temporal distance on subjective confidence and with Trope and Liberman’s (2003) suggestion that task hypotheticality and temporal proximity may be effectively analogous as manipulations of “psychological distance.” The results of these studies go beyond those of Gilovich et al., however, by showing that predictions for real, proximal events are not only less biased than predictions for hypothetical events, but that they are more accurate as well. By measuring performance as a criterion for prediction accuracy, estimates of accuracy, error, and bias could be assessed and compared across conditions. It is worth noting that the measures of relative appraisals—that is, the measures of how well participants thought they would perform in comparison with other people—yielded results that, in Studies 1 and 2 at least, were perfectly misleading about the accuracy of participants’ predictions. Although responses to these measures revealed a consistent decline in optimism in the taskexpected conditions, they also suggested that participants in the task-unexpected conditions were the ones who were accurate (having claimed that they would perform about the same as others on average) and that participants in the task-expected conditions were biased (having claimed that they would perform worse than others). Although these results were not expected, they are not unprecedented: Savitsky et al. (1998) obtained similar results in the context of a study of the effects of temporal distance. One possible explanation is that the performance tasks we used in these studies may have been seen as particularly difficult and that the hypothetical nature of the task may have led participants in the taskunexpected conditions to overcome their normal tendency to be pessimistic when contemplating difficult tasks (Kruger, 1999; see also Burson, Larrick, & Klayman, 2006; Chambers & Windschitl, 2004). Understanding these differences, and the general correspondence (or lack of correspondence) between various measures of prediction accuracy and bias will be an important challenge for future research. Another interesting result is that, in the three studies in which behavior was measured, we found little evidence of predictions leading to self-fulfilling prophecy. An impressive body of research suggests that, in Sherman’s (1980) words, optimistic errors of prediction can be “self erasing,” with even overly optimistic forecasts helping to bring about the expected outcomes (see also Armor & Taylor, 2003; Cervone & Peake, 1986; Sherman, Skov, Hervitz, & Stock, 1981; for a review, see Armor & Taylor, 2002). However, other studies have shown that even successful manipulations of people’s expectations do not always lead to changes in behavior (e.g., Wilson & LaFleur, 1995). In our experiments, participants in the task-unexpected conditions made predictions that were dramatically more optimistic than the predictions made by participants in the task-expected conditions, but this optimism did not carry over to influence performance. Understanding when and why predictions sometimes do and sometimes do not lead to self-fulfilling prophecy is another important question for future

597

research. One possible hypothesis is that predictions for real, proximate behavior (which were here found to be highly correlated with performance) may be sufficiently tied to people’s behavior to be self-fulfilling but that predictions for hypothetical behaviors (which were not found to be correlated with performance) may be relatively untethered and therefore not self-fulfilling (cf. Oettingen & Mayer, 2002). In other words, the truly unrealistic predictions made by participants in our task-unexpected conditions may have been too far removed from reality to have influenced subsequent behavior.

A Construal-Level Interpretation Studies 3 and 4 also provide some initial evidence about why participants’ predictions were unrealistically optimistic in the taskunexpected conditions but realistic in the task-expected conditions. Although many factors are likely to contribute to unrealistic optimism (Armor & Taylor, 1998; Buehler et al., 2002; Sackett & Armor, 2005b) and to prediction accuracy (Osberg & Shrauger, 1986), the results of Studies 3 and 4 suggest that the level of abstraction at which people represent real and hypothetical actions and events may contribute to the degree of accuracy, error, and bias present in people’s predictions. As predicted by construallevel theory (Trope & Liberman, 2003), participants in the taskunexpected conditions appeared to represent the hypothetical tasks in more abstract, high-level terms than did participants in the task-expected conditions: They saw the GRE test as more meaningful yet remembered fewer particular details about the test and, when listing factors that might influence their performance, tended to list more stable and fewer fleeting factors, which suggests that the participants were representing their performance on the task in more general, gist-based terms. In some respects, the construal-level results, too, were anticipated by LaPiere (1934), who argued that a participant’s response to a question is merely “a symbolic response to a symbolic situation” (p. 230), though construal level theory offers greater specificity about how these situations (and people’s imagined responses to them) will be symbolized. Results of Study 4 suggest that participants in the context-unspecified conditions were not only representing the “symbolic situation” (i.e., the performance task) in more abstract, high-level terms, but that they were also imagining their “symbolic response” to that task (i.e., their behavior) in more abstract, higher level terms as well. Research on self-evaluation has shown that, when people are asked to evaluate themselves in increasingly abstract terms, they tend to evaluate themselves much more positively than if they had been asked to evaluate themselves in more concrete terms (e.g., Dunning, Meyerowitz, & Holzberg, 1989). An intriguing question for future research is whether people are more inclined to rely on these abstract (and presumably positive) self-assessments when making predictions for hypothetical or otherwise psychologically distant tasks (cf. Ehrlinger & Dunning, 2003). Although construal level theory provides a reasonable account of the real versus hypothetical effects obtained in these studies, several questions remain. For example, it is not clear how construal level theory would explain why participants in the taskexpected conditions appeared to be pessimistic on the relative appraisal measures (though these results may be more of a challenge for the interpretation of relative appraisal measures of opti-

598

ARMOR AND SACKETT

mism than for the application of construal-level theory). The construal theory account would also be further strengthened by additional data showing that the level of construal mediates the effects of task hypotheticality on optimistic predictions.

How Unrealistic Is Unrealistic Optimism? The results of the studies reported here suggest a pair of very different conclusions about the nature of unrealistic optimism. On one hand, the predictions made by participants in the taskunexpected conditions demonstrate that people’s predictions can be truly unrealistic. These data would appear to offer further validation to a growing list of studies that have characterized people’s predictions as unrealistically optimistic (Armor & Taylor, 1998, 2002) and, like data from studies of the planning fallacy (Buehler et al., 2002), show how inaccurate people’s predictions can be when their predictions are compared to performance. On the other hand, the results of these experiments could be taken to suggest that the prevalence of unrealistic optimism may be overstated, and that the frequently documented expressions of excessive optimism may be partly enabled by the anonymous, consequence-free prediction environments in which optimistic forecasts are so commonly studied. Our studies show that despite a strong reputation for being optimistic, people appear to have a clear capacity to put optimistic biases aside and to be quite accurate when making predictions. We do not wish to claim, however, that optimistic biases are therefore necessarily the artifacts of anonymity, or that these biases are “hothouse creations” that will not be expressed in more consequential settings. Although a majority of studies on optimistic biases have been conducted in contexts free of consequence, these biases have been documented in increasingly diverse settings, including those in which consequences for inaccurate predictions are more immediate and real (e.g., Camerer & Lovallo, 1999; Glare et al., 2003; Lim, 2001; for review, see Dunning, Heath, & Suls, 2004). Our central argument is that context matters. Examining optimistic biases across meaningfully different contexts can help establish the generalizability of these biases, but cross-context comparisons can also usefully inform theory about the underlying processes involved in making predictions that are accurate, erroneous, or biased. We would also like to emphasize that, despite the contextually situated nature of unrealistic optimism, the predictions people make when thinking about hypothetical events may nonetheless be quite important. Although these “symbolic responses” appear to be much revised in situations in which people recognize that their predictions may be tested, these predictions may nonetheless influence behavior and outcomes in the situations in which they are expressed. People are often asked to make predictions about events or outcomes that are, at the time of prediction, essentially hypothetical. People may also rely on optimistic estimates of what they would have done in a particular situation when judging what another person actually did in that situation, which could lead to overly critical interpersonal assessments. Further, psychologically, these optimistic forecasts may be important because they symbolize a future that one hopes to achieve and, thus, may be a source of comfort, pride, or self-esteem.

Trains in the Distance In his own melancholy way, Paul Simon evoked our central findings when he sang that “everybody loves the sound of a train in the distance” (Simon, 1983, track 7). The success of the song, and this lyric in particular, is that we can all conjure up an image of the distant rumblings of a rolling train and agree that the sound evokes pleasing feelings of intrigue, adventure, and forward purpose. Like Simon’s doomed-to-be-disillusioned lovers, participants in our psychologically distant, task-unexpected conditions were found to romanticize about even the most mundane of tasks, spinning optimistic fantasies about their performance on scavenger hunts and GREs. Moreover, participants in these task-unexpected conditions appeared to make their predictions without considering that if they were standing on the railroad tracks of immediate action, they might see things quite differently.

References Armor, D. A., & Taylor, S. E. (1998). Situated optimism: Specific outcome expectancies and self-regulation. Advances in Experimental Social Psychology, 30, 309 –379. Armor, D. A., & Taylor, S. E. (2002). When predictions fail: The dilemma of unrealistic optimism. In T. Gilovich, D. W. Griffin, & D. Kahneman (Eds.), Heuristics and biases: The psychology of intuitive judgment (pp. 334 –347). New York: Cambridge University Press. Armor, D. A., & Taylor, S. E. (2003). Mindset, prediction and performance: Self-regulation in deliberative and implemental frames of mind. Personality and Social Psychology Bulletin, 29, 86 –95. Berra, Y. (1998). The Yogi book. New York: Workman. Buehler, R., & Griffin, D. (2003). Planning, personality, and prediction: The role of future focus in optimistic time predictions. Organizational Behavior and Human Decision Processes, 92, 80 –90. Buehler, R., Griffin, D., Otsubu, Y., Lehman, D., & Heine, S. (2000). A cross-cultural comparison of the planning fallacy. Unpublished manuscript, Wilfred Laurier University, Waterloo, Ontario. Buehler, R., Griffin, D., & Ross, M. (1994). Exploring the “planning fallacy”: Why people underestimate their task completion times. Journal of Personality and Social Psychology, 67, 366 –381. Buehler, R., Griffin, D., & Ross, M. (2002). Inside the planning fallacy: The causes and consequences of optimistic time predictions. In T. Gilovich, D. W. Griffin, & D. Kahneman (Eds.), Heuristics and biases: The psychology of intuitive judgment (pp. 250 –270). New York: Cambridge University Press. Burson, K. A., Larrick, R. P., & Klayman, J. (2006). Skilled or unskilled, but still unaware of it: How perceptions of difficulty drive miscalibration in relative comparisons. Journal of Personality and Social Psychology, 90, 60 –77. Calderon, T. G. (1993). Predictive properties of analysts’ forecasts of corporate earnings. The Mid-Atlantic Journal of Business, 29, 41–58. Camerer, C., & Lovallo, D. (1999). Overconfidence and excess entry: An experimental approach. American Economic Review, 89, 306 –318. Cervone, D., & Peake, P. K. (1986). Anchoring, efficacy, and action: The influence of judgmental heuristics on self-efficacy judgments and behavior. Journal of Personality and Social Psychology, 50, 492–501. Chambers, J. R., & Windschitl, P. D. (2004). Biases in social comparative judgments: The role of nonmotivated factors in above average and comparative optimism effects. Psychological Bulletin, 130, 813– 838. Dunning, D., Heath, C., & Suls, J. M. (2004). Flawed self-assessment: Implications for health, education, and the workplace. Psychological Science in the Public Interest, 5, 69 –106. Dunning, D., Meyerowitz, J. A., & Holzberg, A. (1989). Ambiguity and self-evaluation: The role of idiosyncratic trait definitions in self-serving

REAL VS. HYPOTHETICAL assessments of ability. Journal of Personality and Social Psychology, 57, 1082–1090. Educational Testing Service. (2005). Graduate Record Examinations information and registration bulletin, 2005–2006. Princeton, NJ: Author. Ehrlinger, J., & Dunning, D. (2003). How chronic self-views influence (and potentially mislead) estimates of performance. Journal of Personality and Social Psychology, 84, 5–17. Eyal, T., Liberman, N., Trope, Y., & Walther, E. (2004). The pros and cons of temporally near and distant actions. Journal of Personality and Social Psychology, 86, 781–795. Gilovich, T., Kerr, M., & Medvec, V. H. (1993). Effect of temporal perspective on subjective confidence. Journal of Personality and Social Psychology, 64, 552–560. Glare, P., Virik, K., Jones, M., Hudson, M., Eychmuller, S., Simes, J., & Christakis, N. (2003). A systematic review of physicians’ survival predictions in terminally ill cancer patients. British Medical Journal, 327, 195–201. Heath, C., & Jourden, F. J. (1997). Illusion, disillusion, and the buffering effect of groups. Organizational Behavior and Human Decision Processes, 69, 103–116. Heine, S. J., & Lehman, D. R. (1995). Cultural variation in unrealistic optimism: Does the West feel more invulnerable than the East? Journal of Personality and Social Psychology, 68, 595– 607. Helweg-Larsen, M., & Shepperd, J. A. (2001). Do moderators of the optimistic bias affect personal or target risk estimates? A review of the literature. Personality and Social Psychology Review, 5, 74 –95. Hoch, S. J. (1985). Counterfactual reasoning and accuracy in predicting personal events. Journal of Experimental Psychology: Learning, Memory, and Cognition, 11, 719 –731. Josephs, R. A., Larrick, R. P., Steele, C. M., & Nisbett, R. E. (1992). Protecting the self from the negative consequences of risky decisions. Journal of Personality and Social Psychology, 2, 26 –37. Kelley, H. H. (1967). Attribution theory in social psychology. In D. Levine (Ed.), Nebraska Symposium on Motivation (Vol. 15, pp. 192–240). Lincoln: University of Nebraska Press. Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social psychology. In D. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), Handbook of social psychology (4th ed., Vol. 1, pp. 233–265). New York: McGraw-Hill. Kraus, S. J. (1995). Attitudes and the prediction of behavior: A metaanalysis of the empirical literature. Personality and Social Psychology Bulletin, 21, 58 –75. Kruger, J. (1999). Lake Wobegon be gone! The “below-average effect” and the egocentric nature of comparative ability judgments. Journal of Personality and Social Psychology, 77, 221–232. Kruger, J., & Burrus, J. (2004). Egocentrism and focalism in unrealistic optimism (and pessimism). Journal of Experimental Social Psychology, 40, 332–340. LaPiere, R. T. (1934). Attitudes vs. actions. Social Forces, 13, 230 –237. Larrick, R. P. (1993). Motivational factors in decision theories: The role of self-protection. Psychological Bulletin, 113, 440 – 450. Lerner, J. S., & Tetlock, P. E. (1999). Accounting for the effects of accountability. Psychological Bulletin, 125, 255–275. Liberman, N., Sagristano, M. D., & Trope, Y. (2002). The effect of temporal distance on level of mental construal. Journal of Experimental Social Psychology, 38, 523–534. Liberman, N., & Trope, Y. (1998). The role of feasibility and desirability considerations in near and distant future decisions: A test of temporal construal theory. Journal of Personality and Social Psychology, 75, 5–18. Lim, T. (2001). Rationality and analysts’ forecast bias. Journal of Finance, 56, 369 –385. Linn, L. S. (1965). Verbal attitudes and overt behavior: A study of racial discrimination. Social Forces, 44, 353–364.

599

Norem, J. K., & Cantor, N. (1986). Defensive pessimism: Harnessing anxiety as motivation. Journal of Personality and Social Psychology, 51, 1208 –1217. Nussbaum, S., Trope, Y., & Liberman, N. (2003). Creeping dispositionism: The temporal dynamics of behavior prediction. Journal of Personality and Social Psychology, 84, 485– 497. Oettingen, G., & Mayer, D. (2002). The motivating function of thinking about the future: Expectations versus fantasies. Journal of Personality and Social Psychology, 83, 1198 –1212. Orne, M. (1962). On the social psychology of the psychological experiment. American Psychologist, 17, 776 –783. Osberg, T. M., & Shrauger, J. S. (1986). Self-prediction: Exploring the parameters of accuracy. Journal of Personality and Social Psychology, 51, 1044 –1057. Perloff, L. S., & Fetzer, B. K. (1986). Self– other judgments and perceived vulnerability to victimization. Journal of Personality and Social Psychology, 50, 502–511. Preacher, K. J., & Hayes, A. F. (2005). Asymptotic and resampling strategies for assessing and comparing indirect effects in simple and multiple mediator models. Unpublished manuscript, University of North Carolina, Chapel Hill. Quadrel, M. J., Fischhoff, B., & Davis, W. (1993). Adolescent (in)vulnerability. American Psychologist, 48, 102–116. Regan, J. W., Gosselink, H., Hubsch, J., & Ulsh, E. (1975). Do people have inflated views of their own ability? Journal of Personality and Social Psychology, 31, 295–301. Sackett, A. M. (2002). Optimism and accuracy in performance predictions: An experimental test of the self-protection hypothesis. Unpublished master’s thesis, Yale University. Sackett, A. M., & Armor, D. A. (2005a). Manipulating the reasons for optimism: Reversing bias by shifting consequences. Unpublished manuscript, Yale University. Sackett, A. M., & Armor, D. A. (2005b). Reasons for optimism: Strategic bias or unwanted error? Unpublished manuscript, Yale University. Savitsky, K., Medvec, V. H., Charlton, A. E., & Gilovich, T. (1998). “What, me worry?” Arousal, misattribution and the effect of temporal distance on confidence. Personality and Social Psychology Bulletin, 24, 529 –536. Scheier, M. F., Carver, C. S., & Bridges, M. W. (1994). Distinguishing optimism from neuroticism (and trait anxiety, self-mastery, and selfesteem): A reevaluation of the life orientation test. Journal of Personality and Social Psychology, 67, 1063–1078. Schwarz, N. (1994). Judgment in a social context: Biases, shortcomings, and the logic of conversation. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 26, pp. 123–162). San Diego, CA: Academic Press. Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judgments of well-being: Informative and directive functions of affective states. Journal of Personality and Social Psychology, 45, 513–523. Schwarz, N., & Clore, G. L. (2003). Mood as information: 20 years later. Psychological Inquiry, 14, 296 –303. Sedikides, C., Herbst, K. C., Hardin, D. P., & Dardis G. J. (2002). Accountability as a deterrent to self-enhancement: The search for mechanisms. Journal of Personality and Social Psychology, 83, 592– 605. Shepperd, J. A., Sweeny, K., & Carroll, P. J. (2006). Abandoning optimism in predictions about the future. In L. J. Sanna & E. Chang (Eds.), Judgments over time: The interplay of thoughts, feelings, and behaviors (pp. 13–33). New York: Oxford University Press. Shepperd, J. A., Grace, J., Cole, L. J., & Klein, C. (2005). Anxiety and outcome predictions. Personality and Social Psychology Bulletin, 31, 267–275. Shepperd, J. A., Ouellette, J. A., & Fernandez, J. K. (1996). Abandoning unrealistic optimism: Performance estimates and the temporal proximity

600

ARMOR AND SACKETT

of self-relevant feedback. Journal of Personality and Social Psychology, 70, 844 – 855. Sherman, S. J. (1980). On the self-erasing nature of errors of prediction. Journal of Personality and Social Psychology, 39, 211–221. Sherman, S. J., Skov, R. B., Hervitz, E. F., & Stock, C. B. (1981). The effects of explaining hypothetical future events: From possibility to actuality and beyond. Journal of Experimental Social Psychology, 17, 142–158. Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental studies: New procedures and recommendations. Psychological Methods, 7, 422– 445. Simon, P. (1983). Train in the distance. On Hearts and bones [CD]. New York: Warner Brothers. (1990) Sniezek, J. A., Paese, P. W., & Switzer, F. S. (1990). The effect of choosing on confidence and choice. Organizational Behavior and Human Decision Processes, 46, 264 –282. Taylor, K. M., Shepperd, J. A. (1998). Bracing for the worst: Severity, testing, and feedback timing as moderators of the optimistic bias. Personality and Social Psychology Bulletin, 24, 915–926. Tetlock, P. E. (1992). The impact of accountability on judgment and choice: Toward a social contingency model. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 25, pp. 331–376). New York: Academic Press. Tetlock, P. E., & Kim, J. I. (1987). Accountability and judgment processes in a personality prediction task. Journal of Personality and Social Psychology, 52, 700 –709. Tetlock, P. E., & Lerner, J. (1999). The social contingency model: Identifying empirical and normative boundary conditions on the error-andbias portrait of human nature. In S. Chaiken & Y. Trope (Eds.), Dual-

process theories in social psychology (pp. 571–585). New York: Guilford Press. Trope, Y., & Liberman, N. (2003). Temporal construal. Psychological Review, 110, 403– 421. Watson, D., Clark, L., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Weinstein, N. D. (1980). Unrealistic optimism about future life events. Journal of Personality and Social Psychology, 39, 806 – 820. Weinstein, N. D. (1987). Unrealistic optimism about susceptibility to health problems: Conclusions from a community-wide sample. Journal of Behavioral Medicine, 10, 481–500. Weinstein, N. D. (1998). References on optimistic biases about risk, unrealistic optimism, and perceived invulnerability. Unpublished manuscript, Rutgers University. Weinstein, N. D., & Klein, W. M. (1995). Resistance to personal risk perceptions to debiasing interventions. Health Psychology, 14, 132–140. Wicker, A. W. (1969). Attitudes versus actions: The relationship of verbal and overt behavioral responses to attitude objects. Journal of Social Issues, 25, 41–78. Wilson, T. D., & LaFleur, S. J. (1995). Knowing what you’ll do: Effects of analyzing reasons on self-prediction. Journal of Personality and Social Psychology, 68, 21–35. Woodzicka, J. A., & LaFrance, M. (2001). Real versus imagined gender harassment. Journal of Social Issues, 57, 15–30.

Received September 1, 2005 Revision received December 24, 2005 Accepted January 8, 2006 䡲

600

ARMOR AND SACKETT

of self-relevant feedback. Journal of Personality and Social Psychology, 70, 844 – 855. Sherman, S. J. (1980). On the self-erasing nature of errors of prediction. Journal of Personality and Social Psychology, 39, 211–221. Sherman, S. J., Skov, R. B., Hervitz, E. F., & Stock, C. B. (1981). The effects of explaining hypothetical future events: From possibility to actuality and beyond. Journal of Experimental Social Psychology, 17, 142–158. Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental studies: New procedures and recommendations. Psychological Methods, 7, 422– 445. Simon, P. (1983). Train in the distance. On Hearts and bones [CD]. New York: Warner Brothers. (1990) Sniezek, J. A., Paese, P. W., & Switzer, F. S. (1990). The effect of choosing on confidence and choice. Organizational Behavior and Human Decision Processes, 46, 264 –282. Taylor, K. M., Shepperd, J. A. (1998). Bracing for the worst: Severity, testing, and feedback timing as moderators of the optimistic bias. Personality and Social Psychology Bulletin, 24, 915–926. Tetlock, P. E. (1992). The impact of accountability on judgment and choice: Toward a social contingency model. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 25, pp. 331–376). New York: Academic Press. Tetlock, P. E., & Kim, J. I. (1987). Accountability and judgment processes in a personality prediction task. Journal of Personality and Social Psychology, 52, 700 –709. Tetlock, P. E., & Lerner, J. (1999). The social contingency model: Identifying empirical and normative boundary conditions on the error-andbias portrait of human nature. In S. Chaiken & Y. Trope (Eds.), Dual-

process theories in social psychology (pp. 571–585). New York: Guilford Press. Trope, Y., & Liberman, N. (2003). Temporal construal. Psychological Review, 110, 403– 421. Watson, D., Clark, L., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Weinstein, N. D. (1980). Unrealistic optimism about future life events. Journal of Personality and Social Psychology, 39, 806 – 820. Weinstein, N. D. (1987). Unrealistic optimism about susceptibility to health problems: Conclusions from a community-wide sample. Journal of Behavioral Medicine, 10, 481–500. Weinstein, N. D. (1998). References on optimistic biases about risk, unrealistic optimism, and perceived invulnerability. Unpublished manuscript, Rutgers University. Weinstein, N. D., & Klein, W. M. (1995). Resistance to personal risk perceptions to debiasing interventions. Health Psychology, 14, 132–140. Wicker, A. W. (1969). Attitudes versus actions: The relationship of verbal and overt behavioral responses to attitude objects. Journal of Social Issues, 25, 41–78. Wilson, T. D., & LaFleur, S. J. (1995). Knowing what you’ll do: Effects of analyzing reasons on self-prediction. Journal of Personality and Social Psychology, 68, 21–35. Woodzicka, J. A., & LaFrance, M. (2001). Real versus imagined gender harassment. Journal of Social Issues, 57, 15–30.

Received September 1, 2005 Revision received December 24, 2005 Accepted January 8, 2006 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 601– 611

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.601

Regulatory Fit as Input for Stop Rules Leigh Ann Vaughn, Jill Malik, Sandra Schwartz, Zhivka Petkova, and Lindsay Trudeau Ithaca College Three experiments show that the motivational effects of regulatory fit (consistency between regulatory state and strategic means) are context dependent. With no explicit decision rule about when to stop (Experiment 1) or an explicit enjoyment stop rule (Experiments 2 and 3), participants exerted more effort on tasks when experiencing regulatory fit than when experiencing regulatory nonfit. With an explicit sufficiency stop rule (Experiments 2 and 3), participants exerted less effort when experiencing regulatory fit than when experiencing regulatory nonfit. The interactive effect of regulatory fit and stop rules can be explained by misattribution of rightness feelings from regulatory fit; the effect was eliminated by drawing participants’ attention to an earlier event as a source of rightness feelings (Experiments 1 and 3). Keywords: regulatory focus, fit, motivation, input, stop rules

or services to ensure a profit. When our focus is on performance, we may put effort into a task until we feel that we have done at least an adequate job, in part because the situation requires it (e.g., Hirt et al., 1996; Martin et al., 1993). The effort we put into a task is important to our optimal enjoyment or performance at it. For example, a crossword puzzle that is not interesting probably is not worth doing, and if playing the piano, running, thinking about a new business venture, or any other taskoriented activity is not enjoyable today, perhaps it is best to stop early and try more tomorrow. However, if one is trying to meet a performance goal, it is important to do as much as it takes to achieve an adequate level of performance in the current situation, whether one’s goal is doing well in a course, getting a good score on the GRE, staying on the track team, or making a profit. What affects how much effort we put into an activity when we are either task or performance focused? Research suggests that feelings can serve as input for stop rules: decision rules about when to stop working on a task (Hirt et al., 1996; Martin & Stoner, 1996; Martin et al., 1993; Martin & Whitaker, 2000; Sanna, Parks, & Chang, 2003; Sanna, Turley, & Mark, 1996; Startup & Davey, 2001). Task-focusing situations appear to lend themselves to enjoyment stop rules: decision rules such as “Am I enjoying this task?” and “Do I feel like continuing?” that promote stopping when the task is no longer enjoyable (e.g., Hirt et al., 1996; Martin et al., 1993; Sanna et al., 1996). When we are focused on doing a task for its own sake, desirable feelings appear to indicate that the task is going well and that we enjoy it, making it more likely that we will continue than when we experience undesirable feelings. By contrast, performance-focusing situations appear to lend themselves to sufficiency stop rules: decision rules such as “Have I reached my goal?” and “Have I done all I can?” that promote stopping when we have attained an adequate level of performance (e.g., Hirt et al., 1996; Martin et al., 1993; Sanna et al., 1996). When we are focused on achieving a particular criterion or standard of performance, desirable feelings appear to indicate that we have met our goal, making it less likely that we will continue than when we experience undesirable feelings. A type of subjective experience that may serve effectively as input for stop rules is a feeling of rightness or wrongness. When we are

Every day we engage in various activities that are more or less enjoyable or boring, intrinsically or extrinsically motivating, and task-orienting or performance-orienting, and along the way, we make decisions about how much effort to put into them. Sometimes our focus is simply on engaging in a task. In such cases, we might be doing something that is interesting and enjoyable in its own right, unconcerned about evaluation, or in a situation that conveys the impression that ability is malleable and can improve with interest and effort (e.g., Deci & Ryan, 2000; Dweck & Leggett, 1988; Elliott & Dweck, 1988; Utman, 1997). For example, one might be teaching one’s self how to play the piano, doing a crossword puzzle that is not too difficult or too easy, running for the fun of it, or coming up with ideas for a new business venture. When our focus is on the task, we may continue it as long as it is enjoyable, in part because the situation allows us to (e.g., Hirt, Melton, McDonald, & Harackiewicz, 1996; Martin, Ward, Achee, & Wyer, 1993). Other times, our focus is more on meeting a standard or criterion of performance. We might be thinking about rewards or punishments (such as how our performance will make others feel about us), concerned about how we compare with others, aware that a certain level of performance is expected or required, or in a situation that conveys the impression that ability is fixed (e.g., Deci & Ryan, 2000; Dweck & Leggett, 1988; Elliott & Dweck, 1988; Hirt et al., 1996; Martin et al., 1993; Nicholls, 1984; Utman, 1997). For example, one might be practicing the piano because one needs to get a good grade in a course, practicing word problems to get a good verbal score on the GRE, running to do well enough to stay on a track team, or trying to sell a sufficient amount of goods

Leigh Ann Vaughn, Jill Malik, Sandra Schwartz, Zhivka Petkova, and Lindsay Trudeau, Department of Psychology, Ithaca College. We thank Melanie Green, John Luginsland, Thomas O’Rourke, Darcy Reich, Hugh Stephenson, and Gifford Weary for their helpful comments and advice and Lauren Graber for help with data collection. Correspondence concerning this article should be addressed to Leigh Ann Vaughn, Department of Psychology, 1119 Williams Hall, Ithaca College, Ithaca, NY 14850-7290. E-mail: [email protected] 601

602

VAUGHN, MALIK, SCHWARTZ, PETKOVA, AND TRUDEAU

focused on doing a task for its own sake, rightness feelings might indicate that the task is going well and that we enjoy it, making it more likely that we will continue than when we experience wrongness feelings. By contrast, when we are focused on achieving a particular criterion or standard of performance, rightness feelings might indicate that we have met our goal, making it less likely that we will continue than when we experience wrongness feelings.

Regulatory Focus and Regulatory Fit One source of rightness or wrongness feelings is a sense of a good or a poor fit between one’s regulatory focus and one’s goal-pursuit strategies. Regulatory-focus theory (Higgins, 1997, 1998) distinguishes between two self-regulatory states. Promotion focus is a state concerned with ideals, hopes, and aspirations (and more generally with the presence or absence of positive outcomes). Prevention focus is a state concerned with “oughts,” duties, and obligations (and more generally with the absence or presence of negative outcomes). Promotion-focused people prefer to use eagerness-related strategies of goal pursuit (e.g., doing extra reading for a class), which naturally fit a concern with aspirations and accomplishment. By contrast, prevention-focused people prefer to use vigilance-related strategies of goal pursuit (e.g., avoiding distractions while studying), which naturally fit a concern with responsibilities and protection (Crowe & Higgins, 1997; for reviews, see Higgins, 2000, 2005; Higgins & Spiegel, 2004). People experience regulatory fit when their goal-pursuit strategy fits and sustains their regulatory focus. There is evidence that experiencing regulatory fit produces feelings of rightness and importance (Camacho, Higgins, & Luger, 2003; Higgins, Idson, Freitas, Spiegel, & Molden, 2003), enjoyableness and excitement (Freitas & Higgins, 2002; Freitas, Liberman, & Higgins, 2002), processing fluency and ease (Lee & Aaker, 2004), and confidence in one’s judgments (Cesario, Grant, & Higgins, 2004). As is the case with other subjective experiences (e.g., Clore, 1992; Schwarz & Clore, 1983, 1996), people can use regulatory-fit feelings as information for judgments as long as they attribute those feelings to the judgment task (Cesario et al., 2004). For example, evaluative judgments tend to be more positive when one’s strategy of goal pursuit is congruent with one’s regulatory focus (and thus feels right) than when it is incongruent (and thus feels wrong; Camacho et al., 2003; Cesario et al., 2004; Freitas & Higgins, 2002; Higgins et al., 2003; Lee & Aaker, 2004). However, when regulatory fit is manipulated in an initial task and participants’ attention is drawn to that task as the source of those feelings, the regulatory-fit effect on later judgments is eliminated (Cesario et al., 2004). Attributing the feelings to the earlier source renders them irrelevant for the later judgments (also see Clore, 1992; Schwarz & Clore, 1983, 1996). Regulatory-fit theory proposes that regulatory fit increases the personal value of a goal pursuit relative to regulatory nonfit (Higgins, 2000, 2005; Higgins & Spiegel, 2004). A postulate of regulatory-fit theory that is strongly supported by existing research is that the more regulatory fit people experience, the higher their motivation during actual or imagined goal pursuit (Forster, Higgins, & Idson, 1998; Freitas et al., 2002; Higgins et al., 2003; Idson, Liberman, & Higgins, 2004; Shah, Higgins, & Friedman, 1998; Spiegel, Grant-Pillow, & Higgins, 2004). This effect of regulatory fit on motivational intensity appears to be independent of the valence of the outcome, mood, perceived effectiveness of the goal-pursuit strategy, and recalled ex-

pectations of success (Forster et al., 1998; Higgins et al., 2003; Idson et al., 2004; Shah et al., 1998; Spiegel et al., 2004). Furthermore, it appears that feelings associated with regulatory fit might mediate the positive effect of regulatory fit on motivation (Freitas et al., 2002; Higgins, 2000). The possibility that regulatory fit could diminish motivation relative to regulatory nonfit under any circumstances is neither mentioned by regulatory-fit theory nor evident in the results of currently published research. However, if people can use regulatory-fit feelings as information for evaluative judgments, they should be able to use them as input for stop rules as well. Therefore, we should see that the effect of regulatory fit on motivation depends on the context. When people are in taskfocusing situations, they should tend to use enjoyment stop rules, in which case regulatory fit (which feels right) should indicate that the task is enjoyable and be more motivating than regulatory nonfit (which feels wrong). Such task-focusing situations might include those in which regulatory fit has been found to enhance motivation or enjoyment relative to regulatory nonfit: doing solvable anagrams (Forster et al., 1998; Shah et al., 1998), decrypting messages or doing simple arithmetic problems (Freitas et al., 2002), generating strategies for improving experiences in middle school (Higgins et al., 2003), writing and turning in a report about how one spent a Saturday (Spiegel et al., 2004), and eating more fruits and vegetables (Spiegel et al., 2004). When people are in a context in which they have a sufficiency stop rule, however, we should see a reversal of the typical pattern of higher motivational intensity under regulatory fit than under regulatory nonfit. For example, when generating words from the letters of longer words under a sufficiency stop rule (e.g., “Continue until you cannot think of any more”), feelings of rightness from regulatory fit should suggest that the goal is successfully attained, whereas feelings of wrongness from regulatory nonfit should suggest that it is not. People experiencing regulatory nonfit in this situation should generate more words than those experiencing regulatory fit. These predictions are quite similar to what we would predict for positive and negative moods: greater motivation with positive mood under an enjoyment stop rule, and greater motivation with negative mood under a sufficiency stop rule (Hirt et al., 1996; Martin & Stoner, 1996; Martin et al., 1993; Sanna et al., 2003, 1996; Startup & Davey, 2001). However, although positive mood and feeling right from regulatory fit are more desirable feelings than negative mood and feeling wrong from regulatory nonfit, mood is not the same thing as regulatory fit (Higgins, 2000). In previous research, mood has not accounted for regulatory-fit effects on evaluative judgments or on motivation (Camacho et al., 2003; Cesario et al., 2004; Forster et al., 1998; Higgins et al., 2003; Shah et al., 1998; Vaughn et al., in press). Thus, mood should not account for the motivational effects of using regulatory-fit feelings as input for stop rules. Additional evidence for the context dependency of regulatory-fit effects on motivation would emerge if drawing attention to an earlier event, such as a source of rightness feelings, eliminated regulatory-fit effects on motivation in a subsequent task. Research suggests that feelings can serve as information for judgments if (a) one explicitly or implicitly asks, “How do I feel about it?,” (b) one cannot distinguish between preexisting feelings and reactions to the judgment target, (c) the feelings seem appropriate to the judgment, and (d) one cannot attribute the feelings to another

REGULATORY FIT AS INPUT

source (e.g., Cesario et al., 2004; Schwarz & Clore, 1983, 1996). If regulatory fit is varied in an initial task, and people are confused about the source of their regulatory-fit feelings, this source confusion should allow them to use the feelings from the initial task as input for decisions about when to stop working on a subsequent task. However, if we draw participants’ attention to the regulatoryfit manipulation as a cause of those feelings, it should reduce source confusion and eliminate the motivational impact of regulatory-fit feelings in the later task. The regulatory-fit experience would no longer be relevant because its source was an initial event independent of the current activity.

The Current Research In the three experiments we present below, we seek to demonstrate that the effect of regulatory fit on motivation is malleable depending on the stop rule presented and that this effect is attributable to use of regulatory-fit feelings as information for the stop rule. In Experiment 1, we show that when no decision rule is presented to participants, the typical positive effect of regulatory fit on motivation is replicated in a word-listing task (Martin et al., 1993). In addition, we show that drawing participants’ attention to an initial event as a source of regulatory-fit feelings eliminates this effect, supporting a feelings-as-information interpretation. In Experiment 2, we show that when participants receive an enjoyment stop rule, they list more examples of objects if they experience regulatory fit rather than nonfit, and that the effect reverses when participants receive a sufficiency stop rule. In Experiment 3, we replicate the findings of Experiment 2 with a word-generation task. In addition, we show that drawing participants’ attention to an initial event as a source of regulatory-fit feelings eliminates this effect, again supporting a feelings-as-information interpretation. Together, the results of these experiments provide support for a general model that extends the implications of regulatory-fit theory to motivation in performance-focusing situations and predicts when regulatory fit will be less motivating than regulatory nonfit.

Experiment 1 We had several goals in this experiment. One was to examine whether the typical positive effect of regulatory fit on motivation replicates with a word-generation task in the absence of an explicit stop rule. We expected that Martin et al.’s (1993) bird-listing activity would be task focusing enough to activate an implicit enjoyment stop rule, as it appears to have been in their research. Another goal was to test the hypothesis that drawing participants’ attention to an earlier event as a source of regulatory-fit feelings eliminates the regulatory-fit effect on motivation. In addition, we sought evidence for the hypothesis that mood would not account for the anticipated effect of regulatory fit on motivation (Forster et al., 1998; Shah et al., 1998). To test these hypotheses, we set up a multitask study in which we manipulated regulatory fit and measured mood in an initial task. Then we drew some participants’ attention to the regulatoryfit manipulation as a source of rightness feelings. Finally, we assessed motivation in a subsequent word-generation task.

Method Participants and Design Ninety-eight undergraduate students participated in the study for extra credit in their psychology courses. They were randomly assigned to regulatory fit (fit

603

vs. nonfit) and attention (attention drawn to the true source of rightness feelings vs. no attention) conditions. We excluded an outlier (in the fit–noattention condition) who was more than 6 standard deviations above the mean of listed birds. The final sample size thus was 97 (22 men, 55 women, and 20 with gender unrecorded because of a programming error).

Procedure Students participated in groups of 2 to 15 in a computer lab, with at least one seat separating each student from the next. Participants learned that they would complete several different tasks on the Web. Regulatory-fit manipulation. The first section of the questionnaire was titled “Hopes and Aspirations” (or “Duties and Obligations”). After reading a brief introduction stating that this part of the questionnaire was about students’ goals at this time of the semester and after answering two questions about their year in college and their age, participants completed a regulatory-fit manipulation developed by Freitas and Higgins (2002). Regulatory-fit conditions paired prevention-focused goals with vigilant strategies or promotion-focused goals with eager strategies. Regulatorynonfit conditions paired prevention-focused goals with eager strategies or promotion-focused goals with vigilant strategies. The promotion orientation version (titled “Hopes and Aspirations”) asked participants to “Please think about something you ideally would like to do. In other words, please think about a hope or aspiration you currently have. Please list the hope or aspiration in the space below.” The prevention orientation version (titled “Duties and Obligations”) asked participants to “Please think about something you believe you ought to do. In other words, please think about a duty or obligation you currently have. Please list the duty or obligation in the space below.” Then, we asked participants to list five strategies for achieving that goal. Specifically, in the eager-strategies condition, we asked participants to “Please list some strategies you could use to make sure everything goes right and helps you realize your hope or aspiration [duty or obligation].” In the vigilant-strategies condition, we asked participants to “Please list some strategies you could use to avoid anything that could go wrong and stop you from realizing your duty or obligation [hope or aspiration].” Participants completed the task twice for two different goals, keeping the orientation and strategy type consistent. Mood measures. After participants reported their first goal and their five strategies to attain it, they completed the first set of mood measures. Participants read that we were interested in learning more about the current duty or obligation (or hope or aspiration) they had just listed (i.e., not the individual strategies, but the duty or obligation [or hope or aspiration] itself). Then they were asked to report how happy, relaxed, and good they felt when pursuing that goal, on scales ranging from 1 (not at all) to 7 (extremely). We repeated the instructions and mood measures after the second goal and strategy list. In other words, we assessed participants’ mood between the regulatory-fit manipulation and the focal bird-listing task. To avoid raising suspicion, we did not ask a more direct question about mood (e.g., “What is your current mood?”) after each goal and strategy list. Attention manipulation. Next, we directed some participants’ attention to the true source of their feelings of rightness with attention instructions developed by Cesario et al. (2004). Participants read that “Sometimes thinking about using the right means to attain each goal can make people ‘feel right’ about their goal pursuit. On the following scale, indicate how much you ‘feel right’ about your goal pursuit.” The scale ranged from 1 (not at all) to 6 (extremely). Participants in the no-attention condition went straight from the regulatory-fit manipulation to the filler task. Filler task. In a 3-min “Consumer Survey,” participants reported their favorite brand of various types of products (toothpaste, shampoo, fast food, soft drinks) and why they preferred that brand (price, quality, or other; Vaughn & Weary, 2003). The purpose of this task was to put some time between the regulatory-fit manipulation and the target judgments. Pilot studies indicated that, possibly because all questions were presented on a single questionnaire via the Web, participants discounted their regulatory-fit experience if the

VAUGHN, MALIK, SCHWARTZ, PETKOVA, AND TRUDEAU

604

fit feelings vs. no attention) ANOVA revealed no significant effects on the mood index (all ps ⬎ .23). Furthermore, treating mood as a covariate in the Regulatory Fit ⫻ Attention analysis of the number of birds listed revealed no significant effect for mood, F(1, 92) ⫽ 2.74, p ⬎ .10, and no change in the effects of regulatory fit, the attention manipulation, or their interaction, F(1, 92) ⫽ 6.71, p ⫽ .01. Overall, participants reported feeling good (M ⫽ 4.82, SD ⫽ 1.03, on the 7-point scale).3

Discussion

Figure 1. Number of birds listed as a function of regulatory-fit condition and attention condition (Experiment 1).

regulatory-fit manipulation immediately preceded the judgment task (also see Martin, Abend, Sedikides, & Green, 1997; Martin et al., 1993; McFarland & Buehler, 1998; McFarland, White, & Newth, 2003). Bird-listing task. Participants then completed a section entitled “Information That Comes to Mind,” which contained a bird-listing task similar to that used by Martin et al. (1993). Participants read that, in this task, our interest was in learning about things that come to people’s minds (specifically, types of birds that come to mind). The dependent variable was the number of birds participants listed, defined broadly as any general or specific type of birds participants wrote in the text box we provided (e.g., penguin, robin red breast, songbird).

Results We expected to find that the typical effect of regulatory fit on motivation would replicate with the listing task we used. Specifically, we predicted that participants who experienced regulatory fit in an earlier task would list more birds than participants experiencing regulatory nonfit. Furthermore, we predicted that if the positive effect of regulatory fit on motivation is caused by using rightness feelings as input for an enjoyment stop rule, then drawing people’s attention to an earlier task as a source of regulatory-fit feelings should eliminate the regulatory-fit effect on motivation. As shown in Figure 1, results supported predictions. A 2 (regulatory fit vs. nonfit) ⫻ 2 (attention drawn to the true source of regulatory-fit feelings vs. no attention) analysis of variance (ANOVA) revealed a significant Regulatory Fit ⫻ Attention interaction on the number of birds listed, F(1, 93) ⫽ 6.85, p ⫽ .01. Simple effects analyses explored the nature of this interaction. Participants whose attention was not drawn to an earlier event as a source of regulatory-fit feelings showed the typical positive effect of regulatory fit on motivation, with those experiencing regulatory fit listing more birds (M ⫽ 12.42, SD ⫽ 5.92) than those experiencing regulatory nonfit (M ⫽ 8.96, SD ⫽ 5.47), F(1, 94) ⫽ 4.01, p ⫽ .05. Among participants whose attention was drawn to an earlier event as a source of regulatory-fit feelings, those experiencing regulatory fit listed nonsignificantly fewer birds (M ⫽ 10.41, SD ⫽ 6.37) than those experiencing regulatory nonfit (M ⫽ 13.27, SD ⫽ 5.96), F(1, 94) ⫽ 2.97, p ⬎ .08.1,2 These effects appear not to have been due to mood. Because the six mood items were highly related (Cronbach’s alpha ⫽ .82), we averaged them to create an index of positive mood. A 2 (regulatory fit vs. nonfit) ⫻ 2 (attention drawn to the true source of regulatory-

As predicted, the typical positive effect of regulatory fit on motivation (Forster et al., 1998; Freitas et al., 2002; Higgins et al., 2003; Idson et al., 2004; Shah et al., 1998; Spiegel et al., 2004) was replicated with a word-generation task in the absence of an explicit stop rule. This finding is consistent with the hypothesis that the bird-listing activity is task focusing enough to activate an implicit enjoyment stop rule (Martin et al., 1993). As expected, drawing participants’ attention to an initial regulatory-fit manipulation as a source of rightness feelings eliminated this effect. This finding is consistent with the hypothesis that the effect of regulatory fit on motivation can be explained by use of rightness feelings from regulatory fit as information for decisions about when to stop working on a task. In addition, we found that mood did not account 1

There was a nonsignificant reversal of the regulatory-fit effect in the attention conditions of Experiment 1. Such nonsignificant reversals have occurred with this attention question before (e.g., Cesario et al., 2004) and might reflect overadjustment for regulatory-fit feelings due to overestimation of their possible informational impact. 2 A Prime ⫻ Strategy ⫻ Attention ANOVA on the number of birds generated revealed a significant three-way interaction, F(1, 89) ⫽ 4.83, p ⫽ .03. No-attention participants showed a significant Prime ⫻ Strategy simple interaction, F(1, 94) ⫽ 4.01, p ⫽ .05; regulatory-fit participants (i.e., prevention–vigilant, M ⫽ 12.71, SD ⫽ 7.52, or promotion– eager, M ⫽ 12.08, SD ⫽ 3.55) listed more birds than nonfit participants (i.e., prevention– eager, M ⫽ 10.25, SD ⫽ 6.18, or promotion–vigilant, M ⫽ 8.27, SD ⫽ 5.13). Attention participants showed no significant Prime ⫻ Strategy simple interaction, F(1, 94) ⫽ 2.97, p ⬎ .08 (prevention–vigilant, M ⫽ 9.31, SD ⫽ 6.14; promotion– eager, M ⫽ 12.00, SD ⫽ 6.73; prevention– eager, M ⫽ 9.73, SD ⫽ 5.08; promotion–vigilant, M ⫽ 15.87, SD ⫽ 5.29). The ANOVA also revealed an unexpected Attention ⫻ Prime interaction F(1, 89) ⫽ 5.66, p ⫽ .02. Simple contrasts showed that promotion–attention participants (M ⫽ 14.42, SD ⫽ 6.04) listed significantly more words than either promotion–no-attention participants (M ⫽ 9.96, SD ⫽ 4.82), F(1, 96) ⫽ 7.04, p ⫽ .009, or prevention–attention participants (M ⫽ 9.50, SD ⫽ 5.56), F(1, 96) ⫽ 8.48, p ⫽ .004. Neither simple contrast with prevention–noattention participants (M ⫽ 11.82, SD ⫽ 7.02) was significant ( ps ⬎ .17). 3 In addition, we carried out separate sets of analyses for the index of the three mood measures from the first goal (Cronbach’s alpha ⫽ .83) and for the index of the three mood measures from the second goal (Cronbach’s alpha ⫽ .87). A pair of Attention ⫻ Regulatory Fit ANOVAs revealed no significant effects on the mood index for the first goal (all ps ⬎ .58) or on the mood index for the second goal (all ps ⬎ .35). The analysis of covariance (ANCOVA) with the index of mood measures from the first goal revealed that, although there was a significant covariate effect on the number of words generated, F(1, 92) ⫽ 9.42, p ⫽ .003, the Attention ⫻ Fit interaction remained significant, F(1, 92) ⫽ 6.52, p ⫽ .01. The ANCOVA with the index of mood measures from the second goal revealed no significant covariate effect on the number of words generated, F(1, 92) ⫽ 0.11, p ⬎ .74; the Attention ⫻ Fit interaction remained significant, F(1, 92) ⫽ 6.44, p ⫽ .01.

REGULATORY FIT AS INPUT

for this pattern of effects, which is consistent with prior regulatoryfit research showing that mood does not account for regulatory-fit effects on evaluative judgments or on motivation (Camacho et al., 2003; Cesario et al., 2004; Forster et al., 1998; Higgins et al., 2003; Shah et al., 1998; Vaughn et al., in press). Our next objective was to examine whether the typical positive effect of regulatory fit on motivation would replicate if we focused people explicitly on a task by presenting them with an enjoyment stop rule. In addition, we examined whether this pattern would reverse if we focused people explicitly on performance by presenting them with a sufficiency stop rule.

Experiment 2 The primary goal of Experiment 2 was to test the hypothesis that the effect of regulatory fit on motivation depends on the stop rule they receive. When participants explicitly receive an enjoyment stop rule, those experiencing regulatory fit should expend more effort on an idea-generation task than those experiencing regulatory nonfit. By contrast, when participants explicitly receive a sufficiency stop rule, those experiencing regulatory fit should expend less effort on an idea-generation task than those experiencing regulatory nonfit. In Experiment 2, we used an idea-generation task that would permit varying regulatory fit within the task itself, rather than entirely in an initial task. (In many real-world situations, the experience of regulatory fit probably would come from the process of working on a task itself, rather than carry over from an earlier event.) Participants in Experiment 2 generated ideas about types of food that one could eat more of to achieve good health (consistent with a promotion focus on attaining gains) or that one could avoid to prevent poor health (consistent with a prevention focus on preventing losses). We varied stop rules by asking some participants to continue as long as they felt like it (enjoyment stop rule; Sanna et al., 2003, 1996; Startup & Davey, 2001) or until they could not think of any more examples (sufficiency stop rule; Sanna et al., 2003, 1996; Startup & Davey, 2001).

Method Participants and Design One hundred twenty-nine students participated in the study for extra credit in their psychology courses. They were randomly assigned to regulatory fit (fit vs. nonfit) and stop rule (enjoyment vs. sufficiency) conditions. Data from two participants were excluded for not following instructions (e.g., listing non-food items like “rocks, glass. . .”). In addition, we excluded 3 outliers (2 in nonfit– enjoyment, and 1 in fit–sufficiency conditions) who were 3–5 standard deviations above the mean of foods listed. This resulted in a final sample of 124 participants (33 men, 91 women).

Procedure Participants were run in sessions of 1–15 people in a computer lab where they were seated with at least one computer separating each participant from the next. Participants learned that they would be completing two different tasks on the Web. The tasks were a regulatory-focus prime and a food-listing task. Regulatory-fit conditions consisted of either a promotion prime paired with a listing of foods to eat more of or a prevention prime paired with a listing of foods to avoid. Regulatory-nonfit conditions consisted of either a promotion prime paired with a listing of foods to avoid or a prevention prime paired with

605

a listing of foods to eat more of. The stop-rules manipulation occurred within the instructions for the food-listing task. Regulatory-focus prime. The first section of the questionnaire was titled “Hopes and Aspirations” (or “Duties and Obligations”). After reading a brief introduction stating that this part of the questionnaire was about students’ goals at this time of the semester and after answering two questions about their year in college and their age, participants were asked to list five hopes and aspirations (or duties and obligations) they had. Then they rated how much they ideally would like to achieve (or believed they ought to achieve) each of the five hopes and aspirations (or duties and obligations) they listed, using a 1 (not at all) to 7 (very much) scale for each rating. Food-listing task. Next, participants received a section of the questionnaire entitled, “Information that Comes to Mind.” The introduction to that section stated that students at their college tend to be very health conscious, and “In this task we are interested in learning about foods you can think of. Do not pay attention to what other people are doing, because they are getting different instructions from you. We would like to find out what examples you bring to mind of . . . .” The remainder of the sentence was printed on its own line, in large font, and constituted the manipulation of the type of food to list. In the promotion-fitting condition, participants were asked to list “examples of foods one can eat more of to attain good health.” In the prevention-fitting condition, they were asked to list “examples of foods one could avoid to prevent poor health.” On the next line, also in large font, participants received either an enjoyment stop rule or a sufficiency stop rule. In the sufficiency stop rule condition, they received the following instructions: As you are making your list of foods, ask yourself, “Have I listed as many as I can?” As long as the answer is “no,” continue listing. When the answer becomes “yes,” then stop. There is no objectively best or worst time to stop. Stop when you feel that you have listed as many foods that one can avoid eating in order to prevent poor health [eat more of to attain good health] as you can. In the enjoyment stop rule condition, they received the following instructions: As you are making your list of foods, ask yourself, “Do I feel like continuing with this task?” As long as the answer is “yes,” continue listing. When the answer becomes “no,” then stop. There is no objectively best or worst time to stop. Stop when you feel that you no longer enjoy listing foods that one can avoid eating in order to prevent poor health [can eat more of in order to attain good health]. We provided a text box for participants to list their examples. The dependent variable was operationally defined to be fairly broad. We counted specifically mentioned ingredients and both superordinate and subordinate categories of foods as separate items (e.g., “chemicals – fake sugar, processed, or preserved foods, anything in excess” counted as five items), as long as the specifically mentioned ingredient was sufficient to put that food in the category of foods requested (e.g., “popcorn with butter” was only counted as one food to avoid, not two).

Results We expected to find that the motivational effect of regulatory fit would differ depending on the stop rule in mind. Regulatory fit should be more motivating than regulatory nonfit under an enjoyment stop rule but less motivating than regulatory nonfit under a sufficiency stop rule. As Figure 2 shows, that is what we found. A 2 (regulatory fit vs. nonfit) ⫻ 2 (enjoyment vs. sufficiency stop rule) ANOVA revealed the predicted Regulatory Fit ⫻ Stop Rule interaction on the number of foods brought to mind, F(1, 120) ⫽ 9.73, p ⫽ .002. Analyses of simple effects explored the nature of this interaction. With an enjoyment stop rule, participants experiencing

606

VAUGHN, MALIK, SCHWARTZ, PETKOVA, AND TRUDEAU

Method Participants and Design One hundred eighty-seven students participated in the study for extra credit in their psychology courses. They were randomly assigned to regulatory fit (fit vs. nonfit), attention (attention drawn to the true source of rightness feelings vs. no attention), and stop rule (enjoyment vs. sufficiency) conditions. We excluded the data from 3 participants because of equipment problems. In addition, we excluded an outlier (in the attention– nonfit–sufficiency condition) who was more than 6 standard deviations above the mean of generated words. The final sample consisted of 183 participants (43 men, 140 women).

Figure 2. Number of foods listed as a function of regulatory-fit condition and stop-rule condition (Experiment 2).

regulatory fit listed significantly more foods (M ⫽ 10.15, SD ⫽ 4.71) than those experiencing regulatory nonfit (M ⫽ 7.57, SD ⫽ 3.63), F(1, 121) ⫽ 5.58, p ⫽ .02. With a sufficiency stop rule, participants experiencing regulatory fit generated significantly fewer words (M ⫽ 7.93, SD ⫽ 3.59) than those experiencing regulatory nonfit (M ⫽ 10.22, SD ⫽ 5.14), F(1, 121) ⫽ 4.30, p ⫽ .04.4,5

Discussion Results show that the effect of regulatory fit on task effort differs according to stop rule. Regulatory fit appears to be more motivating than regulatory nonfit with an enjoyment stop rule but less motivating than regulatory nonfit with a sufficiency stop rule. These results are consistent with the hypothesis that regulatory-fit feelings can serve as input for decisions about when to stop working on a task. However, because our procedure did not permit assessment of mood before presentation of the stop rule, it is possible that mood could account for these effects. In addition, if use of feelings of rightness from regulatory fit is a process underlying these effects, we should see that these effects are eliminated by drawing participants’ attention to an earlier regulatory-fit manipulation as a source of rightness feelings.

Procedure The procedure was nearly identical to that in Study 1 except for two differences. One was that we explicitly gave participants either an enjoyment or a sufficiency stop rule, and the other was that we used a wordgeneration task rather than a bird-listing task. After the regulatory-fit manipulation, mood measures, attention manipulation, and “Consumer Preferences” filler task, participants completed a section of the questionnaire entitled “Words That Come to Mind” that contained the stop-rules manipulation and word-generation task. The wordgeneration task is similar to a distraction task used by McFarland and Buehler (1998; McFarland et al., 2003), and the stop rules were similar to those used in Experiment 2 (Martin et al., 1993; Sanna et al., 2003, 1996; Startup & Davey, 2001). Upon reaching the section entitled “Words That Come to Mind,” participants read the following introduction: In this task we are interested in learning about things that come to people’s mind. Do not pay attention to what other people are doing, because they are getting different instructions from you. We would like to find out what words people generate from the letters of each of the longer words below. For example, from the letters of the word: Starboard, one could generate words like “star,” “a,” or “dart”. After this introduction, participants received the stop rule. So that participants would read it, the stop rule was displayed in large, bold font on its own line. Participants read, “As you are generating words, ask yourself. . .” Then in the sufficiency stop rule condition, they received the following instructions: “Have I generated as many words as I can?” If the answer is “yes”, then stop. If the answer is “no,” then continue listing. There is no best or worst time to stop. Stop when you feel that you have generated as many words as you can.

Experiment 3 The goal of this experiment was to test the hypothesis that use of regulatory-fit feelings as input for stop rules can account for the interactive effects of stop rules and regulatory fit on motivation. If so, those effects should disappear when participants’ attention is drawn to an earlier task as a source of rightness feelings. In addition, this experiment tested the hypothesis that mood does not account for regulatory-fit effects on motivation (Forster et al., 1998; Shah et al., 1998). As in Study 1, we manipulated regulatory fit and measured mood in an initial task and then drew some participants’ attention to that event as a source of rightness feelings. In a later task, we gave participants either an enjoyment or a sufficiency stop rule to use as they generated shorter words from the letters of longer words. We used a different task than those in Studies 1 and 2 to provide participants with an activity that pilot testing revealed to be highly interesting and meaningful to many participants and to show that our findings can be applied to several different tasks.

4

In addition, the means for the two regulatory-fit groups differed significantly under sufficiency and enjoyment stop rules, F(1, 121) ⫽ 4.17, p ⫽ .04, as did the means for the two regulatory-nonfit groups, F(1, 121) ⫽ 5.76, p ⫽ .02. 5 A Prime ⫻ Strategy ⫻ Stop Rule ANOVA on the number of foods listed revealed only a significant three-way interaction, F(1, 116) ⫽ 10.07, p ⫽ .002. Enjoyment participants showed a significant Prime ⫻ Strategy simple interaction, F(1, 121) ⫽ 5.58, p ⫽ .02; regulatory-fit participants (i.e., prevention– vigilant, M ⫽ 10.36, SD ⫽ 4.92, or promotion– eager, M ⫽ 10.00, SD ⫽ 4.68) listed more foods than nonfit participants (i.e., prevention– eager, M ⫽ 8.00, SD ⫽ 3.74, or promotion–vigilant, M ⫽ 7.13, SD ⫽ 3.58). Sufficiency participants also showed a significant Prime ⫻ Strategy simple interaction, F(1, 121) ⫽ 4.30, p ⫽ .04; nonfit participants (i.e., prevention– eager, M ⫽ 11.62, SD ⫽ 5.85, or promotion–vigilant, M ⫽ 9.22, SD ⫽ 4.47) listed more foods than regulatory-fit participants (i.e., prevention–vigilant, M ⫽ 8.33, SD ⫽ 3.58, or promotion– eager, M ⫽ 7.67, SD ⫽ 3.68).

REGULATORY FIT AS INPUT

607

vs. no attention) ANOVA revealed a significant main effect for stop rule, indicating that participants generated more words under the sufficiency stop rule (M ⫽ 42.27, SD ⫽ 24.33) than under the enjoyment stop rule (M ⫽ 25.81, SD ⫽ 15.59), F(1, 175) ⫽ 37.71, p ⬍ .001. As expected, however, the ANOVA also revealed a significant Regulatory Fit ⫻ Stop Rule ⫻ Attention interaction on the number of words generated, F(1, 175) ⫽ 10.87, p ⫽ .001. Among participants who received an enjoyment stop rule and whose attention was not drawn to the true source of their regulatory-fit feelings, those experiencing regulatory fit generated significantly more words (M ⫽ 31.48, SD ⫽ 14.10) than those experiencing regulatory nonfit (M ⫽ 23.00, SD ⫽ 14.77), t(45) ⫽ ⫺2.01, p ⫽ .05. By contrast, when participants’ attention was drawn to the true source of their regulatory-fit feelings, this pattern was not significant; those experiencing regulatory fit generated nonsignificantly fewer words (M ⫽ 22.55, SD ⫽ 15.41) than those experiencing regulatory nonfit (M ⫽ 26.17, SD ⫽ 17.23), t(44) ⫽ 0.75, p ⬎ .45. Among participants who received a sufficiency stop rule and whose attention was not drawn to the true source of their regulatory-fit feelings, those experiencing regulatory fit generated significantly fewer words (M ⫽ 36.67, SD ⫽ 21.04) than those experiencing regulatory nonfit (M ⫽ 51.79, SD ⫽ 22.09), t(46) ⫽ 2.43, p ⫽ .02. By contrast, when participants’ attention was drawn to the true source of their regulatory-fit feelings, this pattern was not significant; those experiencing regulatory fit generated nonsignificantly more words (M ⫽ 44.38, SD ⫽ 21.62) than those experiencing regulatory nonfit (M ⫽ 36.24, SD ⫽ 16.54), t(40) ⫽ ⫺1.37, p ⬎ .17.6,7 Figure 3. Number of words generated as a function of regulatory-fit condition, attention condition, and stop-rule condition (Experiment 3; top panel represents enjoyment stop-rule conditions and bottom panel represents sufficiency stop-rule conditions).

In the enjoyment stop rule condition, they received the following instructions: “Do I feel like continuing with this task?” As long as the answer is “yes,” continue listing. When the answer becomes “no,” then stop. There is no best or worst time to stop. Stop when you feel that you no longer enjoy generating words. Participants generated sets of shorter words from the letters of each of the following words: “artichoke,” “archaeology,” “insurance,” “introduction,” and “topographic.” We provided a text box beside each word for that purpose. The dependent variable was the number of words generated, defined broadly to include incorrectly spelled words (e.g., “nuse”) but not explanations (e.g., “none,” or “I can’t think of any more”).

Results We expected to find that the effect of regulatory fit on motivation would differ depending on the stop rule, as in Experiment 2. In addition, we expected that drawing participants’ attention to an earlier task as a source of regulatory-fit feelings would cause the interactive effect of regulatory fit and stop rules to disappear. As Figure 3 shows, this is what we found. A 2 (regulatory fit vs. nonfit) ⫻ 2 (enjoyment vs. sufficiency stop rule) ⫻ 2 (attention to the true source of regulatory-fit feelings

6 In Experiment 3, the within-group variances were heterogeneous. Therefore, we used independent-sample t tests to examine differences between pairs of means, so that error terms would only be based on the within-group variability of the observations being directly compared (Keppel, 1991). 7 A Prime ⫻ Strategy ⫻ Attention ⫻ Stop Rule ANOVA on the number of words generated revealed a significant main effect for stop rule, F(1, 167) ⫽ 36.53, p ⬍ .001. The ANOVA also revealed a significant four-way interaction, F(1, 167) ⫽ 7.53, p ⫽ .007. We examined simple interactions within the sufficiency and enjoyment conditions, because scores were considerably higher and more variable in the sufficiency conditions (Keppel, 1991). In the no-attention–sufficiency conditions, the Regulatory Focus ⫻ Strategy simple interaction was significant, F(1, 87) ⫽ 6.53, p ⫽ .01; regulatory-fit participants (i.e., prevention-focused–vigilant, M ⫽ 47.56, SD ⫽ 28.09, or promotion– eager, M ⫽ 30.13, SD ⫽ 12.43) generated fewer words than nonfit participants (i.e., prevention– eager, M ⫽ 56.67, SD ⫽ 21.80, or promotion–vigilant, M ⫽ 43.67, SD ⫽ 21.27). In the attention–sufficiency conditions, this pattern did not approach significance, F(1, 87) ⫽ 1.66, p ⬎ .20 (prevention–vigilant, M ⫽ 39.64, SD ⫽ 10.80; promotion– eager, M ⫽ 49.60, SD ⫽ 29.17; prevention– eager, M ⫽ 34.46, SD ⫽ 12.87; promotion–vigilant, M ⫽ 39.13, SD ⫽ 21.96). In the no-attention– enjoyment conditions, the Regulatory Focus ⫻ Strategy simple interaction was marginally significant, F(1, 90) ⫽ 3.50, p ⫽ .065; regulatory-fit participants (i.e., prevention–vigilant, M ⫽ 31.07, SD ⫽ 16.65, or promotion– eager, M ⫽ 32.25, SD ⫽ 8.31) generated more words than nonfit participants (i.e., prevention– eager, M ⫽ 25.89, SD ⫽ 21.03, or promotion–vigilant, M ⫽ 21.27, SD ⫽ 9.84). In the attention– enjoyment conditions, this pattern did not approach significance, F(1, 90) ⫽ 0.59, p ⬎ .44 (prevention–vigilant, M ⫽ 21.80, SD ⫽ 15.26; promotion– eager, M ⫽ 24.14, SD ⫽ 16.83; prevention– eager, M ⫽ 27.80, SD ⫽ 19.28; promotion–vigilant, M ⫽ 23.44, SD ⫽ 13.78).

608

VAUGHN, MALIK, SCHWARTZ, PETKOVA, AND TRUDEAU

Analyses of Mood These effects appear not to have been due to participants using their mood as input for stop rules. Because the six mood items were highly related (Cronbach’s alpha ⫽ .75), we averaged them to form an index of positive mood. A 2 (regulatory fit vs. nonfit) ⫻ 2 (enjoyment vs. sufficiency stop rule) ⫻ 2 (attention drawn to the true source of regulatory-fit feelings vs. no attention) ANOVA revealed no significant effects on the mood index (all ps ⬎ .27). In addition, treating mood as a covariate in the Regulatory Fit ⫻ Stop Rules ⫻ Attention analysis of the number of words generated revealed no significant effect for mood, F(1, 174) ⫽ 0.74, p ⬎ .39, and no change in the effects of regulatory fit, attention, or their interaction: stop-rule main effect, F(1, 174) ⫽ 38.24, p ⬍ 001; interaction, F(1, 174) ⫽ 11.25, p ⫽ .001. Overall, participants reported feeling good (M ⫽ 4.60, SD ⫽ 1.00, on the 7-point scale).8

Discussion Results supported the hypothesis that regulatory fit enhances motivation relative to regulatory nonfit when people have an enjoyment stop rule, but that it diminishes motivation relative to regulatory nonfit when people have a sufficiency stop rule. Moreover, the effects of regulatory fit under each stop rule were eliminated by drawing participants’ attention to an earlier task as a

The ANOVA also revealed an unexpected Attention ⫻ Prime interaction, F(1, 167) ⫽ 4.45, p ⫽ .04. Simple contrasts showed that prevention– no-attention participants (M ⫽ 41.19, SD ⫽ 24.40) generated more words than either promotion–no-attention participants (M ⫽ 30.26, SD ⫽ 15.10), F(1, 181) ⫽ 5.62, p ⫽ .02, or prevention–attention participants (M ⫽ 30.15, SD ⫽ 16.24), F(1, 181) ⫽ 3.34, p ⫽ .07. Neither simple contrast with promotion–no-attention participants (M ⫽ 34.97, SD ⫽ 23.71) approached significance ( ps ⬎ .19). This interaction might seem to replicate the Attention ⫻ Prime interaction in Experiment 1 (see Footnote 2). However, in Experiment 1, the Attention ⫻ Prime interaction occurred within a default enjoyment stop rule, and in Experiment 3, an Attention ⫻ Prime interaction occurred when averaging across stop rules. The ANOVA also revealed a significant Attention ⫻ Prime ⫻ Stop Rule interaction, F(1, 167) ⫽ 3.92, p ⫽ .05. Among enjoyment conditions, there were no significant differences (promotion–no-attention, M ⫽ 25.09, SD ⫽ 10.59; prevention–no-attention, M ⫽ 29.13, SD ⫽ 18.15; promotion– attention, M ⫽ 23.75, SD ⫽ 14.64; prevention–attention, M ⫽ 24.80, SD ⫽ 17.35; all ps ⬎ .32). Among sufficiency conditions (promotion–no attention, M ⫽ 35.21, SD ⫽ 17.21; prevention–no-attention, M ⫽ 53.25, SD ⫽ 24.17; promotion–attention, M ⫽ 44.94, SD ⫽ 26.04; prevention–attention, M ⫽ 36.83, SD ⫽ 12.02), prevention–no-attention participants listed significantly more words than either promotion–no-attention participants, t(46) ⫽ 2.98, p ⫽ .005, or prevention–attention participants, t(46) ⫽ 2.98, p ⫽ .005. No other differences reached significance ( ps ⬎ .15). It appears to be sufficiency–no-attention–prevention participants who contributed most to the Attention ⫻ Prime interaction in Experiment 3. Because the Attention ⫻ Prime interaction in Experiment 1 took place within a default enjoyment stop rule, the most appropriate comparison is between Experiment 1 and the enjoyment conditions of Experiment 3. In Experiment 1, promotion–attention participants provided an anomalously large number of responses. There is no trace of such a pattern among promotion–attention– enjoyment participants in Experiment 3. These effects involving focus and attention remain theoretically unexplained.

source of regulatory-fit feelings, which rendered them irrelevant for use as input for stop rules. This study also supported the hypothesis that mood does not account for this pattern of effects, replicating what we found in Experiment 1. This finding is consistent with prior research, which has shown that mood (measured in various ways) does not account for regulatory-fit effects on evaluative judgments or motivation (Camacho et al., 2003; Cesario et al., 2004; Forster et al., 1998; Higgins et al., 2003; Shah et al., 1998; Vaughn et al., in press).

General Discussion The goal of the current research was to examine how regulatory fit can influence motivation when people are focused on enjoying a task or on sufficiently meeting a performance criterion. When participants either received no stop rule (Experiment 1) or an enjoyment stop rule (Experiments 2 and 3), they accomplished more in a task when they experienced regulatory fit than when they experienced regulatory nonfit. By contrast, when participants received a sufficiency stop rule, they accomplished less in a task when they experienced regulatory fit than when they experienced regulatory nonfit (Experiments 2 and 3). The process underlying these results appears to be use of regulatory-fit feelings as information for decisions about whether to continue. Drawing participants’ attention to an earlier regulatory-fit manipulation as a source of those feelings eliminated the effects of regulatory fit on participants’ productivity in a later task (Experiments 1 and 3). Apparently, the attention focusing procedure eliminated confusion about the source of those feelings and rendered them irrelevant for decisions about when to stop working on the task. This research contributes to regulatory-fit theory by providing evidence that the motivational effects of regulatory fit are context dependent and that they can be explained by use of regulatory-fit feelings as input for stop rules. Together, the results of these experiments offer support for a general model that extends the implications of regulatory-fit theory to motivation in explicitly performance-focusing situations and predicts when regulatory fit will be more or less motivating than regulatory nonfit. If one is focused on task enjoyment, regulatory-fit feelings of rightness can indicate that one is enjoying the task (Freitas & Higgins, 2002; Freitas et al., 2002) and enhance motivation relative to regulatory-nonfit feelings of wrongness. By contrast, if one is focused on attaining a sufficiently adequate performance, regulatory-fit feelings of rightness can indicate 8 In addition, we carried out separate sets of analyses for the index of the three mood measures from the first goal (Cronbach’s alpha ⫽ .78) and for the index of the three mood measures from the second goal (Cronbach’s alpha ⫽ .86). A pair of Attention ⫻ Regulatory Fit ANOVAs revealed no significant effects on the mood index for the first goal (all ps ⬎ .20) or on the mood index for the second goal (all ps ⬎ .17). The ANCOVA with the index of mood measures from the first goal revealed that there was no significant covariate effect on the number of words generated, F(1, 174) ⫽ 0.07, p ⬎ .78; the stop-rules main effect remained significant, F(1, 174) ⫽ 37.40, p ⬍ .001, as did the Attention ⫻ Stop Rules ⫻ Fit interaction, F(1, 174) ⫽ 10.77, p ⫽ .001. The ANCOVA with the index of mood measures from the second goal revealed no significant covariate effect on the number of words generated, F(1, 174) ⫽ 2.32, p ⬎ .12; the stop-rules main effect remained significant, F(1, 174) ⫽ 40.12, p ⬍ .001, as did the Attention ⫻ Stop Rules ⫻ Fit interaction, F(1, 174) ⫽ 11.33, p ⫽ .001.

REGULATORY FIT AS INPUT

that one has done well enough and reduce motivation relative to regulatory-nonfit feelings of wrongness. Several caveats are in order. Although a sufficient explanation for the effect of regulatory fit on motivation is use of rightness feelings (from promotion or prevention regulatory fit) as information for stop rules, the current research does not indicate that such a process is necessary for the effect of regulatory fit on motivation to occur. There are various other feelings associated with regulatory fit (in addition to feelings of rightness) that could be used as information for decisions about when to stop working on a task, including, but not limited to, importance (Camacho et al., 2003; Higgins et al., 2003), perceptual fluency (Lee & Aaker, 2004), and confidence (Cesario et al., 2004). It is possible that any feelings associated with regulatory fit could serve as input for stop rules. In addition, fit between strategies of goal pursuit and other regulatory orientations or task instructions (besides prevention and promotion; Avnet & Higgins, 2003; Bianco, Higgins, & Klem, 2003) could serve as input for enjoyment or sufficiency stop rules and produce similar results. Furthermore, although we neither set out to manipulate mood nor found that mood accounted for the current set of results, there could be times that mood and regulatory-fit feelings have additive or interactive effects on motivation when used as input for stop rules. For example, if mood and regulatory fit are independently manipulated in the same study, mood might have more powerful effects on motivation when interpreted via mood-related stop rules (e.g., “Keep going until you feel good about stopping”), and regulatory fit might have more powerful effects on motivation when interpreted via regulatory-fit-related stop rules (e.g., “Keep going until you feel right about stopping”). Moreover, and perhaps most importantly, it is possible that there are certain situations in which the regulatory-fit effect on motivation is not based on use of feelings as information for decisions about when to stop or to continue. Future research would need to identify such situations and provide evidence for other processes underlying regulatory-fit effects on motivation. The current research also expands the “mood-as-input” extension of the feelings-as-information approach (e.g., Martin et al., 1997; Martin & Stoner, 1996; Martin et al., 1993; Martin & Whitaker, 2000) by providing evidence that regulatory-fit feelings can be used as input for stop rules. As is the case with mood (e.g., Martin et al., 1993) and accessibility experiences (Martin & Whitaker, 2000), regulatory-fit feelings appear to have context dependent implications. This finding is consistent with the apparent use of regulatory-fit feelings as input for evaluative judgments of persuasive communications (Cesario et al., 2004). When Cesario et al. directed participants to evaluate the persuasiveness of the message they read, feelings of rightness from regulatory fit apparently validated the participants’ positive or negative thoughts about the message. Participants experiencing regulatory fit were more persuaded when they had more positive thoughts about a message and were less persuaded when they had more negative thoughts about a message, relative to participants experiencing regulatory nonfit. In other words, the implications of regulatory-fit feelings for persuasion depended in part on participants’ spontaneous interpretations of their rightness feelings in light of their positive or negative reactions to the message. Cesario et al.’s (2004) research suggests that spontaneous interpretations of situations could also influence the effect of regulatory fit on decisions about when to stop working on a task (also see

609

Bohner & Weinerth, 2001; George & Zhou, 2002; Startup & Davey, 2001). It appears that some tasks and situations lend themselves to enjoyment stop rules (Forster et al., 1998; Freitas et al., 2002; Higgins et al., 2003; Shah et al., 1998; Spiegel et al., 2004). Others might lend themselves to sufficiency stop rules. Even without an explicit stop rule, situations that enhance personal relevance or responsibility for one’s judgments or that explicitly activate an accuracy goal might be enough to elicit a performance focus and a sufficiency stop rule (e.g., “Are my judgments sufficiently accurate?”). If so, then compared with feelings of rightness from regulatory fit, feelings of wrongness from regulatory nonfit might suggest that judgments are insufficiently accurate. That could enhance motivation to engage in careful information processing, which can reduce low-effort biases in judgment (e.g., Chaiken, Lilberman, & Eagly, 1989; Fiske & Neuberg, 1990; Petty & Cacioppo, 1986; Weary, Jacobson, Edwards, & Tobin, 2001; Wegener & Petty, 1997; Wilson & Brekke, 1994). For example, in a recent research program, Vaughn et al. (in press) found evidence that, when asked to make sure their judgments are accurate and unbiased, people tend to show less biased judgments when they experienced regulatory nonfit than regulatory fit in a previous task. Vaughn et al. used a correction study procedure in which all participants rated the attractiveness of several highly attractive individuals (biasing context), then some participants received an accuracy motive induction, and then all participants rated the attractiveness of several moderately attractive targets (e.g., Petty & Wegener, 1993; Stapel, Martin, & Schwarz, 1998; Vaughn & Weary, 2003; Wegener & Petty, 1995). When an accuracy motive is not activated, this procedure commonly results in a negative contrast effect on the moderate targets, making them seem worse than without a biasing context. Evidence for bias correction emerges in more positive target judgments. When Vaughn et al. (in press) asked participants to make sure their judgments were accurate and unbiased, participants who had experienced regulatory nonfit in a previous task gave more positive judgments of moderately attractive targets. This pattern of target judgments is the opposite of the value transfer pattern typically seen when participants use regulatory-fit feelings as information for evaluative judgments (namely, more positive judgments under regulatory fit than nonfit; Camacho et al., 2003; Cesario et al., 2004; Freitas & Higgins, 2002; Higgins et al., 2003; Lee & Aaker, 2004). However, it is consistent with more correction for a negative contrast effect under regulatory nonfit. Moreover, the effect of regulatory fit on target judgments was eliminated when researchers drew participants’ attention to an earlier regulatory-fit manipulation as a source of rightness feelings. The results were consistent with the hypothesis that when considering whether one’s judgments are sufficiently accurate, feelings of wrongness from regulatory nonfit can suggest that the answer is no and make correction of judgments for bias more likely. More generally, the current research supports the hypothesis that the effect of regulatory fit on motivation is context dependent. Implications of this finding could extend to a variety of endeavors. Every day we engage in various task-oriented or performanceoriented activities; the current research suggests that the interplay of regulatory fit and stop rules can affect how much effort we put into them. Task-focused activities, such as teaching one’s self how to play the piano, doing a crossword puzzle that is not too difficult or too easy, running for the fun of it, or coming up with ideas for

610

VAUGHN, MALIK, SCHWARTZ, PETKOVA, AND TRUDEAU

a new business venture, might lend themselves to enjoyment stop rules. If so, then experiencing rightness feelings from regulatory fit should suggest that the task is enjoyable and that it is worth continuing, whereas wrongness feelings from regulatory nonfit should suggest the opposite. By contrast, performance-focused activities, such as practicing piano to get a good grade in a class, practicing verbal problems to get a good verbal score on the GRE, running to do well enough to stay on a track team, or selling a sufficient amount of goods and services to ensure a profit, might lend themselves to sufficiency stop rules. If so, then experiencing wrongness feelings from regulatory nonfit should suggest that the task is not yet accomplished and that one needs to continue, whereas rightness feelings from regulatory fit should suggest the opposite.

Conclusions The effect of regulatory fit on motivation depends on the context. When one has an enjoyment stop rule (e.g., “Continue as long as you feel like it”), regulatory fit appears to enhance motivation relative to regulatory nonfit, but when one has a sufficiency stop rule (e.g., “Continue until you cannot do any more”), regulatory fit appears to diminish motivation relative to regulatory nonfit. The current research also provides support for the hypothesis that these effects are due to use of regulatory-fit feelings of rightness as input for stop rules, because drawing attention to an earlier experience as a source of rightness feelings eliminates these effects. In short, although prior research has shown that it is more enjoyable to do things in a way that feels right (Freitas & Higgins, 2002; Freitas et al., 2002), the current research suggests that there are performancefocusing situations in which one may achieve more by doing things in a way that feels wrong.

References Avnet, T., & Higgins, E. T. (2003). Locomotion, assessment, and regulatory fit: Value transfer from “how” to “what.” Journal of Experimental Social Psychology, 39, 525–530. Bianco, A. T., Higgins, E. T., & Klem, A. (2003). How “fun/importance” fit affects performance: Relating implicit theories to instructions. Personality and Social Psychology Bulletin, 29, 1091–1103. Bohner, G., & Weinerth, T. (2001). Negative affect can increase or decrease message scrutiny: The affect interpretation hypothesis. Personality and Social Psychology Bulletin, 27, 1417–1428. Camacho, C. J., Higgins, E. T., & Luger, L. (2003). Moral value transfer from regulatory fit: What feels right is right and what feels wrong is wrong. Journal of Personality and Social Psychology, 84, 498 –510. Cesario, J., Grant, H., & Higgins, E. T. (2004). Regulatory fit and persuasion: Transfer from “feeling right.” Journal of Personality and Social Psychology, 86, 388 – 404. Chaiken, S., Lilberman, A., & Eagly, A. (1989). Heuristic and systematic information processing within and beyond the persuasion context. In J. Uleman & J. Bargh (Eds.), Unintended thought (pp. 212–252). New York: Guilford. Clore, G. L. (1992). Cognitive phenomenology: Feelings and the construction of judgment. In L. L. Martin & A. Tesser (Eds.), The construction of social judgment (pp. 133–163). Hillsdale, NJ: Erlbaum. Crowe, E., & Higgins, E. T. (1997). Regulatory focus and strategic inclinations: Promotion and prevention in decision-making. Organizational Behavior and Human Decision Processes, 69, 117–132. Deci, E. L., & Ryan, R. M. (2000). The “what” and “why” of goal pursuits:

Human needs and the self-determination of behavior. Psychological Inquiry, 11, 227–268. Dweck, C. S., & Leggett, E. L. (1988). A social-cognitive approach to motivation and personality. Psychological Review, 95, 256 –273. Elliott, E. S., & Dweck, C. S. (1988). Goals: An approach to motivation and achievement. Journal of Personality and Social Psychology, 54, 5–12. Fiske, S. T., & Neuberg, S. L. (1990). A continuum of impression formation, from category-based to individuating processes: Influences of information and motivation on attention and interpretation. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 23, pp. 1–74). New York: Academic Press. Forster, J., Higgins, E. T., & Idson, L. C. (1998). Approach and avoidance strength during goal attainment: Regulatory focus and the “goal looms larger” effect. Journal of Personality and Social Psychology, 75, 1115– 1131. Freitas, A. L., & Higgins, E. T. (2002). Enjoying goal-directed action: The role of regulatory fit. Psychological Science, 13, 1– 6. Freitas, A. L., Liberman, N., & Higgins, E. T. (2002). Regulatory fit and resisting temptation during goal pursuit. Journal of Experimental Social Psychology, 38, 291–298. George, J. M., & Zhou, J. (2002). Understanding when bad moods foster creativity and good ones don’t: The role of context and clarity of feelings. Journal of Applied Psychology, 87, 687– 697. Higgins, E. T. (1997). Beyond pleasure and pain. American Psychologist, 52, 1280 –1300. Higgins, E. T. (1998). Promotion and prevention: Regulatory focus as a motivational principle. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 30, pp. 1– 46). New York: Academic Press. Higgins, E. T. (2000). Making a good decision: Value from fit. American Psychologist, 55, 1217–1230. Higgins, E. T. (2005). Value from regulatory fit. Current Directions in Psychological Science, 14, 209 –213. Higgins, E. T., Idson, L. C., Freitas, A. L., Spiegel, S., & Molden, D. C. (2003). Transfer of value from fit. Journal of Personality and Social Psychology, 84, 1140 –1153. Higgins, E. T., & Spiegel, S. (2004). Promotion and prevention strategies for self-regulation: A motivated cognition perspective. In R. F. Baumeister & K. D. Vohs (Eds.), Handbook of self-regulation: Research, theory, and applications. New York: Guilford. Hirt, E. R., Melton, R. J., McDonald, H. E., & Harackiewicz, J. M. (1996). Processing goals, task interest, and the mood-performance relationship: A mediational analysis. Journal of Personality and Social Psychology, 71, 245–261. Idson, L. C., Liberman, N., & Higgins, E. T. (2004). Imagining how you’d feel: The role of motivational experiences from regulatory fit. Personality and Social Psychology Bulletin, 30, 926 –937. Keppel, G. (1991). Design and analysis: A researcher’s handbook (3rd ed.). Englewood Cliffs, NJ: Prentice Hall. Lee, A. Y., & Aaker, J. L. (2004). Bringing the frame into focus: The influence of regulatory fit on processing fluency and persuasion. Journal of Personality and Social Psychology, 86, 205–218. Martin, L. L., Abend, T., Sedikides, C., & Green, J. D. (1997). How would I feel if. . .? Mood as input to a role fulfillment evaluation process. Journal of Personality and Social Psychology, 73, 242–253. Martin, L. L., & Stoner, P. (1996). Mood as input: What we think about how we feel determines how we think. In L. L. Martin & A. Tesser (Eds.), Striving and feeling: Interactions among goals, affect, and selfregulation (pp. 279 –301). Mahwah, NJ: Erlbaum. Martin, L. L., Ward, D. W., Achee, J. W., & Wyer, R. S. (1993). Mood as input: People have to interpret the motivational implications of their moods. Journal of Personality and Social Psychology, 64, 317–326. Martin, L. L., & Whitaker (2000). Availability as input: The experience of cognitive effort can either strengthen or weaken evaluations. In H. Bless

REGULATORY FIT AS INPUT & J. P. Forgas (Eds.), The message within: The role of subjective experience in social cognition and behavior. (pp. 88 –106). New York: Psychology Press. McFarland, C., & Buehler, R. (1998). The impact of negative affect on autobiographical memory: The role of self-focused attention to moods. Journal of Personality and Social Psychology, 75, 1424 –1440. McFarland, C., White, K., & Newth, S. (2003). Mood acknowledgement and correction for the mood-congruency bias in social judgment. Journal of Experimental Social Psychology, 39, 483– 491. Nicholls, J. G. (1984). Achievement motivation: Conceptions of ability, subjective experience, task choice, and performance. Psychological Review, 91, 328 –346. Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 19, pp. 123–205). New York: Academic Press. Petty, R. E., & Wegener, D. T. (1993). Flexible correction processes in social judgment: Correcting for context-induced contrast. Journal of Experimental Social Psychology, 29, 137–165. Sanna, L. J., Parks, C. D., & Chang, E. C. (2003). Mixed-motive conflict in social dilemmas: Mood as input to competitive and cooperative goals. Group Dynamics: Theory, Research, and Practice, 7, 26 – 40. Sanna, L. J., Turley, K. J., & Mark, M. M. (1996). Expected evaluation, goals, and performance: Mood as input. Personality and Social Psychology Bulletin, 22, 323–335. Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judgments of well-being: Informative and directive functions of affective states. Journal of Personality and Social Psychology, 45, 513–523. Schwarz, N., & Clore, G. L. (1996). Feelings and phenomenal experiences. In E. T. Higgins & A. Kruglanski (Eds.), Social psychology: A handbook of basic principles. New York: Guilford. Shah, J., Higgins, E. T., & Friedman, R. S. (1998). Performance incentives and means: How regulatory focus influences goal attainment. Journal of Personality and Social Psychology, 74, 285–293.

611

Spiegel, S., Grant-Pillow, H., & Higgins, E. T. (2004). How regulatory fit enhances motivational strength during goal pursuit. European Journal of Social Psychology, 34, 39 –54. Stapel, D. A., Martin, L. L., & Schwarz, N. (1998). The smell of bias: What instigates correction processes in social judgments? Personality and Social Psychology Bulletin, 24, 797– 806. Startup, H. M., & Davey, G. C. L. (2001). Mood as input and catastrophic worrying. Journal of Abnormal Psychology, 110, 83–96. Utman, C. H. (1997). Performance effects of motivational state: A metaanalysis. Personality and Social Psychology Review, 1, 170 –182. Vaughn, L. A., O’Rourke, T., Schwartz, S., Malik, J., Petkova, Z., & Trudeau, L. (in press). When two wrongs can make a right: Regulatory nonfit, bias, and correction of judgments. Journal of Experimental Social Psychology. Vaughn, L. A., & Weary, G. (2003). Causal uncertainty and correction of judgments. Journal of Experimental Social Psychology, 39, 516 –524. Weary, G., Jacobson, J. A., Edwards, J. A., & Tobin, S. J. (2001). Chronic and temporarily activated causal uncertainty beliefs and stereotype usage. Journal of Personality and Social Psychology, 81, 206 –219. Wegener, D. T., & Petty, R. E. (1995). Flexible correction processes in social judgment: The role of naı¨ve theories in corrections for perceived bias. Journal of Personality and Social Psychology, 68, 36 –51. Wegener, D. T., & Petty, R. E. (1997). The flexible correction model: The role of naı¨ve theories of bias in bias correction. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 29, pp. 141–208). New York: Academic Press. Wilson, T. D., & Brekke, N. (1994). Mental contamination and mental correction: Unwanted influences on judgments and evaluations. Psychological Bulletin, 116, 117–142.

Received June 17, 2005 Revision received January 4, 2006 Accepted January 17, 2006 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 612– 625

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.612

See What You Want to See: Motivational Influences on Visual Perception Emily Balcetis and David Dunning Cornell University People’s motivational states—their wishes and preferences—influence their processing of visual stimuli. In 5 studies, participants shown an ambiguous figure (e.g., one that could be seen either as the letter B or the number 13) tended to report seeing the interpretation that assigned them to outcomes they favored. This finding was affirmed by unobtrusive and implicit measures of perception (e.g., eye tracking, lexical decision tasks) and by experimental procedures demonstrating that participants were aware only of the single (usually favored) interpretation they saw at the time they viewed the stimulus. These studies suggest that the impact of motivation on information processing extends down into preconscious processing of stimuli in the visual environment and thus guides what the visual system presents to conscious awareness. Keywords: motivation, visual perception, motivated reasoning, New Look, ambiguous figures

pebble at the bottom of one’s shoe is never nearly the rock it feels like when one steps on it. Moreover, perception is malleable. It is responsive to top-down influences that flow from the perceiver’s cognitive and psychological states or from environments (Henderson & Hollingworth, 1999). To be sure, much of perception is bottom-up, with sense organs and perceptual systems working inflexibly and automatically to form a representation of a stimulus that the perceiver passively accepts. The perceptual system pieces together the finegrained bits of information the senses acquire to create a coherent percept, analyzing and synthesizing basic components of objects (Kosslyn & Koenig, 1992; Michelon & Koenig, 2002), including focal areas, critical features (Long & Olszweski, 1999), fixation points (Meng & Tong, 2004; Toppino, 2003), and spatial proximity or crowding (Pelli, Palomares, & Majaj, 2004). But a substantial volume of psychological research reveals that top-down influences also inform perception. For example, context matters. Prior exposure to images of animals or people biases what people see when they view classic ambiguous figures, such as the rat–man and old woman–young woman figures so often featured in introductory psychology textbooks (Bugelski & Alampay, 1961; Leeper, 1935). Estimates of a man’s walking speed are biased after thinking about fast animals like cheetahs or slow animals like turtles (Aarts & Dijksterhuis, 2002). Interpretations of an ambiguous figure that can be seen as a woman’s face or as a man playing a saxophone depend on whether perceivers have been recently primed with the concepts of “flirtation” or “music” (Balcetis & Dale, 2003). Perceptions of how steep a hill is become more extreme after participants jog vigorously for an hour (Bhalla & Proffitt, 1999). The distance to a goal seems longer if people strap on a heavy backpack (Proffitt, Stefanucci, Banton, & Epstein, 2003). In the current article, we explore one possible top-down influence on perception that has been shown to have a profound and ubiquitous impact in other arenas of social cognition. That influence is the perceiver’s motivational states—more specifically, the motivation to think of one’s self and one’s prospects in a favorable way, to believe that one will achieve positive outcomes while

The world that people know is the one they take in through their senses. This is the world they react to—the one their conscious thoughts, feelings, and actions are predicated on. People act on the presumption that the world they are consciously aware of is a comprehensive and accurate representation of the environment that exactly copies the outside world as it truly is. Decades of research in psychology, however, tend to undermine the assumption that what people see or hear is an exact replica of what is out in the world, in two different ways. First, perception is selective. People are not aware of everything that is going on around them. Consider, for example, recent studies of attentional blindness. Of undergraduates asked to monitor how many times people in a videotape pass a basketball among themselves, 40% failed to see the woman in a gorilla suit saunter into the middle of the group, turn to the camera, beat her chest, and then walk out (Simons & Chabris, 1999). Second, perception is often biased. Hills are not as steep as they appear to be (Bhalla & Proffitt, 1999; Creem & Proffitt, 1998; Proffitt, Creem, & Zosh, 2001). Distances are not as short as they look (Baird & Biersdorf, 1967; Durgin, Proffitt, Olson, & Reinke, 1995; Gilinsky, 1951; Tittle, Todd, Perotti, & Norman, 1995; Todd & Bressan, 1990; Todd & Norman, 1991). Large objects are not as tall as they seem (Yang, Dixon, & Proffitt, 1999). Everyone knows that the speck of a

Emily Balcetis and David Dunning, Department of Psychology, Cornell University. This research was supported financially by National Institute of Mental Health Grant RO1 56072 awarded to David Dunning. We thank Kathy Deng, Chelsea Finn, Agata Gluszek, Sirisha Nandipati, Giorgio Piccoli, Lorraine Ricci, Michael Van Wert, and Steven Zhang for their assistance in conducting the experiments reported in this article as well as Nathan Novemsky for suggesting the procedure used in Study 5. Richard Eibach provided useful commentary on a previous version of this article. Correspondence concerning this article should be addressed to Emily Balcetis who is now at the Department of Psychology, 200 Porter Hall, Ohio University, Athens, OH 45701 or David Dunning, Department of Psychology, Uris Hall, Cornell University, Ithaca, NY 14853. E-mail: [email protected] or [email protected] 612

MOTIVATIONAL INFLUENCES ON VISUAL PERCEPTION

being able to avoid aversive ones, and to enhance self-worth and esteem. This motivation in the psychological literature has several names, such as motivated reasoning, self-affirmation, wishful thinking, and defensive processing, and has been shown to have a widespread influence in shaping how people think about their world, that is, how they interpret information of which they are consciously aware. This motive has been shown to influence such higher order tasks as judging other people, evaluating the self, predicting the future, and making sense of the past (for reviews, see Baumeister & Newman, 1994; Dunning, 2001; Kunda, 1990; Pittman, 1998). In the studies that follow, we examine the scope of motivated reasoning to see if it crosses the boundary between how people think about their outside world and how they perceive it. Certainly, motivated reasoning influences conscious, deliberate, and effortful judgments, but we ask if it can constrain what information reaches consciousness in the first place. Does the impact of motivated reasoning or wishful thinking, more specifically, extend down to preconscious processing of visual information? We test, in essence, whether people literally are prone to see what they want to see.

The Impact of Motivational States There exist some indirect hints that the motives underlying wishful thinking have an impact on visual perception. Recent work focusing on more biologically oriented motivational states shows that they influence the perception of visual stimuli. For example, Changizi and Hall (2001) demonstrated that participants who were thirsty perceived more transparency in ambiguous visual stimuli than did those who were not thirsty, presumably because transparency is a characteristic associated with water. Women during periods of high fertility were faster to categorize male photographs than female ones by gender, relative to those not in such a fertile state (Macrae, Alnwick, Milne, & Schloerscheidt, 2002). It is important that the same comparative enhancement was not present for women taking a contraceptive pill or those who were pregnant (Johnston, Arden, Macrae, & Grace, 2003). Both of these examples suggest an enhanced perceptual sensitivity for features in visual stimuli that are relevant to biological drives or desires. But would a drive toward wishful thinking similarly influence perception? In a sense, this question is a revisiting and a reopening of one of the focal issues of the New Look approach to perception that arose in psychology during the 1940s and 1950s (Bruner & Minturn, 1955). According to New Look theorists, perception was an active and constructive process influenced by many top-down factors. One class of such factors was the needs and values of the perceiver. For example, Bruner and Goodman (1947) asked children in diverse social economic conditions to estimate the size of monetary coins by manipulating the diameter of a beam of light. Poorer children, for whom the value of money was greater, overestimated the size of the coins compared with more affluent children, who were presumed to place less value on the same coins. In studies of perceptual defense, New Look theorists concluded that participants inhibited the recognition of threatening stimuli, such as troubling words (Postman, Bruner, & McGinnies, 1948). These initial demonstrations of motivational influences on perception were met with much enthusiasm, which was then followed

613

by withering criticism. To be sure, much of what the New Look theorists proposed has lasted through today and informs contemporary cognitive and perceptual psychology in fundamental ways. Psychologists uniformly agree with the New Look tenet that much of cognition happens nonconsciously, that is, outside a person’s awareness, monitoring, or control (Greenwald, 1992; Wegner & Bargh, 1998). Many modern textbooks describe the New Look proposal that perception is filtered: that the representation of the environment that people have in consciousness has omitted a good deal of information that is actually in the environment (Allport, 1989; Miller, 1987). Similarly, perception of an object is importantly influenced by the perceiver’s expectations as well as the context surrounding that object (Biederman, Mezzanotte, & Rabinowitz, 1982; Boyce & Pollatsek, 1992; Li & Warren, 2004; Long & Toppino, 2004). However, the specific New Look assertion that motivational states influence perception did not achieve the same stature and longevity as these other insights. It, instead, ran aground in the 1950s on the rocky shoals of methodological difficulties and theoretical controversies (Eriksen, 1958, 1962; Eriksen & Browne, 1956; Goldiamond, 1958; Prentice, 1958; Wohlwill, 1966). Critics pointed out that poorer children might misjudge the size of coins because they were not as familiar with them, or that their misjudgments might involve problems of memory rather than perception (McCurdy, 1956). Critics also noted in studies of perceptual defense that participants might have taken longer to report troubling words not because it took them longer to perceive them but rather because it took longer to get over the surprise of seeing them or the embarrassment of saying them (Erdelyi, 1974, 1985). Others lamented that the relative unfamiliarity of threatening words, and not their motivational punch, was the key ingredient that slowed participants’ recognition responses (Adkins, 1956; Howes & Solomon, 1950). As such, the influence of motivational states on perception was never firmly established. And as the 1950s closed the study of the relation between motivational states and perception, this pursuit fell by the wayside and ceased to have the major impact—if any at all— enjoyed by other insights from the New Look tradition (Dunning, 2001; Erdelyi, 1974; Gilbert, 1998; Jones, 1985; Nisbett & Ross, 1980).

Perception of Ambiguous Figures In the present research, we examined the impact of motivational states on perception by focusing on interpretations of ambiguous or reversible figures—visual stimuli, like the famous Necker cube, that people can interpret in two different ways but for which they tend to see only one interpretation at any given time (Long & Toppino, 2004; Rock & Mitchener, 1992). In each of five studies, we told participants that they were about to be assigned to one of two experimental tasks, one being much more desirable than the other. We also told participants that a computer sitting in front of them was about to present them a stimulus that would indicate which task they were assigned to. In fact, in each study, the computer presented a figure that could be interpreted in two different ways: one way that would assign participants to their favored task and one that would assign them to the opposite. We expected that participants would tend to see the interpretation that assigned them to the outcome they favored.

BALCETIS AND DUNNING

614

Because our experimental stimuli, like much of the contents of our surroundings, lack clarity and contain multiple interpretations, potential interpretations of a visual stimulus can be likened to a hypothesis (Gregory, 1974). Given a constrained set of bottom-up features and top-down influences, the perceptual system considers certain ideas of what an ambiguous stimulus might be and ultimately selects one interpretation. For example, given the distinct features of a four-legged shape in a distant field, one can entertain different hypotheses about the identity of the shape. For example, to test whether the shape is a cow, the perceiver might examine whether the shape has a stocky snout and black spots. Just as expectancies and contexts can suggest a testable perceptual hypothesis, a preference or desire might privilege a favored interpretation or hypothesis over a disfavored one. Wishful thinking might shape the specific hypothesis that individuals test when given such ambiguous information. In particular, the perceiver might scan the visual stimulus in a biased manner, searching for features that match those of the desired animal rather than those that match an undesired one. The net effect of focusing on a hypothesis is that the perceiver tends to seek out information that would confirm it rather than disconfirm it (Pyszczynski & Greenberg, 1987; Sanitioso, Kunda, & Fong, 1990). Alternatively, a motivated preference might lower the threshold required for the visual system to decide it matches the favored interpretation. Other work in motivated reasoning has shown that information consistent with a favored conclusion is held to a lower standard of scrutiny than information consistent with an unwanted one (Dawson, Gilovich, & Regan, 2002; Ditto & Lopez, 1992; Trope & Ferguson, 2001). It could be then that those features most representative of the desired animal category are recognized faster or more easily because the perceiver requires less of a match between what he or she hopes to see and what is offered by the stimulus. The key to whatever process is at play is that it takes place preconsciously. People are not aware that they have selected one interpretation over another. Indeed, they are not even aware of the alternative interpretation. Whatever work the visual system has done to bias the interpretation that people see involves processes below the level of awareness.

Overview of Studies Studies 1 and 2 demonstrated that participants tended to report seeing the interpretation of an ambiguous figure that fit with their wishes and preferences over one that did not. Studies 3 and 4 added implicit measures to ensure that participants truly saw the interpretation they reported rather than simply reporting the preferred interpretation. Study 5 added a procedural twist to affirm that participants saw only the interpretation they usually wanted to see as they viewed the stimulus and that it was not the case that they saw both interpretations and then only reported the favored one. In short, people tended to honestly see only that interpretation that was suggested, in part, by their motivational state.

Study 1: Disambiguating an Ambiguous Figure Study 1 was designed to provide an initial demonstration that wishful thinking could influence the interpretation of an ambiguous stimulus. Participants were brought into the laboratory and told

that they would be assigned to one of two tasks. One was favored (i.e., drinking freshly squeezed orange juice); the other was not (i.e., drinking a noxious-smelling and vile-looking health food drink). They were told that the computer would assign their beverage by presenting either a number or a letter. For roughly half of participants, a letter would indicate that they were assigned to the desirable beverage. For the other half, the reverse was true. However, what the computer flashed very briefly was an ambiguous figure that could be interpreted either as a number or letter. Our prediction was that participants would tend to report seeing the interpretation that offered them the coveted beverage.

Method Participants. Participants were 88 undergraduates at Cornell University who earned extra credit in their psychology or human development courses for taking part in the study. Procedure. In what was advertised as a taste-testing experiment, an experimenter explained that participants would predict taste sensations for two beverages, consume only one beverage, and describe their actual taste sensation of that one beverage. On the table in front of participants sat the two beverages. The first was the desirable one: freshly squeezed orange juice. The second was the less desirable alternative: a gelatinous, chunky, green, foul-smelling, somewhat viscous concoction labeled as an “organic veggie smoothie.”1 The experimenter invited participants first to smell each beverage. Then, participants spent 3 min predicting what they might experience if asked to drink 8 ounces (about 240 ml) of each beverage to heighten the appeal of the orange juice and strengthen their disgust with the veggie smoothie. Participants were seated in front of a 15-in. G3 iBook. The experimenter then explained that a computer program would randomly select a beverage for the participant to consume. Specifically, the computer would select either a single letter or a single number from a set of 26 letters and 26 numbers. Roughly half of the participants, those in the number-desirable condition, were told that if the computer selected a number from the set, they would drink 8 ounces (about 240 ml) of orange juice, and if a letter was selected, they would drink 8 ounces (about 240 ml) of veggie smoothie. The remaining participants in the letter-desirable condition learned that a letter would result in their assignment to the orange juice and a number to the veggie smoothie. After inviting the participant to review these directions on a computer screen, the experimenter stepped away to ostensibly complete some paperwork. Participants focused on the center of the monitor on which was displayed a static fixation point. After 3 s, this fixation point was replaced with an ambiguous figure (1 in. in height, 1 in. in width) that could be interpreted as either the capital letter B or the number 13 (see Figure 1) for 400 ms. The presentation of this figure was followed by a 200-ms mask and then finally by an image that was meant to look as though the computer program had crashed. The experimenter continued to focus on the paperwork until the participant called her attention to the computer crash. The experimenter feigned surprise, exclaimed that “this always happens to old Macs,” and stated that she would have to ask the graduate student she worked for what she should do. As the experimenter approached the door to leave the lab, she asked if the program displayed anything before crashing. At this point, most participants reported whether they saw a B or a 13. If participants did not offer a response, the experimenter asked again if anything was shown or if it immediately crashed. If at this point participants still refused an answer, the experimenter left the room and returned a few minutes later to ask a final time if anything was shown. After receiving an answer, the experimenter handed the participant a questionnaire to complete while she supposedly left to prepare the bever-

1

Recipe available on request.

MOTIVATIONAL INFLUENCES ON VISUAL PERCEPTION

Figure 1.

Ambiguous B–13 figure used in Study 1.

age. This questionnaire probed for suspicion of the purpose of the study, suspicion of the computer crash, and in a funneled manner queried participants to see if they realized the ambiguity in the figure shown before the computer crash.

Results A priori, we established conditions for the inclusion of participants’ data. Participants were excluded if they recognized the figure was ambiguous, were able to explain the purpose of the study in debriefing, or mentioned they wished to be assigned to what was considered by most participants to be the less desired task (i.e., consumption of veggie smoothie). Given these criteria, 15 people were excluded for recognizing the ambiguity in the figure when viewing the figure, 4 for explaining that we were interested in how their desires could influence the way they saw the figure, 3 for stating they hoped to consume the smoothie, and 3 simply refused to participate when they heard that they might be asked to consume the smoothie. This left data from 63 participants for analysis. Although a few participants indicated the computer crash was suspicious, none of these participants were able to describe the purpose of the study or the reason for the crash. Responses from those 63 participants were coded by means of the following method. Reports of the letter B were given a score of ⫹1, and reports of the number 13 a score of ⫺1. Those who did not offer a response or indicated that nothing was shown before the crash received a score of 0. We then subjected these scores to an ordinal logistical regression analysis (the constrained range of the coding system made more usual statistical procedures less appropriate) to see if participants tended to see different interpretations of the ambiguous figure depending on which interpretation was more desirable. As expected, participants’ desire to see either letters or numbers influenced their interpretation of the B–13 ambiguous figure, ␹2(1, N ⫽ 63) ⫽ 23.92, p ⬍ .001. In particular, when hoping to see a letter, 72% (n ⫽ 18) of participants reported seeing the capital letter B, whereas 0% reported seeing a 13. When hoping to see a number, 60.5% (n ⫽ 23) reported seeing a 13 and 23.7% (n ⫽ 9) reported seeing the B. Some people in each condition reported that in fact nothing was shown before the crash (28%, n ⫽ 7, in the letter-favorable condition; 15.8%, n ⫽ 6, in the number-favorable condition).

615

Our specific prediction focuses on the responses of those who offered an interpretation of the figure. When excluding those responses from participants who reported that nothing was shown before the crash, participants’ desire to see either letters or numbers influenced their interpretation of the B–13 ambiguous figure, ␹2(1, N ⫽ 50) ⫽ 23.96, p ⬍ .001. Additionally, we can collapse across the specific character participants were motivated to see and look at just the reported interpretation for those participants who offered one. In fact, 82% (n ⫽ 41) of participants reported the desired interpretation, ␹2(1, N ⫽ 50) ⫽ 20.48, p ⬍ .001. In addition, including those people in the analyses who indicated that the figure was ambiguous does not change this pattern, as similar numbers of participants across both motivational conditions reported the ambiguity of the figure (n ⫽ 8, when hoping to see letters; n ⫽ 7, when hoping to see numbers). That is, we gave a score of 0 to those people who indicated the figure was ambiguous and again conducted an ordinal logistic regression. Still, participants’ desire to see either a letter or a number influenced their interpretation of the ambiguous figure, ␹2(1, N ⫽ 78) ⫽ 22.95, p ⬍ .001.

Discussion In sum, Study 1 provided evidence that people’s motivational states can influence their interpretation of ambiguous objects in their environment. When faced with an ambiguous figure that could be interpreted as either a number or letter, the interpretation that reached consciousness and was reported tended to be the one that placed participants in a desirable circumstance rather than in an unwanted one. However, it is possible that the participants’ responses did not reflect their true percept. Instead of reporting what they saw, they instead just offered a report that assigned them to the orange juice. Put simply, participants may have lied about what they saw. Although we suspect this is not the case, we conducted a follow-up to assess this counterexplanation. In a design similar to Study 1, 28 participants were either motivated to see letters or numbers to avoid the veggie smoothie but were then shown unambiguous figures of B or 13, rather than an ambiguous figure, during the computer assignment process. For half of the participants, a letter assigned them to the orange juice, whereas for the other half a number assigned them to the veggie smoothie. Crossed with this, half of the participants were shown a B and the other half were shown a 13, resulting in a 2 (desired character: letter or number) ⫻ 2 (character shown: B or 13) factorial. The alternative account predicts that participants’ reports of the figure shown to them would be influenced by which character was desired as well as what character was shown to them. However, inconsistent with that account, we found that what participants reported depended only on the character shown to them. In all conditions, 100% of participants (n ⫽ 7 in every cell) reported the actual figure shown, regardless of what figure was shown to them and what participants were motivated to see.

Study 2: Replication Study 2 was designed as a conceptual replication involving a different ambiguous figure and a different procedure. In addition, in Study 1, we noted that a small but notable minority of partici-

616

BALCETIS AND DUNNING

pants was able to spot the ambiguity of the figure we showed them. In Study 2, we used a figure whose ambiguity was more opaque and thus not as likely to be noticed by participants.

Method Participants. Participants were 52 undergraduates at Cornell University who received extra credit in their psychology course for taking part. Procedure. Participants completed a task ostensibly about differences in predictions of and actual taste experiences. The experimenter explained that participants would be experiencing and describing different taste sensations. Participants would predict taste sensations for three food items but actually consume only one of them. First, participants predicted what each of the following items would taste like: a bottle of Aquafina water, a bag of Jelly Belly candies, and a bag of gelatinous and partially liquified canned beans. After participants predicted taste sensations of each item, participants were seated in front of a 17-in. iMac 64 desktop computer. Again, supposedly to eliminate bias from the selection process, a computer program would randomly assign the item participants would consume. The experimenter explained that participants would play a game, and their final score would determine what item was consumed. In this game, the computer displayed pictures of animals worth positive and negative points. On the top of their response sheet was a table listing every animal that could be selected and the specific number of points each animal was worth. For half of the participants, farm animals were worth positive points, whereas sea creatures were worth negative points. For the other half of participants, this was reversed. Black and white drawings of the full bodies, heads, and artistic renditions of animals were displayed in the rounds that preceded the final round. Although the computer would be keeping an ongoing tally of the points accumulated, participants recorded the animal shown to them, the points that animal was worth, and their ongoing score ostensibly to corroborate the computer program. If their score at the end of 15 cards was zero, participants would consume the water. If their score was positive, they would consume the candies, but if their score at the end was negative, participants would consume the canned beans. Although participants were told that the program randomly selected animals from a set of four farm animals and four sea animals, the program was actually rigged such that every participant experienced one of two sequences of animals and point tallies, depending on what category of animal was worth positive point values. As the game progressed, ongoing scores, predetermined and consistent across participants, fluctuated between positive and negative. However, the last three rounds brought increasingly negative point totals. That is, ongoing scores became ever more suggestive that participants would consume the canned beans. Ongoing scores at the end of the penultimate round were such that only one animal was worth enough positive points to be able to pull participants from the negative and bring a positive final score, thus avoiding the canned beans. For half of the participants, this animal was a horse; for the other half, it was a seal. The animal displayed during the final trial was in fact an ambiguous figure (2.75 in. wide, 3.75 in. tall) that could be interpreted as either the head of a horse or the full body of a seal (see Figure 2; from Fisher, 1968). All animals, including the last figure, remained on the screen for 1,000 ms. After the game, participants completed a funneled debriefing that probed for suspicion of the purpose of the study, possible alternate interpretations of the figure, and asked if they had seen the figure before.

Results Given the criteria we established a priori, 5 participants were excluded for articulating the purpose of the study and 4 for mathematical errors that precluded them from desiring the target

Figure 2. Ambiguous horse–seal figure used in Studies 2– 4. From “Ambiguity of Form: Old and New,” by G. H. Fisher, 1968, Perception & Psychophysics, 4, p. 191. Copyright 1968 by the Psychonomics Society. Reprinted with permission.

animal. No one reported seeing both interpretations of the ambiguous figure. These omissions left data from 43 participants for analysis. We used the same type of coding scheme for interpretations as in the previous studies. Given the natural bias of this ambiguous figure was to see a horse, those who reported a horse received a score of ⫹1. Because the less common interpretation of the figure was as a seal, those who reported a seal received a score of ⫺1. Using an ordinal logistic regression, we found that participants’ interpretations depended on what category of animal was worth positive points, ␹2(1, N ⫽ 43) ⫽ 6.89, p ⫽ .009. When hoping to see a horse, 66.7% (n ⫽ 14) of participants saw the figure as a horse, and 33.3% (n ⫽ 7) saw a seal. However, this bias reversed when hoping to see a seal. Only 27.3% (n ⫽ 6) of this group saw a horse, but 72.7% (n ⫽ 16) reported a seal, ␹2(1, N ⫽ 23) ⫽ 6.70, p ⫽ .01.

Discussion In sum, Study 2 replicated the findings of the first study with a different figure and experimental procedure. Participants tended to see the interpretation of the figure that they desired to see, rather than one they wished to avoid. In addition, no participant, either spontaneously or in debriefing, noted the ambiguous nature of the figure they saw. However, a reader can propose one counterexplanation for these findings, one that we decided to test in a control study. Given that the three rounds preceding the ambiguous figure included animals

MOTIVATIONAL INFLUENCES ON VISUAL PERCEPTION

that brought participants’ scores down, it is possible that participants’ expectations about the next type of animal and not their desire predisposed them to see an animal worth positive points. That is, participants fell prey to a gambler’s fallacy, assuming that a run of negative scores made positive-scoring animals more likely to appear next. To test this alternative explanation, we reran a version of Study 2, asking participants to follow along with the computer game and to record their points on a response sheet. However, we made clear to them that they would not be consuming any products after the game and that there would be no consequence for the final score they earned. Instead, they were to act as proofreaders, reading the directions thoroughly and evaluating the clarity of them. As was the case in Study 2, half of the participants encountered a game that made the horse the most valuable animal, whereas the other half were led to believe the seal was the most valuable animal. Thus, this group of participants, aware of the point structure and the progression of animals, would also be susceptible to the gambler’s fallacy but would have little reason to be motivated to see the most valuable animal in the final round. In this control study, interpretations of the figure were not biased by what animals were most valuable. Those for whom farm animals would have been the most valuable were not more likely to see a horse than were those for whom sea animals would have been the most valuable, ␹2(1, N ⫽ 40) ⫽ 0.11, p ⫽ .74. When farm animals were the most valuable, 65% (n ⫽ 13) of participants saw the figure as a horse, and 35% (n ⫽ 7) saw it as a seal. When sea creatures were the most valuable, 70% (n ⫽ 14) saw the figure as a horse, and 30% (n ⫽ 6) saw it as a seal. The results of this study can be compared with those of Study 2 to suggest that reducing desire to see a particular animal can reduce the bias in interpretations. Because we are making comparisons across studies, it is necessary to use a Stouffer’s Z test (see Darlington & Hayes, 2000, for a review) to test if the effect of desire in Study 2 is sufficiently different from the effect of desire in this control study. That turns out to be the case (Z ⫽ 2.58, p ⬍ .005).

Study 3: Adding an Unobtrusive Measure Study 3 was designed to provide convergent evidence that the interpretations participants reported were, indeed, the sole interpretations that came to consciousness as they viewed the ambiguous stimulus. One can propose, instead, that participants saw both interpretations and then simply chose the one to tell the experimenter that placed them in a happier circumstance. One way to test whether participants saw only one versus both interpretations is to collect more unobtrusive measures that participants would not suspect were designed to test which interpretation they had seen—if they knew the measure was being taken at all. As was the case in the previous studies, we asked participants to provide a verbal or written report of whether they had seen a horse or a seal after being shown a figure that could be interpreted as either. However, in addition, we also measured participants’ eye movements to see if they would give clues as to how participants had interpreted the figure. Recent evidence suggests that initial eye movements on presentation of a stimulus are not influenced by conscious processing (Allopenna, Magnuson, & Tanenhaus, 1998; Richardson & Spivey, 2000; Tanenhaus, Spivey-Knowlton, Eber-

617

hard, & Sedivy, 1995). Thus, we examined whether the first saccade (eye movement) after presentation of the ambiguous figure would be to a label on the computer screen marked “farm animal” or one marked “sea creature.” We expected that such saccades would indicate that participants had interpreted the figure in a way that placed them in a favorable circumstance.

Method Participants. Participants were 79 undergraduates at Cornell University completing the study in exchange for extra credit. Procedure. Participants came into the lab alone and were seated approximately 20 in. from a 21-in. Apple cinema-display monitor (17 in. viewable). As was the case in previous studies, participants completed a task ostensibly about differences in predictions of and actual taste experiences of Aquafina, orange juice, and veggie smoothie. After participants predicted taste sensations of each item, the experimenter explained that to eliminate bias from the selection process, a computer program would randomly assign the item they would consume on the basis of their score at the end of a game similar to the one used in Study 2. As described in the previous study, the computer displayed pictures of farm and sea animals counterbalanced between participants to be worth either positive or negative points. Participants kept a record of the animal shown to them, the points that the animal was worth, and their ongoing score, ostensibly to corroborate with the computer program. Participants were told that although the computer would be keeping an ongoing tally of the points accumulated, they would still categorize the animal as either a farm animal or sea creature by clicking on a box on the computer screen to advance the computer to the next animal. The program displayed each animal for 1,000 ms, followed by a 500-ms blank screen, and finally a request to categorize the figure, which remained on the screen until participants responded. On the extreme left side of the categorization screen was a box labeled “farm animal,” and on the extreme right was a box labeled “sea creature.” Participants were instructed to categorize the animals on the computer correctly to avoid point penalties. In addition to losing points for incorrect categorization, participants learned that a portion of their final score would be determined by the speed of their categorization; thus, they were advised to categorize animals as quickly as possible. Unbeknownst to the participants, a video camera was hidden approximately 15 in. behind the monitor and trained on participants’ eyes. Thus, every time the categorization task appeared on the cinema-display monitor, we were able to capture participants’ initial eye movements. As practice to familiarize them with the task of viewing and categorizing animals, participants categorized filler animals eight times. After this practice session, participants completed 15 trials, the last of which displayed the ambiguous figure. Thus, participants were well-acquainted with the three-step process to complete a single trial: (a) view the animal, (b) categorize the animal on the computer screen, and (c) record the animal and points on the written response sheet. We were interested in the way in which participants interpreted the ambiguous figure. Their interpretation was measured in two ways: the written self-report and participants’ eye movements immediately on perceiving the categorization screen. Given that initial eye movements are not influenced by conscious processing (Richardson & Spivey, 2000), we can suppose that immediate looks at either the farm animal or sea creature box are representative of participants’ interpretations of the figure without concern for conscious, calculated response selection. We expected then that desire to see a particular animal would influence the way that the ambiguous figure was reported on the response sheet. Specifically, we expected that participants, hoping to drink orange juice, would see the most valuable animal. In addition, we expected that participants’ eye movements would corroborate their self-reports such that initial saccades would be toward the box labeled as the most desired animal.

618

BALCETIS AND DUNNING

Coder reliability. A coder, blind to condition, hypotheses, and purpose of the study watched the videotaped eye movements and noted the initial direction of movement for half of the data set. For the other half of the data set, a second coder, blind to condition, coded the videotaped eye movements. A third coder, blind to condition, randomly selected 18 participants from the complete data set and noted the initial direction of eye movement. Eye movements recorded by this third coder then served as a measure of interrater reliability. Across 213 individual trials from the 18 randomly selected participants, the third coder and the original coder agreed in 92% of the cases. If there was disagreement, the direction of eye movement as indicated from the original coder was used in analyses. In addition, to assess the validity of our nonconscious measure of initial eye movement and to see whether eye movements corresponded to what participants later reported, we randomly selected 48 participants and coded their eye movements in response to the 10 unambiguous animals that preceded the ambiguous figure. Across 480 trials, initial eye movements went to the correct categorization box 86% of the time.

Results Explicit reports. Using the same coding scheme as in the previous studies that used the horse–seal ambiguous figure, we again ran an ordinal logistic regression. As expected, desire facilitated the disambiguation of the figure, ␹2(1, N ⫽ 79) ⫽ 5.62, p ⬍ .02. When hoping to see farm animals, 83.7% (n ⫽ 36) of participants saw the figure as a horse, and 16.3% (n ⫽ 7) saw a seal. However, the pattern changed when participants hoped to see sea creatures. That is, 58.3% (n ⫽ 21) of this group saw a horse, 33.3% (n ⫽ 12) reported a seal, and 8.3% (n ⫽ 3) of participants did not indicate their interpretation. When looking only at the interpretations of those who offered one, it appears that desire influenced the disambiguation of the figure. Those who were motivated to see farm animals were more likely to report seeing a horse than were those who were motivated to see sea animals, ␹2(1, N ⫽ 76) ⫽ 4.02, p ⬍ .05. Eye movements. We used the same coding scheme in analyzing the interpretations gathered from participants’ eye movements. Again, those whose initial look was to the farm animal box received a score of 1, those who initially looked to the sea creature box received a score of ⫺1, and those who looked down to their response sheet and not to either the farm animal or sea creature box received a score of 0. We conducted an ordinal logistic regression and found that desire facilitated the disambiguation of the figure, ␹2(1, N ⫽ 79) ⫽ 10.24, p ⬍ .001. When hoping to see farm animals, 62.8% (n ⫽ 27) of participants looked to the farm animal box, 14.0% (n ⫽ 6) looked to the sea creature box, and 23.3% (n ⫽ 10) looked down to their score sheet. However, the pattern changed when participants hoped to see sea animals. That is, 30.6% (n ⫽ 11) looked to the farm animal box, 41.7% (n ⫽ 15) looked to the sea creature box, and 27.8% (n ⫽ 10) looked down to their score sheet. When looking only at the interpretations of those who looked to either box, it appears that desire influenced the disambiguation of the figure. Those who were motivated to see farm animals were more likely to look to the farm animal box than were those who were motivated to see sea animals, ␹2(1, N ⫽ 59) ⫽ 9.90, p ⫽ .002. We should note that scores on our eye-tracking measure significantly correlated with the score participants received from their explicit reports (Spearman’s ␳ ⫽ .42, p ⬍ .001).

Study 4: Converging Evidence from Lexical Decision Data Study 4 served as a conceptual replication of Study 3 but used a different type of indirect measure of perception. A good deal of research (e.g., Neely, 1991) suggests that a picture of an object serves as a prime for concepts associated with that object, even if people are not aware that they have seen the object (e.g., Loach & Mari-Beffa, 2003; Raymond, Shapiro, & Arnell, 1992). Thus, in Study 3, we motivated participants to interpret an ambiguous figure as either a horse or a seal. Participants again provided an explicit report of the interpretation they saw. However, we also collected reaction time data to gain an additional measure of whether participants had specifically seen the interpretation they had reported—and only that interpretation. Just after viewing the figure, participants completed a lexical decision task (LDT) in which they were presented with letter strings and had to decide whether those letter strings formed English words. Each participant saw a word related to the concept of “horse” (e.g., cowboy) or “seal” (e.g., blubber). We predicted that participants would respond more quickly to a word in the LDT exercise when that word was related to the interpretation they preferred to see rather than to the opposite interpretation. If participants actually saw both interpretations, no such difference should be seen in participants’ decision speed to words related to desired versus undesired interpretations. We also wanted to make sure that participants’ interpretations of the ambiguous figure were indeed responsible for priming their reactions in the LDT, rather than an overall desire to see a farm animal or sea creature. Thus, as a control condition, roughly half of the participants responded to the LDT just before they saw the ambiguous figure rather than just afterward. If participants responded more quickly to desired-concept words to a greater degree after they viewed the ambiguous figure, that fact would suggest that the interpretation participants saw was the one influencing the speed of their lexical decisions. However, if just a desire to see one type of animal over the other is enough to prime performance in the LDT, then desired-concept words should be facilitated in both before and after conditions to an equal degree. This design also allowed us to investigate one mechanism by which participants’ perceptions were influenced. Collecting LDT reaction times just before participants viewed the ambiguous figure allowed us to gauge whether people’s preferences suggested a perceptual set (Bruner & Minturn, 1955), that is, a preparedness to see the ambiguous figure as the desired object rather than the alternative. If participants provided quicker reaction times to words associated with the desired object than they did to words associated with the undesirable object, that pattern would be suggestive of a perceptual set.

Method Participants. Participants were 166 undergraduates at Cornell University who received extra credit in their psychology courses for taking part. Procedure. Participants came into the lab in groups of 2 to 4 to complete a task ostensibly about differences in internal and external evaluations of vocal abilities. The experimenter explained that approximately 75% of participants would evaluate various aspects of a person’s vocal performance, whereas the remaining 25% would be asked to perform a tune as if in a karaoke bar. The experimenter clarified that these

MOTIVATIONAL INFLUENCES ON VISUAL PERCEPTION percentages meant that approximately 1 person in each session would be the singer and subject of evaluation, whereas the remaining people would be observers. After performing a tune, singers would evaluate their own vocal abilities on rhythmic ability, skill, and general appeal. The experimenter explained that these scores would be corroborated against those provided by the observers on the same dimensions. At this point, participants were shown a 60-s video clip ostensibly of past participants and observers completing the performance evaluation portion of the experiment to heighten anxiety about the potential assignment to the singer role. In this video, a stocky Italian man in his early 20s held a microphone while singing and dancing along to Gloria Gaynor’s 1979 rendition of “I Will Survive.” Participants were seated approximately 24 –26 in. from a 17-in. iMac G4 or a 17-in. eMac desktop computer. As was the case in previous experiments, the experimenter explained that to eliminate bias from the selection process, a computer program would randomly assign participants to either the role of singer or observer. Participants played the same animal game as described in Study 3, ostensibly to determine whether they danced or observed. Again, participants kept a record of the animal shown to them, the points that the animal was worth, and their ongoing score, ostensibly to corroborate with the computer program. Additionally, participants categorized the animal as either a farm animal or sea creature on the computer. Finally, participants completed a number of LDTs during the animal categorization task, supposedly meant to impair their ability to categorize the animals. That is, participants categorized strings of letters as words or nonwords. In a go/no-go paradigm, participants hit the space bar if the string of letters was a word and did nothing if the string of letters was not a word. All strings of letters disappeared from the screen if no key was hit within 2,000 ms. Participants randomly assigned to the control condition completed the LDT at the beginning of each trial, that is, before seeing each animal. Participants randomly assigned to the experimental condition completed the LDT at the end of each trial, after seeing each animal but before categorizing it on the computer or recording it on their response sheet. Participants completed between one and three lexical decisions during each trial for the first 12 trials. In the last round, participants responded to three strings of letters. In this last trial, all participants responded to one word related to farm animals, one related to sea animals, and one nonword, the order of which were counterbalanced between subjects. Although for each participant only a single farm- and sea-relevant word was included in the last trial, the particular word selected was counterbalanced between subjects. Specifically, there were four words related to farm animals (cowboy, saddle, stallion, pasture), four words related to sea animals (blubber, flipper, ocean, whale), and four nonwords (blevre, yaver, dreas, puli) that were varied between subjects. That is, a participant would react to a single word from each of these sets. Again, the ongoing score at the end of the penultimate round were such that only one animal was worth enough positive points to produce an assignment to the observer role. For roughly half of the participants, the only animal capable of this was a horse, whereas for the other half, it was a seal. The last animal displayed was again the horse–seal ambiguous figure. We presumed that participants, going into the final trial with a negative score, would be hoping to see the animal worth the greatest number of positive points. We expected then that desire to see a particular animal would influence the way that the ambiguous figure was interpreted. Additionally, we expected that the desire to see a particular set of animals would influence the speed at which the target words was categorized, but only after participants had viewed the ambiguous figure. In particular, we expected that the control group that completed the LDTs before seeing the ambiguous figure would be equally likely to categorize the horse-relevant fragments and seal-relevant fragments as words. However, we expected that the experimental condition that completed the LDTs after having seen the ambiguous figure and interpreted it as the desired animal would be

619

faster to categorize words related to the desired animal type. Specifically, those participants in the experimental condition for whom farm animals were worth positive points were expected to categorize the farm-relevant words faster than sea-relevant words.

Results Although a small number of participants questioned why they had to play a computer game to determine their role, no participant was able to explain the purpose of the study. Additionally, in debriefing, some indicated disbelief that the performance evaluation component of the experiment would take place. Again, these people were unable to explain the purpose of the study. Thus, no participant was excluded for either of these reasons. Explicit reports. Omitting the one participant who did not offer an interpretation, we calculated the proportion of participants who had reported seeing a horse in each cell in a 2 (desired animal type: farm or sea) ⫻ 2 (task order: LDT before or after figure) design. Performing arcsin transforms on these proportions by means of the procedure outlined by Langer and Abelson (1972), allowed us to assess all main effects and interactions inherent in the design. This analysis indicated that desire facilitated the disambiguation of the figure. Whether or not participants saw a horse or a seal depended on whether participants were motivated to see farm animals or sea animals (z ⫽ 4.15, p ⬍ .001). No other effects were significant. When hoping to see farm animals, 97.2% (n ⫽ 69) of participants saw the figure as a horse, and 2.8% (n ⫽ 2) saw a seal. However, the pattern changed when participants hoped to see sea creatures. That is, 76.0% (n ⫽ 73) of this group saw a horse, 22.9% (n ⫽ 22) reported a seal, and 1.0% (n ⫽ 1) of participants did not indicate their interpretation. LDT. However, we were most interested in the speed with which strings of letters were categorized as words. The complete design was a 2 (word type: related to farm or sea animals) ⫻ 2 (desired animal: farm or sea) ⫻ 2 (task order: LDT before or after the ambiguous figure) with the first variable being within-subjects. Two participants (1 in the farm animal control condition, 1 in the sea animal experimental condition) made errors in categorizing during the LDT; their data are omitted. Given the skewed nature of the reaction time data, we conducted all analyses on natural log transformations. However, note all means reported in the text and tables are the original reaction times. In general, participants were no faster at responding to farm or sea words, F(1, 159) ⫽ 2.14, p ⫽ .15. Likewise, participants were no faster at responding to words when motivated to see either farm or sea animals, F(1, 159) ⬍ 1, p ⫽ .54. However, unexpectedly, it appears that those who completed the LDT before seeing the figure were generally faster (M ⫽ 778 ms, SE ⫽ 15) to respond than those completing the LDT after seeing the figure (M ⫽ 890 ms, SE ⫽ 27), F(1, 159) ⫽ 13.97, p ⬍ .001. Presumably, after viewing the ambiguous figure, participants were slowed somewhat, knowing that they would soon have to report the category of the creature they had seen. More interesting, the 2-way interaction between word type and desired animal was significant, F(1, 161) ⫽ 4.00, p ⫽ .05; but this interaction was qualified by the predicted 3-way interaction between word type, desired animal, and LDT time, F(1, 159) ⫽ 5.99, p ⫽ .02. As seen in Table 1, when completing the LDT before seeing the figure, the motivation to see a particular type of animal influenced the speed at which participants reacted to the words, as

BALCETIS AND DUNNING

620

Table 1 Reaction Times (ms) to Identify Word as a Function of Timing of Lexical Decision Task (LDT) and Desired Interpretation of the Ambiguous Figure Target word Farm related Timing of LDT task and desired interpretation Before Horse Seal After Horse Seal

Sea related

M

SE

M

SE

Difference

746 815

43 29

819 730

53 36

⫺73 85

716 958

37 46

1034 853

42 52

⫺318 105

confirmed by a significant Desired Animal ⫻ Word Type interaction that focused only on participants in the before condition, F(1, 91) ⫽ 5.49, p ⫽ .02. Participants responded to words associated with the desired category more quickly than they did to words associated with the undesired category. However, this advantage for words associated with desired categories was significantly stronger for participants completing the LDT after viewing the ambiguous figure, as evidenced by a significant Desired Animal ⫻ Word Type interaction, F(1, 68) ⫽ 25.05, p ⬍ .001. That is, those motivated to see farm animals responded faster to farm-related words than sea-related words by some 318 ms. Those motivated to see sea animals were faster to respond to sea-related words than farm-related words by some 105 ms.2 Unlike Study 3, for participants in the group who viewed the ambiguous figure before completing the LDT, scores on this implicit measure (reaction time to farm words minus reaction time to sea words) did not correlate with their explicit reports (point-biserial r ⫽ .05). Summary. In sum, Study 4 provided more convergent evidence that participants were more likely to interpret an ambiguous figure in line with their preferences. Participants again were more likely to explicitly report seeing a horse or a seal when they preferred to see that animal relative to when they did not. Their performance on an LDT also indicated that they had interpreted the ambiguous figure in a manner consistent with their desires. After seeing the ambiguous figure, participants recognized those words associated with the desired animal more quickly than they did words associated with an undesired animal, indicating that they had seen only the interpretation consistent with their desires. This performance advantage for words associated with desired animals was not as evident when participants completed an LDT before they viewed the ambiguous figure. However, participants who completed the LDT before they viewed the ambiguous figure still classified words associated with the desired interpretation more quickly than they did words associated with the opposite, although this tendency was much more muted relative to participants completing the lexical task after viewing the figure. This last result suggests a hint of a perceptual set: Participants showed some preparation or bias to see the desired interpretation over the undesired one before viewing the stimulus. However, this result is preliminary and tentative, and there is much more to explore regarding the processes that lead people to see what they want to see.

Study 5: Ruling Out Participant Deception This study was also designed to reduce suspicions about participants’ possible construction of responses to ensure favorable outcomes. If participants saw both interpretations and selectively reported the favorable interpretation, then both percepts in previous studies (e.g., horse vs. seal) would have to be accessible to them, in that participants would have to have seen both interpretations and selected only one when asked for an interpretation. To test this possibility, we again told participants that they were here to predict and describe taste sensations of freshly squeezed orange juice and organic veggie smoothie. They were shown an ambiguous figure, but before they could report what they had seen, the experimenter reported that he or she had made a mistake: that the participant would be assigned to the orange juice condition if the computer had shown him or her the other category of animal. Of key importance was what interpretation participants would report: the one they desired at the time they viewed the ambiguous figure or the one desired at the time they had to report what they saw. If participants saw only one interpretation in consciousness as they viewed the figure and if that interpretation was influenced by their motivational state, they should be more likely to report the figure they desired at the time the figure was presented to them. However, if they saw both interpretations and just reported the one that was desired when the experimenter asked for their report, then they should more likely report the figure that ran counter to their desires at the time they viewed the figure.

2 The effects and conclusions reported in the text remained virtually the same if we controlled for the specific words participants reacted to in the LDT. We should also note that all participants in a particular session, when multiple participants ran, were assigned to the same condition. Thus, although participants were assigned randomly, they were not assigned independently. This led cell sizes to differ somewhat. We should note that we ran supplemental analyses to gauge whether any of our results were due to session effects. When we controlled for the particular session in which participants ran (by conducting analyses in which session was added as a random variable nested within our conditions), we found our findings remained intact.

MOTIVATIONAL INFLUENCES ON VISUAL PERCEPTION

Method Participants. Participants were 27 undergraduates at Cornell University who received extra credit in their psychology courses for taking part. Procedure. The procedures for this experiment were modeled closely on those used in Study 1. Again, the computer would assign the participant to drink freshly squeezed orange juice or an off-putting veggie smoothie on the basis of the single item that it randomly selected from a database. In this study, though, for half of the participants, if the computer displayed a farm animal, participants would consume the orange juice, whereas a sea creature would bring the veggie smoothie. For the other half of the participants, this was reversed. After these instructions were explained to participants, the experimenter supposedly calibrated the computer program. In a practice phase, the program displayed four animals as examples of what would be shown. Two of these examples were farm animals, and two were sea creatures. Crossed with this, two of the animals were drawings of the full bodies of animals, whereas two were just of animal heads. Following the examples, participants fixated on a red dot flashing in the center of a 15-in. G3 iBook screen for 3 s. This fixation point was then replaced by the horse–seal ambiguous figure (3.75 in. high, 2.75 in. wide) displayed for 1,000 ms followed by the same staged computer program crash. The experimenter remained preoccupied with paperwork until the participant got her attention. Unlike the previous study, the experimenter did not ask the participant at this point if anything was displayed before the crash. Instead, she immediately offered that the crash was most likely because she made an error during the calibration. For those participants for whom farm animals were valued, she continued by saying the error was that in fact sea creatures were supposed to signal the consumption of orange juice. For those valuing sea creatures, she said the error was that farm animals were in fact supposed to signal the consumption of orange juice. To rephrase, after the crash, the experimenter switched which animals were desired. After explaining this confusion and making the switch, the experimenter asked if anything was shown before the crash.

Results The procedure of Study 5 put two accounts for our data in opposition. Our guiding hypothesis is that participants’ motivational states influence the interpretation of the ambiguous figure that is presented to consciousness at the time the figure is viewed. If motivational states help to disambiguate the figure during the time it is viewed, we would expect that after the switch participants would tend to report seeing the animal from the desired category at the time of viewing the object, even though this animal, after the switch in instructions, ultimately consigned them to drink the veggie smoothie. However, if participants see both interpretations and then just report the one that they favor, then we would expect that participants would be more likely to report seeing an animal from the category that is desirable after the switch. We used the same type of coding scheme for interpretations as in the previous studies. Using an ordinal logistic regression, we found that participants were more likely to report the animal that was originally the most desired even when this meant they would complete the less desirable task, ␹2(1, N ⫽ 27) ⫽ 9.48, p ⫽ .002. When participants originally hoped to see farm animals, 100% (n ⫽ 13) reported seeing a horse even when the horse ultimately meant drinking the veggie smoothie. When participants originally hoped to see sea creatures, 28.6% (n ⫽ 4) reported seeing a seal, 57.1% (n ⫽ 8) saw a horse, and 14.3% (n ⫽ 2) said nothing was shown before the crash. Focusing on only those participants reporting an interpretation, we again found that participants were

621

more likely to report a horse or a seal when they were originally motivated to see that type of animal (Fisher’s exact p ⫽ .039). Although a larger percentage of participants reported seeing a horse than they did a seal when originally hoping to see sea creatures, what is important is that the percentage who saw a seal is biased between conditions on the basis of original desire. That is, when originally hoping to see a horse, none saw a seal, but when originally hoping to see a seal, nearly 30% saw one.

General Discussion The world people know is the one they take in through their senses. In these studies, we examined the extent to which what people take in could be guided by such top-down constraints as personal wishes and preferences. Across these studies, we provided converging evidence to suggest that participants’ desires, hopes, fears, or wishful thinking led them to perceive a representation of the visual environment they desired. Studies 1 and 2 demonstrated that participants tended to interpret an ambiguous figure in a manner that fit with their wishes and preferences over one that did not. Studies 3 and 4 added implicit measures to ensure that participants actually saw the interpretation they favored and not just what they chose to report seeing. Specifically, for a clear majority of participants in Study 3, their first saccade after presentation of an ambiguous stimulus tended to be to the favored category label rather than to the disfavored one. In Study 4, after viewing an ambiguous figure, participants reacted in an LDT to words consistent with a preferred interpretation more quickly than to words consistent with the less preferred one. It is important that this facilitation after seeing the ambiguous stimulus was greater than it was for those performing the LDT before viewing the stimulus, indicating that the ambiguous figure primed concepts associated with the preferred interpretation more than it did the less preferred one. Study 5 added a procedural variation to affirm that participants did not see both interpretations in our experiments and then just report the one that brought about the favored outcome. Participants viewed an ambiguous stimulus while hoping for one outcome, but then the experimenter switched which interpretation was the favored one before participants reported what they had seen. Participants tended to report seeing the interpretation they favored at the time they viewed the stimulus, even though that report, after the switch, assigned them to a less desired task. It is important that Study 5 demonstrated that wishful thinking constrains perceptual processes preconsciously, before the products of those processes become available to conscious awareness.

Alternative Accounts A critic might argue that the paradigms we used might have taken advantage of other psychological processes, rather than motivational states, that could influence participants’ interpretation of ambiguous stimuli. For example, participants’ interpretations of ambiguous figures could have been due to differences in expectation. In Studies 2, 3, and 4, participants were exposed to a series of stimuli they did not want to see just before they viewed the critical ambiguous stimulus. Participants’ might have fallen prey to the gambler’s fallacy and expected that a favored animal was bound to show up after a string of unwanted ones.

622

BALCETIS AND DUNNING

However, Study 1, the control study associated with Study 2, and Study 5 all argue against this explanation. For example, Studies 1 and 5 presented participants with a single stimulus and still found that people tended to see the interpretation they wanted to see over the one they did not. In addition, the control study associated with Study 2 specifically tested whether a gambler’s fallacy alone would influence what they saw in the ambiguous figure when participants had no motivation to interpret the ambiguous figure in a certain way. Further, it is implausible that our results are explained by cognitive or perceptual salience. That is, one could argue that the desired interpretation was highlighted and more easily seen by participants because that perceptual outcome was paired with a desirable event. However, in our experiments we were careful to pair both the favored and less favored interpretations with salient events. In Study 1, for example, seeing a number might be associated with drinking delicious orange juice, but seeing a letter was associated with an event— drinking a foul-smelling and foullooking concoction—that was at least as salient. Thus, salience is not a viable alternative explanation for the pattern of responses we observed.

Notes on the Mechanism Underlying Biased Perception Our results suggest that people’s desires for a particular outcome bias their perceptual set, such that they are more prepared to see what they hope for rather than what they fear. In fact, in a funneled probe for suspicion, one participant offered, “I kept getting ⫹5 and –5 over and over, making me worry about eating the beans. At the last minute, I was sure I would have to eat the heinous beans and I prayed for the horse to give me a ⫹5. I got it! Yes!” Of course, prayer may not always be the precise mechanism biasing all participants’ interpretations, but we do feel this illustrates a possible chain of events leading to differences in what participants saw. A desire to see one stimulus over the other led to the formation of a perceptual set that included features and concepts related to the desired stimulus over the undesired one. Indeed, in Study 4, we discovered initial evidence of a perceptual set biased toward the favored hypothesis. Just before viewing the ambiguous stimulus, participants were slightly— but significantly—faster at recognizing words associated with the favored interpretation than they were words associated with the disfavored interpretation. Two notes are in order concerning this finding and the potential role of perceptual set in motivational influences on perception. First, the specific perceptual hypothesis that participants might be using to disambiguate the stimulus need not be closely tied to the nature of the stimulus. To be sure, in Studies 2, 3, and 4, participants were given a rather narrow hypothesis (i.e., the desirable stimulus will be either a horse or a seal) about what the computer might show them. In this way, our work is reminiscent of previous work concerning contextual effects on visual perception of ambiguous figures, in which participants are given primes whose appearance is quite close to that of the ambiguous stimulus (i.e., they are shown drawings of women) before they view that stimulus (e.g., one seen as either a man or woman; Long & Toppino, 2004). However, in Studies 1 and 5, participants were not given such specific hints about what the stimulus might be. Instead, they were given broad categories (e.g., a letter versus a number; a farm

versus a sea animal). As a result, they were not necessarily able to look for features of a specific stimulus but rather had to search for any number of possible stimuli to satisfy these broad categories. Even in this circumstance, participants tended to see what they wanted to see. This suggests that the top-down influences on perception inspired by motivation can be quite diffuse and nonspecific: that when disambiguating an ambiguous figure, people do not need concrete features specified a priori. Instead, the clues or context surrounding the perceptual judgment can be quite vague, indirect, abstract, or higher order. This conjecture is consistent with other recent evidence showing that priming people with abstract categories (such as “flirting” or “music”) has an impact on how they interpret ambiguous figures they subsequently view (see Balcetis & Dale, 2003). Second, these studies left open one ambiguity about perceptual set that future work could profitably address. Across five studies, we found that people tended to see an interpretation they favored over one they did not. But did this bias arise because the perceptual set associated with their motivational state was an approach one, facilitating processes associated with seeing the favored interpretation, or an avoidance one, inhibiting processes that could lead them to see the disfavored interpretation? Either route—facilitation of the favored interpretation, inhibition of the disfavored one, or a mixture of the two— could lead to the pattern of responses we observed. Future work could potentially tease apart whether the phenomenon we uncovered is one in which people are biased toward seeing wanted stimuli or biased against seeing stimuli they wish to avoid, or both.

Where Does the Bias Reside in the Perceptual System? One remaining question that this work leaves open is determining the stage in the perceptual process at which motivational factors begin to guide perception. Such a question is relevant not only to work on motivation but also to work on other higher order constructs (e.g., stereotypes, expectations, frames) that have been at the focus of social cognitive work. Is the impact of motivation limited to later stages of perception, such as categorization, or does its influence extend to earlier and more primitive tasks the perceptual system faces (e.g., noticing lines and edges in a visual scene)? This question became a major theoretical battle during the New Look period, one that continues to this day. In particular, Bruner and Goodman’s (1947) theory of perceptual defense was criticized by opponents, who asked how a perceiver could selectively defend against a particular stimulus unless the stimulus is already perceived (Eriksen & Browne, 1956; Howie, 1952; Spence, 1967). Critics of Bruner and Goodman (1947) and more recent ones have argued that higher order constraints influence not early perception but rather later stages of the perceptual processes that could be termed postperceptual or perceptual decision making. Pylyshyn (1999), for example, asserted that the act of perceiving an object contains at least two processes. One process, termed early vision works, which is immune to higher order influences, works to provide three-dimensional representations of the surfaces of objects. A later process takes any created representation and then identifies or categorizes it. Pylyshyn (1999) argued that higher

MOTIVATIONAL INFLUENCES ON VISUAL PERCEPTION

order influences have an impact predominantly on this latter stage.3 However, this assertion is a contentious issue (see the commentaries that accompany Pylyshyn, 1999), and more recent evidence suggests that higher order processes can impose their influence on perception very early in the perceptual process. Emerging evidence, for example, suggests that higher order influences can be detected in V1, the area of the primary visual cortex considered to be the simplest, earliest cortical visual area responsible for processing visual stimuli, which is a mere two synapses away from the eye (Boynton, 2005). For example, when perceivers are asked attend to one of two overlapping orthogonal line patterns, functional magnetic resonance imaging activity patterns in early visual areas, including V1, contain information that can predict what the participant consciously perceives (Kamitani & Tong, 2005). Perceptions of patterns in V1 also occur even if participants are clearly unaware that a pattern has been shown to them (Haynes & Rees, 2005).

Implications for Self-Deception The data from these five studies also have implications for another enduring issue in psychology. Over the decades, social, personality, clinical, and cognitive psychologists have catalogued a myriad of ways in which people engage in wishful thinking (for reviews, see Baumeister & Newman, 1994; Dunning, 2001; Kunda, 1990; Mele, 1997; Pittman, 1998). However, people remain seemingly unaware that they do all this cognitive work; they remain innocent of the fact that their fears and desires have shaped how they view themselves and think about the world around them (Ehrlinger, Gilovich, & Ross, 2005; Gilbert, Pinel, Wilson, Blumberg, & Wheatley, 1998; Mele, 1997; Pronin, Gilovich, & Ross, 2004). Indeed, for people to reach their motivational goals, it is imperative that they remain unaware of the distortions they place on their thinking. If they knew that they believed some pleasant thought merely because they wanted to believe it, they would also know, at least in part, how illegitimate that thought was. How, then, do people pull off the self-deception crucial to the execution of motivated reasoning? Our data provide one answer to this riddle. People fail to recognize such self-serving biases if those processes remain outside of conscious awareness, monitoring, or control. If those processes take place preconsciously, before any content of perception and cognition reaches consciousness, people can construct pleasant thoughts yet remain unaware of the construction. The only content that would be available in consciousness would be the product and not the process of motivated reasoning. There exist some shards of evidence that motivational processes operate on a nonconscious level (e.g., Arndt, Greenberg, Pyszczynski, & Solomon, 1997; Fein & Spencer, 1997). The present studies enlarge the types of nonconscious processes that motivational states may influence, and it may be profitable to consider other automatic or nonconscious processes that might be molded, in part, by the motivation toward believing in a masterful self in a congenial world. One also wonders about the full range of nonconscious processes that might be tainted by motivational pressures. The world people know is the one they take in through their senses, but it is also formed by other preconscious processes. To what extent is the

623

representation of the world furnished to conscious awareness by all of these processes one that reproduces the outside world faithfully versus one that people just wish they could inhabit? There is much work to be done to address this question, and we are unsure at the end what picture of the perceiver we will see.

3 Pylyshyn (1999) also allowed for the possibility that higher order processes might guide attentional mechanisms that guide early vision.

References Aarts, H., & Dijksterhuis, A. (2002). Category activation effects in judgment and behaviour: The moderating role of perceived comparability. British Journal of Social Psychology, 41, 123–138. Adkins, L. J. (1956). Critical comment on the measurement of familiarity in personality-perception experiments. Perceptual and Motor Skills, 6, 147–151. Allopenna, P. D., Magnuson, J. S., & Tanenhaus, M. K. (1998). Tracking the time course of spoken word recognition using eye movements: Evidence for continuous mapping models. Journal of Memory and Language, 38, 419 – 439. Allport, A. (1989). Visual attention. In M. I. Posner (Ed.), Visual attention (pp. 631– 682). Cambridge, MA: The MIT Press. Arndt, J., Greenberg, J., Pyszczynski, T., & Solomon, S. (1997). Subliminal exposure to death-related stimuli increases defense of the cultural worldview. Psychological Science, 8, 379 –385. Baird, J. C., & Biersdorf, W. R. (1967). Quantitative function for size and distance judgments. Perception & Psychophysics, 2, 161–166. Balcetis, E., & Dale, R. (2003). There is no naked eye: Higher-order social concepts clothe visual perception. Proceedings of the twenty-fifth annual meeting of the Cognitive Science Society (pp. 109 –114). Mahwah, NJ: Erlbaum. Baumeister, R. F., & Newman, L. S. (1994). Self-regulation of cognitive inference and decision processes. Personality and Social Psychology Bulletin, 20, 3–19. Bhalla, M., & Proffitt, D. R. (1999). Visual–motor recalibration in geographical slant perception. Journal of Experimental Psychology: Human Perception and Performance, 25, 1076 –1096. Biederman, I., Mezzanotte, R. J., & Rabinowitz, J. C. (1982). Scene perception: Detecting and judging objects undergoing relational violations. Cognitive Psychology, 14, 143–177. Boyce, S. J., & Pollatsek, A. (1992). Identification of objects in scenes: The role of scene background in object naming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 531–543. Boynton, G. M. (2005). Imagining orientation selectivity: Decoding conscious perception in V1. Nature Neuroscience, 8, 541–542. Bruner, J. S., & Goodman, C. C. (1947). Value and need as organizing factors in perception. Journal of Abnormal Social Psychology, 42, 33– 44. Bruner, J. S., & Minturn, A. L. (1955). Perceptual identification and perceptual organization. Journal of General Psychology, 53, 21–28. Bugelski, B. R., & Alampay, D. A. (1961). The role of frequency in developing perceptual sets. Canadian Journal of Psychology, 15, 205– 211. Changizi, M. A., & Hall, W. G. (2001). Thirst modulates a perception. Perception, 30, 1489 –1497. Creem, S. H., & Proffitt, D. R. (1998). Two memories for geographical slant: Separation and interdependence of action and awareness. Psychonomic Bulletin & Review, 5, 22–36. Darlington, R. B., & Hayes, A. F. (2000). Combining independent pvalues: Extensions of the Stouffer and binomial methods. Psychological Methods, 5, 496 –515.

624

BALCETIS AND DUNNING

Dawson, E., Gilovich, T., & Regan, D. T. (2002). Motivated reasoning and performance on the Wason selection task. Personality and Social Psychology Bulletin, 28, 1379 –1387. Ditto, P. H., & Lopez, D. F. (1992). Motivated skepticism: Use of differential decision criteria for preferred and nonpreferred conclusions. Journal of Personality and Social Psychology, 63, 568 –584. Dunning, D. (2001). On the motives underlying social cognition. In N. Schwarz & A. Tesser (Eds.), Blackwell handbook of social psychology: Vol. 1. Intraindividual processes (pp. 348 –374). New York: Blackwell. Durgin, F. H., Proffitt, D. R., Olson, T. J., & Reinke, K. S. (1995). Comparing depth from motion with depth from binocular disparity. Journal of Experimental Psychology: Human Perception and Performance, 21, 679 – 699. Ehrlinger, J., Gilovich, T., & Ross, L. (2005). Peering into the bias blind spot: People’s assessment of bias in themselves and others. Personality and Social Psychology Bulletin, 31, 680 – 692. Erdelyi, M. H. (1974). A new look at the new look: Perceptual defense and vigilance. Psychological Review, 81, 1–25. Erdelyi, M. H. (1985). Psychoanalysis: Freud’s cognitive psychology. New York: Freeman. Eriksen, C. W. (1958). Unconscious processes. In M. R. Jones (Ed.), Nebraska symposium on motivation (pp. 169 –227). Lincoln: University of Nebraska Press. Eriksen, C. W. (Ed.). (1962). Behavior and awareness: A symposium of research and interpretation. Durham, NC: Duke University Press. Eriksen, C. W., & Browne, C. T. (1956). An experimental and theoretical analysis of perceptual defense. Journal of Abnormal and Social Psychology, 52, 224 –230. Fein, S., & Spencer, S. J. (1997). Prejudice as self-image maintenance: Affirming the self through derogating others. Journal of Personality and Social Psychology, 73, 31– 44. Fisher, G. H. (1968). Ambiguity of form: Old and new. Perception & Psychophysics, 4, 189 –192. Gilbert, D. T. (1998). Ordinary personology. In D. T. Gilbert, S. Fiske, and G. Lindzey (Eds.), Handbook of social psychology (Vol. 2, pp. 89 –150). New York: McGraw-Hill. Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J., & Wheatley, T. P. (1998). Immune neglect: A source of durability bias in affective forecasting. Journal of Personality and Social Psychology, 75, 617– 638. Gilinsky, A. S. (1951). Perceived size and distance in visual space. Journal of Experimental Psychology, 44, 11–15. Goldiamond, I. (1958). Indicators of perception: I. Subliminal perception, subception, unconscious perception: An analysis in terms of psychophysical indicator methodology. Psychological Bulletin, 55, 373– 411. Greenwald, A. G. (1992). New look 3: Unconscious cognition reclaimed. American Psychologist, 47, 766 –779. Gregory, R. L. (1974). Choosing a paradigm for perception. In E. C. Carterete & M. P. Friedman (Eds.), Handbook of perception: Vol. 1. Historical and philosophical roots of perception (pp. 255–283). New York: Academic Press. Haynes, J.-D., & Rees, G. (2005). Predicting the orientation of invisible stimuli from activity in human primary visual cortex. Nature Neuroscience, 8, 686 – 691. Henderson, J. M., & Hollingworth, A. (1999). High-level scene perception. Annual Review of Psychology, 50, 243–271. Howes, D. H., & Solomon, R. L. (1950). A note on McGinnies’ “emotionality and perceptual defense.” Psychological Review, 57, 229 –234. Howie, D. (1952). Perceptual defense. Psychological Review, 59, 308 – 315. Johnston, L., Arden, K., Macrae, C. N., & Grace, R. C. (2003). The need for speed: The menstrual cycle and person construal. Social Cognition, 21, 89 –100. Jones, E. E. (1985). Major developments in social psychology during the past five decades. In G. Lindzey & E. Aronson (Eds.), The handbook of

social psychology (3rd ed., Vol. 1, pp. 47–108). New York: Random House. Kamitani, Y., & Tong, F. (2005). Decoding the visual and subjective contents of the human brain. Nature Neuroscience, 8, 679 – 685. Kosslyn, S. M., & Koenig, O. (1992). Wet mind: The new cognitive neuroscience. New York: Free Press. Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108, 480 – 498. Langer, E. J., & Abelson, R. P. (1972). The semantics of asking a favor: How to succeed in getting help without really dying. Journal of Personality and Social Psychology, 24, 26 –32. Leeper, R. (1935). A study of a neglected portion of the field of learning— The development of sensory organization. Journal of Genetic Psychology, 46, 41–75. Li, L., & Warren, H. (2004). Path perception during rotation: Influence of instructions, depth range, and dot density. Vision Research, 44, 1879 – 1889. Loach, D., & Mari-Beffa, P. (2003). Post-target inhibition: A temporal binding mechanism? Visual Cognition, 10, 513–526. Long, G. M., & Olszweski, A. D. (1999). To reverse or not to reverse: When is an ambiguous figure not ambiguous? American Journal of Psychology, 112, 41–71. Long, G. M., & Toppino, T. C. (2004). Enduring interest in perceptual ambiguity: Alternating views of reversible figures. Psychological Bulletin, 130, 748 –768. Macrae, C. N., Alnwick, K. A., Milne, A. B., & Schloerscheidt, A. M. (2002). Person perception across the menstrual cycle: Hormonal influences on social– cognitive functioning. Psychological Science, 13, 532– 536. McCurdy, H. G. (1956). Coin perception studies and the concept of schemata. Psychological Review, 63, 160 –168. Mele, A. R. (1997). Real self-deception. Behavioral and Brain Sciences, 20, 91–136. Meng, M., & Tong, F. (2004). Can attention selectively bias bistable perception? Differences between binocular rivalry and ambiguous figures. Journal of Vision, 4, 539. Michelon, P., & Koenig, O. (2002). On the relationship between visual imagery and visual perception: Evidence from priming studies. European Journal of Cognitive Psychology, 14, 161–184. Miller, J. (1987). Priming is not necessary for selective-attention failures: Semantic effects of unattended, unprimed letters. Perception and Psychophysics, 41, 419 – 434. Neely, J. H. (1991). Semantic priming effects in visual word recognition: A selective review of current findings and theories. In G. W. Humphreys & D. Besner (Eds.), Basic processes in reading: Visual word recognition (pp. 264 –336). Hillsdale, NJ: Erlbaum. Nisbett, R. E., & Ross, L. D. (1980). Human inference: Strategies and shortcomings of social judgment. Englewood Cliffs, NJ: Prentice Hall. Pelli, D., Palomares, M., & Majaj, N. J. (2004). Crowding is unlike ordinary masking: Distinguishing feature integration from detection. Journal of Vision, 4, 1136 –1169. Pittman, T. S. (1998). Motivation. In D. Gilbert, S. Fiske, & G. Lindsay (Eds.), Handbook of social psychology (4th ed., Vol. 1, pp. 549 –590). Boston: McGraw-Hill. Postman, L., Bruner, J. S., & McGinnies, E. (1948). Personal values as selective factors in perception. Journal of Abnormal and Social Psychology, 43, 142–154. Prentice, W. C. H. (1958). Perception. Annual Review of Psychology, 9, 1–18. Proffitt, D. R., Creem, S. H., & Zosh, W. (2001). Seeing mountains in mole hills: Geographical slant perception. Psychological Science, 12, 418 – 423. Proffitt, D. R., Stefanucci, J., Banton, T., & Epstein, W. (2003). The role of effort in perceiving distance. Psychological Science, 14, 106 –112.

MOTIVATIONAL INFLUENCES ON VISUAL PERCEPTION Pronin, E., & Gilovich, T., & Ross, L. (2004). Objectivity in the eye of the beholder: Divergent perceptions of bias in self versus others. Psychological Review, 111, 781–799. Pylyshyn, Z. W. (1999). Is vision continuous with cognition? The case for cognitive impenetrability of visual perception. Behavioral and Brain Sciences, 22, 341– 423. Pyszczynski, T., & Greenberg, J. (1987). Toward an integration of cognitive and motivational perspectives on social inference: A biased hypothesis-testing model. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 20, pp. 297–340). New York: Academic Press. Raymond, J. E., Shapiro, K. L., & Arnell, K. M. (1992). Temporary suppression of visual processing in an RSVP task: An attentional blink? Journal of Experimental Psychology: Human Perception and Performance, 18, 849 – 860. Richardson, D. C., & Spivey, M. J. (2000). Representation, space and Hollywood Squares: Looking at things that aren’t there anymore. Cognition, 76, 269 –295. Rock, I., & Mitchener, K. (1992). Further evidence of failure of reversal of ambiguous figures by uninformed subjects. Perception, 21, 39 – 45. Sanitioso, R., Kunda, Z., & Fong, G. T. (1990). Motivated recruitment of autobiographical memories. Journal of Personality and Social Psychology, 59, 229 –241. Simons, D. J., & Chabris, C. F. (1999). Gorillas in our midst: Sustained inattentional blindness for dynamic events. Perception, 28, 1059 –1074. Spence, D. P. (1967). Subliminal perception and perceptual defense: Two sides of a single problem. Behavioral Science, 12, 183–193. Tanenhaus, M. K., Spivey-Knowlton, M. J., Eberhard, K. M., & Sedivy, J. C. (1995, June 16). Integration of visual and linguistic information in spoken language comprehension. Science, 268, 1632–1634.

625

Tittle, J. S., Todd, J. T., Perotti, V. J., & Norman, J. F. (1995). Systematic distortion of perceived three-dimensional structure from motion and binocular stereopsis. Journal of Experimental Psychology: Human Perception and Performance, 21, 663– 678. Todd, J. T., & Bressan, P. (1990). The perception of 3-dimensional affine structure from minimal apparent motion sequences. Perception & Psychophysics, 48, 419 – 430. Todd, J. T., & Norman, J. F. (1991). The visual perception of smoothly curved surfaces from minimal apparent motion sequences. Perception & Psychophysics, 50, 509 –523. Toppino, T. (2003). Reversible-figure perception: Mechanisms of intentional control. Perception & Psychophysics, 65, 1285–1295. Trope, Y., & Ferguson, M. J. (2001). How and when preferences influence inferences: A motivated hypothesis-testing framework. In J. Bargh & D. K. Apsley (Eds.), Unraveling the complexities of social life: A Festschrift in honor of Robert B. Zajonc (pp. 111–130). Washington, DC: American Psychological Association Press. Wegner, D. M., & Bargh, J. A. (1998). Control and automaticity in social life. In D. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), Handbook of social psychology (4th ed., Vol. 1, pp. 446 – 496). New York: McGraw-Hill. Wohlwill, J. F. (1966). Perceptual learning. Annual Review of Psychology, 17, 201–232. Yang, T. L., Dixon, M. W., & Proffitt, D. R. (1999). Seeing big things: Overestimation of heights is greater for real objects than for objects in pictures. Perception, 28, 445– 467.

Received July 15, 2005 Revision received January 17, 2006 Accepted January 30, 2006 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 626 – 641

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.626

Jealousy and the Threatened Self: Getting to the Heart of the Green-Eyed Monster David DeSteno, Piercarlo Valdesolo, and Monica Y. Bartlett Northeastern University Several theories specifying the causes of jealousy have been put forth in the past few decades. Firm support for any proposed theory, however, has been limited by the difficulties inherent in inducing jealousy and examining any proposed mediating mechanisms in real time. In support of a theory of jealousy centering on threats to the self-system, 2 experiments are presented that address these past limitations and argue for a model based on context-induced variability in self-evaluation. Experiment 1 presents a method for evoking jealousy through the use of highly orchestrated social encounters and demonstrates that threatened self-esteem functions as a principal mediator of jealousy. In addition to replicating these findings, Experiment 2 provides direct evidence for jealousy as a cause of aggression. The ability of the proposed theory of jealousy to integrate other extant findings in the literature is also discussed. Keywords: jealousy, self-esteem, aggression, emotion

certain outcomes (Frijda, 2000; Keltner & Gross, 1999; Keltner & ¨ hman & Haidt, 1999; Lazarus, 1991; LeDoux & Phelps, 2000; O Wiens, 2003). The cognitive and physiological changes associated with fear and anxiety, for example, prepare an organism to detect and/or escape from an impending danger more efficiently (LeDoux ¨ hman, 2002). It is important to note, however, & Phelps, 2000; O that organisms whose existence is characterized by high degrees of collective or social living confront not only challenges involving the successful navigation of the physical environment but also those involving the social one (e.g., social exchange, coalition building, social bonding, and relationship maintenance; Bartlett & DeSteno, 2006; Cosmides & Tooby, 2000; Darwin, 1872/1998; Keltner & Busswell, 1997; Keltner & Haidt, 1999; Lewis, 2000). The importance of such challenges suggests the need for specific emotional responses that are intrinsically tied to sociality.

Jealousy, it seems, is a fundamental aspect of human social life. For as far back in time or as widely across civilizations as one can peer, the green-eyed monster has reared its head. From Gilgamesh’s romps retold in the first millennium B.C.E, to Othello’s throes portrayed in the middle part of the last millennium, to modern day soap operas and drama series, fascination with the jealousy motif has not waned among artists and audiences alike. From cultures representing geographically and socially disparate milieus, research documents the pervasiveness of jealousy among men and women from childhood to old age (e.g., Bryson, 1991; Buunk, Angleitner, Oubaid, & Buss, 1996; Geary, Rumsey, BowThomas, & Hoard, 1995; Hupka et al., 1985; Masciuch & Kienapple, 1993). Jealousy’s ubiquity is so well accepted that even Freud (1922/1955) himself suggested that its absence, not its presence (at least within normal levels), is a sign of pathology. From a functional perspective, jealousy stands as an exemplary candidate for a fundamental social emotion. Emotions, like many psychological phenomena, are theorized to exist because they serve some adaptive purpose. That is, although their specific components and sequelae may operate on many different levels (e.g., neurochemical, interpersonal, cultural), emotions are designed to increase the success with which an organism meets specific challenges by shunting cognition and behavior toward

Jealousy: Form and Function For humans, adaptive functioning is intrinsically tied to social interactions through which myriad needs are met (e.g., protection, resource acquisition, reproduction). Accordingly, engagement in interpersonal relationships stands as a fundamental predictor of human physical and psychological health (Baumeister & Leary, 1995; Berscheid & Reis, 1998; Cacioppo et al., 2002) and is fostered by the seemingly universal motive to belong to social groups and be a member of interpersonal relationships (Baumeister & Leary, 1995). Indeed, involvement in social relationships is of such central value to adaptive functioning that it has been documented to increase psychological well-being (Diener, 1984; Myers & Diener, 1995), resistance to cardiovascular disease (Berkman, Vaccarino, & Seeman, 1993), resistance to cancer (Glanz & Lerman, 1992), and immune system function (Booth & Pennebaker, 2000; Kennedy, Kiecolt-Glaser, & Glaser, 1990; Kiecolt-Glaser, 1999). Given the benefits provided by relationships, competition for them frequently arises (Salovey, 1991). Consequently, the exis-

David DeSteno, Piercarlo Valdesolo, and Monica Y. Bartlett, Department of Psychology, Northeastern University. Monica Bartlett is now at the Department of Psychology, Gonzaga University. This research was supported by National Institute of Mental Health Grant MH068240. We thank Nilanjana Dasgupta and members of the Boston Emotion Research Lab for insights and comments regarding this work. Correspondence concerning this article should be addressed to David DeSteno, Department of Psychology, Northeastern University, Boston, MA 02115. E-mail: [email protected] 626

JEALOUSY AND THE THREATENED SELF

tence of a specific emotion designed to protect these relationships from the advances of rivals is to be expected. Accordingly, most researchers agree that jealousy functions to evoke somatic, cognitive, and behavioral responses designed to address relationship threats (Buss, Larsen, Westen, & Semmelroth, 1992; DeSteno & Salovey, 1995; Salovey, 1991; White, 1991).1 With respect to phenomenology, most researchers also agree that the subjective experience of jealousy is quite aversive and best described as a combination or blend of the feelings of anger, anxiety, betrayal, and hurt (Buck, 1999; Hupka, 1984, 1991; Parrott & Smith, 1993; Sharpsteen, 1991; Sharpsteen & Kirkpatrick, 1997). The ubiquity and agony of jealousy stand in direct correspondence to the fundamental threats posed by its eliciting events. For many, the prototypical jealousy-evoking situation involves a romantic triad: An individual becomes jealous as he or she suspects or actually learns that a partner is interested in a rival (Salovey, 1991). Asking individuals about actual or imagined instances of this type of scenario has been one of the more widely used methods in studies of jealousy (e.g., Buss, Larsen, Westen, & Semmelroth, 1992; DeSteno, Bartlett, Braverman, & Salovey, 2002; DeSteno & Salovey, 1996; Harris, 2003; Salovey, 1991). Yet, there is no requirement that the relationship being threatened needs to be a romantic one. All that is central is that a valued relationship of any type may be usurped by a rival (DeSteno, 2004). Research clearly supports the fact that jealousy is not limited solely to romantic relationships but can occur within any type of triadic relationship. Developmental research, for example, has shown that children may be jealous of siblings’ relationships with parents (Masiuch & Kienapple, 1993; Volling, McElwain, & Miller, 2002); workers have been shown to be jealous of their coworkers’ relationships with superiors (Vecchio, 2000). In each case, the nature of the fundamental threat is the same, although the specifics differ. Parents possess a finite amount of personal (e.g., emotional, attentional) and substantive (e.g., economic, food) resources that can be divided among offspring; superiors, likewise, possess a finite amount of privileges they can offer. In both instances, the strength of one’s relationship with such partners holds important implications for survival and advancement. The strength of the relationship dictates the allocation of resources. Accordingly, jealousy aimed at safeguarding such relationships can be expected to play an important role during all phases of life. Indeed, it is a more efficient process to have a single emotion that is sensitive to rival-induced threats to any established or budding relationships than to have discrete systems designed for each specific type of relationship challenge (DeSteno, 2004; DeSteno et al., 2002). This assertion identifies jealousy as a discrete emotional response to a specific type of anticipated or actual social rejection: rejection by a relationship partner in favor of a rival. Yet it is important to note that although social rejection can take many forms (e.g., ostracism from a group, refusal of admission to a group, relationship dissolution not due to a rival), jealousy and any associated behavioral sequelae can be expected to be intrinsically tied only to the triadic relationship pattern noted in the preceding sentences. Put simply, jealous distress stems from a motivation to protect a relationship from being usurped, and resulting behaviors (e.g., derogation of rivals) center on preventing successful advances of rivals (Salovey, 1991). Other types of social rejection may induce negative emotional states (e.g., shame, anger); how-

627

ever, these states differ from jealousy and any associated behaviors do not center directly on issues of usurpation.

Chasing the Monster Given both its prevalence and painfulness, it is not surprising that jealousy has become one of the more studied social emotions during the past few decades (DeSteno, 2004; Salovey, 1991). Broad interest in this emotion stems not only from the distress it engenders but also from its association with aggressive behavior. Indeed, jealousy, more so than many negative emotions, is thought to lead to hostile and abusive behavior aimed at relationship partners (De Weerth & Kalma, 1993; Mullen, 1996; Paul, Foss, & Galloway, 1993; Schackleford, 2001; White, 1991) and stands as a likely contributing factor to homicide-related deaths among women, with over 40% of such female deaths in 2000 stemming from conflict with relationship partners (U.S. Department of Justice, 2003). In light of the distress and violence associated with jealousy, the need to better understand the psychological mechanisms that determine its intensity is of high import. At present, however, little empirical evidence exists that provides strong support for a specific model of jealousy. Researchers possess an understanding of jealousy’s most general environmental elicitors and phenomenological results but lack clear evidence regarding the intrapsychic processes underlying it. Indeed, the previously prevailing view that jealousy stems from sex-specific, evolved modules sensitive to reproductive threats (see Buss et al., 1992) has encountered formidable theoretical and empirical difficulties that limit its viability (DeSteno, Barlett, & Salovey, in press; DeSteno et al., 2002; DeSteno & Salovey, 1996a; Harris, 2003; Harris & Christenfeld, 1996; Sabini & Green, 2004). In the absence of evidence supporting a specific theoretical model, it becomes difficult to provide a clear and parsimonious account for the intra- and interindividual variation in jealousy known to exist. A primary reason for this void is that the majority of previous research, our own included (e.g., DeSteno & Salovey, 1996b), has relied on predictions regarding the intensity of jealousy one would feel if a relationship partner were, hypothetically, to act in some specified way (DeSteno, 2004; Salovey, 1991). Forecasts of emotional intensity in response to hypothetical events have been shown, unfortunately, to be subject to several biases (Gilbert, Pinel, Wilson, Blumberg, & Wheatley, 1998; Wilson, Wheatley, Meyers, Gilbert, & Axsom, 2000), and, therefore, their use as a primary dependent variable to test models of jealousy is problematic. Simply put, how one thinks she or he may feel in response to the presence of a rival need not necessarily reflect reality. Indeed, attempting to understand the functional influence of an emotion on subsequent cognition and behavior without a true in vivo induction of the emotion is a tenuous enterprise at best. Emotions exert their influence on cognition and behavior through conscious and non1 Although the term jealousy is also used in modern parlance to connote begrudging feelings toward another individual due to his or her possession of some desired object or attribute, this feeling state is more appropriately labeled as envy (Parrott, 1991; Smith, 1991). Jealousy is defined as the negative emotional state generated in response to a threatened or actual loss of a valued relationship due to the presence of a real or imagined rival (DeSteno & Salovey, 1995, 1996b; Parrott & Smith, 1993; Salovey, 1991).

628

DESTENO, VALDESOLO, AND BARTLETT

conscious processes (DeSteno, Dasgupta, Bartlett, & Cajdric, ¨ hman & Wiens, 2003; Schwarz 2004; LeDoux & Phelps, 2000; O & Clore, 1996). It seems unlikely, therefore, that the mediators and associated processes stemming from simply estimating how one would feel if the emotion in question were to be evoked would mirror the effects of the true emotional experience. This argument applies equally to studies that use retrospective assessments of jealousy; in addition to memory biases involving intensity, accurate assessment of proposed mediators becomes problematic, as it is unlikely that such mediators would be properly engaged through simple recall of the event in question. Of course, jealousy researchers recognize these problems. The limiting factor to studying jealousy in real time has been the difficulties inherent in inducing it in the lab. Nonetheless, the induction of in vivo jealousy in an experimental context is necessary in order to test hypotheses regarding mediating mechanisms. Only with the ability to manipulate jealousy and subsequently measure a proposed mediator and behavioral outcomes in real time can strong evidence for a specific theory be marshaled. In its absence, one is left with a reliance on data from hypothetical scenarios or the use of correlational measures that may suggest potential mediators or moderators of jealousy (e.g., personality traits, cultural membership) but are, in themselves, insufficient to establish causality. In the present article, we accept this challenge and attempt to test initial hypotheses derived from a theory of jealousy based on threatened self-esteem through real-time experimental inductions of this emotion. In so doing, we hope not only to provide strong support for a specific theory of jealousy but also to suggest how the proposed theory may hold the potential to integrate previous and seemingly disparate findings regarding the influence of idiographic and cultural factors on this emotion.

Self-Esteem Threat as the Mediator of Jealousy It is our contention that threatened self-esteem is the principal mediating mechanism of jealousy. Schematically, this model shares similarities with many appraisal theories of emotion: Awareness of an event is followed by an appraisal of its significance and then by an ensuing emotional state designed to lead to an adaptive response (cf. Ellsworth & Scherer, 2003; Frijda, 1986; Lazarus, 1991; LeDoux & Phelps, 2000). In the specific case of jealousy, events that have the possibility to arouse this emotion must involve the real or imagined interaction of a relationship partner with a rival. Once an individual becomes aware of any such interaction, an appraisal is made regarding the self-esteem threat posed by it. This event serves as the proximate cause for jealousy, which then leads to behaviors designed to remove the threat. Such an appraisal, of course, need not involve a conscious attempt at assessment; appraisals of emotion-relevant stimuli often occur automatically (Ellsworth & Scherer, 2003; LeDoux & Phelps, 2000). Our central point is that the appraisal centers on the selfsystem, and variations in momentary levels of self-esteem stand as the driving force for jealousy. Although it is a relatively parsimonious model, the question of why the induction of jealousy should depend on or use the self-system necessarily arises. The candidacy of threatened self-esteem as a mediator for jealousy is supported by work suggesting self-esteem’s importance in assessing status in social relationships. Indeed, many have noted that a primary and pancultural determinant of self-esteem is the

perception and evaluation provided by others (Cooley, 1902/1956; A. P. Fiske, Kitayama, Markus, & Nisbett, 1998; Goffman, 1959) and that one of self-esteem’s central functions is to provide an ongoing gauge of one’s status vis-a`-vis relationship partners (Leary & Baumeister, 2000; Leary, Tambor, Terdal, & Downs, 1995). Correspondingly, many emotions related to functioning within the context of interpersonal relationships (i.e., social emotions) have been shown to involve awareness and appraisals of self (Leary, 2003; Tangney & Fischer, 1995; Tracy & Robins, 2004). Knowledge involving other individuals’ evaluations of oneself and their motivations for certain behaviors certainly stands as an integral variable in correctly appraising a given social situation (Flavell, 2004; U. Frith & Frith, 2001; Saxe, Carey, & Kanwisher, 2004). Indeed, differences in theory of mind (i.e., the ability to infer and understand the mental states of others) have been directly linked with social emotions involving self-appraisal. For example, autistic individuals evidence a deficit in recognizing self-conscious emotions (e.g., embarrassment, shame) as opposed to more basic ones (e.g., anger, disgust; Heerey, Keltner, & Capps, 2003). Similarly, individuals with damage to the orbitofrontal cortex demonstrate marked deficiencies in the appropriate experience and regulation of several social emotions (Beer, Heerey, Keltner, Scabini, & Knight, 2003). Given the theorized associations of several interlinked regions of the prefrontal cortex with theory of mind, experience of social emotions, and self-reflective abilities (e.g., Beer et al., 2003; Berridge, 2003; Damasio, 1994; C. D. Frith & Frith, 1999; Gallagher & Frith, 2003; Kelley et al., 2002; Macrae, Moran, Heatherton, Banfield, & Kelly, 2004), such evidence suggests a possible role for the self-system in the experience of social emotions. Awareness and evaluation of the self vis-a`-vis one’s social context may stand as a primary gauge for assessing one’s place within changing social environs and, as such, may be intrinsically tied to the induction and regulation of emotions emergent from social interaction. This view is buttressed by developmental studies demonstrating age-related convergences in the appearance of social emotions, individuated self-awareness, and theory of mind abilities in human development (Dunn, 2003; Lewis, 2000). With respect to jealousy, the role played by self-evaluation may be quite specific. Given that the attention one receives from a partner in a valued relationship is usually taken to signify selfworth (Murray, Griffin, Rose, & Bellavia, 2003; Parrott, 1991; cf. Leary & Baumeister, 2000; Leary, Koch, & Hechenbleikner, 2001), a partner’s interest in a rival stands as a signal that the rival is superior in some way to the self, and, consequently, the integrity of the present relationship may be threatened by the value the partner places on the rival (DeSteno & Salovey, 1996b). Accordingly, jealousy occurs not only when relationships are in the active stage of dissolution but also in the lead up to such an eventuality (Parrott, 1991). If jealousy is to prevent the usurpation of a relationship, then one must be privy to the mental states of partners in order to gauge what their behaviors indicate with respect to their evaluations of possible rivals. That is, individuals must be able to assess what, for instance, a smile or a touch signifies with respect to their partners’ intentions and evaluations. These assessments, then, may function to modulate self-esteem so that what constituted a relatively high level of self-esteem within the domain of the relationship suddenly becomes threatened by certain actions of a partner toward a rival. This threat, in turn, results in a negative emotional state, jealousy, designed to redress the threat (cf. Tesser,

JEALOUSY AND THE THREATENED SELF

1988). Of course, devaluations in self-esteem based on the perceived approval of others may also lead to other negative emotions such as shame (Tangney & Fischer, 1995). It is the coupling of self-esteem threat with the appraisal that it stems from the presence of a rival that provides the requisite factors for jealousy as opposed to other aversive social emotions (e.g., embarrassment). Protection of self-esteem, therefore, serves as an efficient proxy mechanism for the benefits accrued through relationship maintenance; its protection leads to successful navigation through challenges to the integrity of valued relationships and, in so doing, to the protection of the physical and psychological benefits associated with relationships (cf. Baumeister & Leary, 1995; Leary et al., 1995; Schackelford, 2001). Put simply, maximizing self-esteem derived from the views of relationship partners safeguards the more tangible resources stemming from these relationships. Linkage of threatened self-esteem with jealousy also provides an explanation for why jealousy is associated with aggression (Mullen, 1996; Parker, Low, Walker, & Gamm, 2005; Paul et al., 1993). As work by Baumeister and colleagues has revealed, threatening an individual’s self-esteem has the potential to produce an aggressive response, especially when that self-esteem is based on external sources (Baumeister, Bushman, & Campbell, 2000; Baumeister, Smart, & Boden, 1996). Self-esteem threats based on the evaluation of a partner certainly qualify as an external source. Jealousy and associated aggression resulting from such threats, therefore, can be understood to impel one to redress the wrong to one’s sense of honor caused by the partner’s attention to a rival. Such aggression, though normally not socially appropriate or acceptable, may nonetheless serve an adaptive function from an individual’s standpoint if it does prevent the relationship and its associated benefits from being usurped.

The Present Studies As noted, almost all previous research investigating jealousy and any proposed mediators has relied on predictions of jealousy intensity to hypothetical events or on correlational methodologies involving retrospective reports.2 Such strategies limit confident testing of candidate models of jealousy. For instance, our past work investigating the links between self-esteem and jealousy revealed that individuals believe they will be more jealous of rivals who excel in areas of high import to these individuals’ selfconcepts (DeSteno & Salovey, 1996b). Individuals who, for example, place great value on their athletic prowess predict they will be more jealous in response to their partner interacting with an athlete rather than with a musician. Taking a self-esteem maintenance perspective, we argued that jealousy intensity is linked to the threat posed by a rival along dimensions central to self-definition; a partner’s attention to such a rival implies that he or she is superior in the domain of import. However, without a measure of variation in the proposed mediator in real time, such conclusions are difficult to substantiate, especially when they are based on imagined as opposed to actual experiences of jealousy. Several other alternative accounts for our findings could easily be put forth. For instance, individuals might be most jealous of rivals who excel in domains important to their self-definitions simply because they believe that their partners find people who excel in these areas attractive. Self-esteem concerns might not play any role; individ-

629

uals might simply be making strategic judgments about which potential rivals are most likely to peak their partners’ interests. To address such limitations, the current experiments differ substantially from past attempts to link self-esteem to jealousy in two important ways. First and foremost, rather than relying on retrospective or prospective reports of jealousy we had participants experience a jealousy-evoking scenario in the lab through the formation and dissolution of a working relationship. Though effortful to orchestrate, an in vivo experience of jealousy is necessary to infer the causal linkages among the variables in question. In addition to providing the ability to manipulate partner and rival behaviors in ways that would not be readily accomplished through the use of preexisting relationships, the use of this technique also allowed us to control for idiographic factors that may have moderated jealousy intensity if existing relationships were used (e.g., relationship duration, level of commitment). Although these working relationships represented new relationships for participants, they were designed to be very enjoyable and productive. The threatening of such a budding relationship by a rival, consequently, would constitute a relevant scenario for jealousy. Second, we decided to assess self-esteem with both implicit and explicit measures; previous work in this area has only involved explicit measures. Assessment of self-esteem through implicit measures promised to provide a more accurate measure in the present experiments given that an implicit measure is more likely to reflect momentary changes in one’s evaluative stance toward one’s self as a function of one’s salient contingency of self-worth (Crocker & Wolfe, 2001; Greenwald & Banaji, 1995; Greenwald & Farnham, 2000). Indeed, past research has clearly documented the sensitivity of implicit measures of evaluation to changes in context that make different features of a concept more salient (Dasgupta & Greenwald, 2001; Lowery, Hardin, & Sinclair, 2001; Wittenbrink, Judd, & Park, 2001). Therefore, as self-evaluation within the context of the current relationship becomes the salient contingency of worth, corresponding alterations in self-evaluation should be readily captured by using implicit measures of selfesteem (cf. Koole, Dijksterhuis, & Knippenberg, 2001). It is important to note that in using an implicit measure, we are not making any assumptions regarding a lack of conscious awareness of self-esteem. As noted by Greenwald and Banaji (1995), implicit self-esteem may be defined as an attitude toward the self that is either inaccurately identified or outside of awareness. In the present studies, individuals may have been aware of their views and feelings toward the self in response to their partners’ actions or, if not immediately aware, may have had ready access to such information upon reflection. The primary benefit of the use of an implicit measure is that it reduces bias stemming from either a lack of awareness or motivation for positive self-presentation. Explicit measures, given the static and more global nature of their questions, may not be as sensitive to context-induced flexibility. Consequently, they may be less sensitive to the threats that our jealousy manipulation may produce. Threats to self resulting from the manipulations we used would not be expected to alter self-esteem for nonrelationship-relevant contingencies of self-worth (e.g., self2 Work by Volling and colleagues has examined sibling jealousy through in vivo inductions with child samples; however, mediational hypotheses were not examined in these studies (Volling et al., 2002).

DESTENO, VALDESOLO, AND BARTLETT

630

evaluations that are based on intellectual, athletic, or other abilities; cf. Crocker & Wolfe, 2001). Additionally, explicit measures of self-esteem usually require a more deliberate consideration of the self. Such measures, because of their greater controllability, are more amenable to strategic attempts meant to obscure threats to self-esteem (Greenwald & Banaji, 1995; Koole et al., 2001). The basic structure of the two studies is quite similar and involved the formation and subsequent threatening of a valued relationship through the interaction of participants with two confederates: one playing the role of partner and one the role of rival. The jealousy manipulation involved whether the partner indicated interest in working with the rival and ended his or her working relationship with the participant. Following this manipulation, participants completed measures of self-esteem and jealousy. In addition, the second study examined the links between jealousy and direct aggression aimed at partners and rivals.

Study 1 The primary goals of this study were to demonstrate that jealousy can be evoked in a laboratory setting and to investigate whether jealousy is mediated by threats to self-esteem. As noted earlier, the occurrence of jealousy is not limited to romantic relationships; it occurs in relationships of all types involving a valued partner. Accordingly, we expected that after participants formed a novel and pleasant relationship with a work partner, threats to that relationship posed by a rival should produce jealousy. Moreover, we expected that jealousy intensity would vary as a direct function of decreases in self-esteem. The intensity of any resulting jealousy can be expected to be relatively mild as the relationship is quite new. Nonetheless, jealousy should occur whenever there is a threat to even a budding relationship of potential value and, thereby, provide an opportunity to examine the functioning of this emotion in real time.

Method Participants Forty-six female undergraduates at Northeastern University participated in this experiment in partial fulfillment of a course requirement.3 Participants were randomly assigned to either the jealousy or the control condition.

Manipulations and Measures Jealousy manipulation. In order to induce jealousy in vivo, a complex triadic interaction involving the participant was staged through specific actions by two confederates playing the respective roles of the partner and the rival. The details of the induction are noted in the procedure description in the following section as they are integrated with the unfolding of the experimental paradigm. In brief, a confederate playing the role of the partner forms an enjoyable working relationship with each participant. At a later point in the experimental session, the bonds of this relationship are threatened and broken because of either the usurpation of the relationship by a confederate playing the rival (i.e., the jealousy condition) or fate (i.e., the control condition). In all conditions, the partner was male and the rival was female. Implicit self-esteem. Implicit self-esteem (ISE) was assessed with an implicit association test (IAT) based closely on that developed by Greenwald and Farnham (2000). This measure has been shown to possess good reliability and predictive validity with respect to both self-report and

behavioral measures (Bosson, Swann, & Pennebaker, 2000; Greenwald & Farnham, 2000). For example, ISE measures that use the IAT have been demonstrated to predict defensive behavior in response to threats to selfesteem when used to assess narcissism in consort with explicit self-esteem measures (Jordan, Spencer, Zanna, Hoshino-Browne, & Correll, 2003; McGregor & Marigold, 2003). This measure has also been shown to predict clinical status with respect to depression and susceptibility of depressed individuals to contextual changes in self-evaluation and mood (Gemar, Segal, Sagrati, & Kennedy, 2000). In this task, the self-versus-other category was represented by 10 selfrelevant versus 10 nonself-relevant items. The evaluative attribute was represented by 10 pleasant (e.g., joy, peace) and 10 unpleasant (e.g., agony, vomit) words (see the Appendix for the complete stimulus set). Stimuli were presented by using DirectRT software (Jarvis, 2004) on PC-type desktop computers (Intel Pentium III, 550 MHz processors) equipped with CRT color monitors. At the start of the ISE task, each participant provided the self-relevant information items (e.g., last name, student ID) in response to prompts by the computer (see the Appendix for the complete set of prompts). Of importance, these items did not possess any intrinsic positive or negative qualities; any valenced associations would only arise through their association with the self. In order to disallow any sense of personal association with the nonself-relevant stimuli, a set of 10 items matching the form of the self-relevant items was provided for all participants (see the Appendix for the complete list). The assumption of lack of any self-association was checked both through the comparison of generated items and debriefing. After providing this information, participants completed an IAT that assessed self-esteem. Participants were instructed to categorize four types of stimuli (self-relevant vs. other-relevant information, pleasant vs. unpleasant words) by using two designated response keys. Errors were always noted by the appearance of the word error on the screen, after which participants had to press the appropriate key to continue to the next trial. Response latencies for error trials were recorded as the time from stimulus onset to the time of correct categorization (Greenwald, Nosek, & Banaji, 2003). In the first block (20 trials), participants categorized items as belonging to the self or other category. In the second block (20 trials), participants categorized words as pleasant or unpleasant. In the third block (20 practice trials followed by 40 critical trials), participants completed a combined categorization task by classifying informational items as self or other and words as pleasant or unpleasant by using the two keys (for a randomly selected half of the participants, pleasant was paired with self and unpleasant with other; for the other half, this pairing was reversed). In the fourth block (20 trials), participants had to categorize pleasant versus unpleasant words by using the opposite keys to those used in the earlier blocks. Finally, in the fifth block (20 practice trials followed by 40 critical trials), participants again completed a combined categorization task by classifying information items as self or other and words as pleasant or unpleasant by using the two keys. In this block, all participants categorized self–nonself and pleasant– unpleasant stimuli in a manner that was opposite to the stimulus pairing combination used in the third block. To the extent that participants held a positive evaluation of themselves, they should have been faster at associating self-related words with pleasant stimuli and slower at associating self-related words with unpleasant stimuli (Dasgupta, Greenwald, McGhee, & Banaji, 2000; Greenwald & Farnham, 2000; Greenwald et al., 1998, 2003). Scoring of the ISE measure was done in accordance with the D algorithm developed by Greenwald et al. (2003). Each participant’s D was computed by subtracting the mean response time for Block 3 from Block 5 and dividing the resulting quantity by the pooled standard deviation of the two blocks. The D measure may be conceptually understood as an index of individual differences in the degree to which

3

The sample was limited to women due to gender constraints in the participant pool.

JEALOUSY AND THE THREATENED SELF responses for the Self ⫹ Bad trials were slower than those for the Self ⫹ Good trials adjusted for individual differences in the variability of response times. Higher D values indicate higher self-esteem as indexed by increased difficulty in completing the Self ⫹ Unpleasant as compared with the Self ⫹ Pleasant trials. The D metric has been shown to be free from contamination effects due to stimuli ordering and to group differences in task-switching ease (Mierke & Klauer, 2003). Therefore, any resulting differences between the experimental conditions that use this metric cannot be attributed to the effects of simple distraction arising from the use of the jealousy manipulation. That is, differences in D scores did not occur because the jealousy manipulation simply occupied cognitive resources in the jealousy group (e.g., rumination) and, thereby, made it more difficult for individuals to respond to the changing stimulus pairings inherent in the IAT. Explicit self-esteem. Explicit self-esteem was assessed by using the State Self-Esteem Scale (Heatherton & Polivy, 1991). Jealousy. Jealousy was assessed by using a feeling state questionnaire in which participants indicated the degree to which each of 10 adjectives described their current state. The questionnaire consisted of both positive and negative items, embedded in which were four items that specifically targeted jealousy: jealous, angry, betrayed, and hurt (Cronbach’s ␣ ⫽ .81). Parrot and Smith (1993) have demonstrated that these feeling descriptors capture the multifaceted experience of jealousy in a way that is distinct from other related negative emotions (e.g., envy). Participants’ jealousy scores reflect the mean score on these four items.

Procedure Participants were run individually for all sessions. Upon arrival at the lab, the participant (S) was greeted by the experimenter and asked to sit in a chair in front of a cubicle containing a PC. The room contained five such cubicles with an accordion wall that partially expanded so as to separate two cubicles from the other three. Immediately after S entered the room, a confederate playing the role of the partner (P) arrived and was similarly greeted. The experimenter then informed them that two other participants were also scheduled to arrive and that they would therefore wait a few minutes before beginning the session. At this point, P introduced himself to S and began use of predetermined conversational probes that were designed to initiate a sense of familiarity and liking. After 3 min, the experimenter returned and noted that the experiment would begin without the other participants. The experimenter informed S and P that the study in which they would take part was designed to examine differences in task performance levels as a function of working alone or in pairs. Moreover, as some of the tasks would be conducted on computers, the experiment would also involve S and P taking two hand– eye coordination tests that would allow the experimenter to adjust scores for individual differences in hand– eye acuity for computer use. After the first such test, S and P would be free to choose to work together or alone on the first problem-solving task. At this point, S and P were instructed to turn to their computers to complete the first hand– eye coordination task. In actuality, this task was an IAT taken from Greenwald et al. (1998) that assesses positive attitudes toward flowers versus insects. Its only purpose was to familiarize participants with the IAT so that it would require less instruction to complete the ISE measure after the introduction of the critical manipulation. When S and P had finished this IAT, the experimenter returned to the room and provided instructions for the first problem-solving task (in actuality, participants would only complete one such task). This was a word unscrambling task. P and S were handed sheets of paper that contained letter matrices at the top of each. The task was to find as many words as possible that were contained in each matrix. After reminding them that as there were only 2 of them they could choose to work together or alone, the experimenter left the room. P then turned to S and asked if she would like to work together.4

631

The problem-solving task served only as a vehicle to foster the formation of a pleasant working relationship. During the next 5 min, P’s task was to ensure that S enjoyed working with him. He did this through repeated smiling and the use of a set of verbal responses. For example, he would provide encouragement (e.g., “let’s see if we can figure out this one”) and validation (e.g,. “that’s a good one” and “I’m glad we’re doing this together”) to the participant. After 5 min had passed, a knock was heard at the door and the experimenter appeared from a side room to answer it. The confederate playing the rival (R) then entered the room and apologized to the experimenter for being late. The experimenter informed R that she would complete the earlier hand– eye coordination task at the end of the experiment, handed her a clipboard containing the materials for the word scramble task, gave brief instructions for it, and left the room. R then grabbed a chair and sat next to S and P. For the next 3 min, the three individuals worked together. However, R was instructed to devote most of her attention and interactions (i.e., validations and encouragements) to P. At this point, the critical manipulation occurred. In the jealousy condition, P suddenly noted that he thought the experimenter said they could only work alone or in pairs. After expressing concern that this could be a problem, he went into the next room and asked the experimenter. The experimenter and P returned to the room at which point the experimenter noted that they could only work in pairs or alone before turning to leave. P then turned toward R and asked if she would like to continue as his partner. R agreed and the two moved to the other side of the room (i.e., behind the partially expanded accordion wall) and continued working within earshot of S for 1 min. In the control condition, P suddenly noted that he had an appointment at the campus medical center that he had forgotten. He then went into the next room to tell the experimenter who could be heard excusing him with the caveat that he return later to finish the study. In this way, the enjoyable working relationship was severed in both conditions. However, in one it was due to the presence of a rival and in the other to consequences of fate. At this point in both conditions, the experimenter then returned to the room and instructed the individuals to turn toward their individual PCs and to follow the instructions provided. Participants then completed the implicit and explicit measures of self-esteem, the jealousy scale, and a questionnaire concerning demographic information. The confederate(s) always left the experimental room before the participant had finished. Upon completion of the study, participants were extensively debriefed and given a small gift of candy for their participation.

Results and Discussion In accord with expectations, the termination of a relationship due to a partner leaving to work with a rival as opposed to leaving for a scheduling conflict was successful in evoking jealousy. Participants reported higher levels of jealousy in the jealousy condition (M ⫽ 1.65, SD ⫽ 0.89) than in the control condition (M ⫽ 1.21, SD ⫽ 0.30), t(44) ⫽ 2.21, p ⫽ .03. It is instructive to note that the variation in reported jealousy is quite large in the jealousy condition relative to the control condition.5 Thus, even though the mean level of self-reported jealousy in the jealousy condition falls in the mild to moderate range, it masks a high degree of variability. Such differences in variability are to be expected given the lack of jealousy in the control condition and individual differences in self-presentational concerns related to the stigmatizing nature associated with admitting to jealous feelings 4 In all cases except one, this proposal was accepted. Data from the participant who chose to work alone were discarded from all analyses. 5 A t test assuming unequal variances for the two groups also showed a significant difference in jealousy (t ⫽ 2.29, p ⫽ .03).

DESTENO, VALDESOLO, AND BARTLETT

632

Figure 1. errors.

Implicit self-esteem as a function of the jealousy manipulation in Study 1. Error bars depict standard

(Mathes et al., 1982; Wiederman, Allgeier, & Ragusa, 1995). Nonetheless, the manipulation was quite strong (Cohen’s d ⫽ 0.73), indicating that the two distributions were relatively separable. Given that jealousy had been successfully induced in the lab, the next question centered on the potential of threatened self-esteem to function as the mediator of this emotion. In accord with predictions, a partner’s leaving for a rival resulted in a decrease in implicit self-esteem. That is, participants demonstrated lower ISE scores when the partner left for the rival (MD ⫽ 0.55) than when he left for an appointment (MD ⫽ 0.82), t(44) ⫽ 2.57, p ⫽ .01, d ⫽ 0.75. For ease of interpretability, Figure 1 presents the response latencies in the ms metric for the IAT (Me ⫹ Good and Me ⫹ Bad) blocks as a function of jealousy condition.6 The extent to which Me ⫹ Bad response times exceed Me ⫹ Good in the control as compared with the jealousy condition stands at 120 ms, thereby indicating a lowered association of the self with positivity in the jealousy condition.7 Examination of explicit self-esteem scores revealed no differences as a function of jealousy condition. As noted, this finding was to be expected given the broader focus of most self-esteem scales across individuals’ sets of contingencies of self-worth. Demonstration of condition differences in ISE do not, of course, directly imply mediation of jealousy by threatened self-esteem. We therefore conducted a mediation analysis following the usual procedures (Kenny, Kashy, & Bolger, 1998). Zero-order correlations and regression beta weights are shown for the predicted mediational model in Figure 2. As expected, significant zero-order

correlations existed among all three variables. However, when jealousy intensity was regressed on ISE and condition, only ISE remained a reliable predictor. Supporting the view of complete mediation, the ability of the actions of the partner and the rival to induce jealousy possessed no causal efficacy beyond that explained by the manipulation’s ability to threaten self-esteem (Sobel Z ⫽ 2.08, p ⫽ .04). Indeed, as self-esteem decreased in response to the partner’s interest in the rival, jealousy intensity correspondingly increased. These findings provide strong initial support for the theory of jealousy that we advocate. They represent the first direct evidence of the role played by threatened self-esteem in the evocation of jealousy. Of great import, they demonstrate a rapid decrease in self-esteem in response to the favorable interaction of the partner and rival that is directly associated with the intensity of jealousy experienced. Nonetheless, given the novelty of the methodology and findings, they bear replication and extension before greater confidence can be placed in the proposed theory.

Study 2 In this study, we sought not only to replicate the findings of Study 1 but also to assess further the proposed theory of jealousy through examining its predictive validity with respect to a frequent behavioral correlate of this emotion: aggression aimed at partners and rivals (Mullen, 1993, 1996; Parker et al., 2005). Finding a positive association between the jealousy induced in our participants and any aggressive behavior would further support the construct validity of both our procedures and model of jealousy. Of greater theoretical import, however, would be the ability of the 6

Figure 2. Implicit self-esteem as a mediator of jealousy. Coefficients in parentheses indicate zero-order correlations. Coefficients not in parentheses represent parameter estimates for a recursive path model containing both predictors. Asterisks indicate parameter estimates that differ from zero at p ⬍ .05. Jealousy condition is dummy coded (control ⫽ 0, jealousy ⫽ 1).

Although we used the psychometrically more robust D scoring procedure for ISE, it is instructive to note that the Condition ⫻ Block Type interaction is also significant ( p ⬍ .05). 7 Karpinski (2004) has noted that use of IAT techniques to assess ISE may be influenced not only by participants’ self-evaluations but also by their evaluation of the target other. That is, participants’ ISE could appear, as opposed to be, lower if the target other varied across individuals. This concern is not relevant in the present case. The information about the other (e.g., name, nationality, student ID) was constrained to be anonymous and held constant across groups. Thus, any resulting group differences in ISE must reflect changes in self-evaluation.

JEALOUSY AND THE THREATENED SELF

experimental nature of our paradigm to examine the causal relations between self-esteem threat, jealousy, and any resulting aggression in this triadic interaction. It is precisely the correlational nature of past research that has suggested an association between jealousy and aggression that has limited its ability to demonstrate that jealousy causes aggression. Indeed, several plausible alternatives exist. Specific individuals, for example, might be both more susceptible to jealousy and more aggressive without one factor mediating the other; threats to self-esteem might influence each factor directly (cf. Baumeister et al., 1996; Twenge, Baumeister, Tice, & Stucke, 2001). The inclusion of a direct aggression measure in the present study has the potential to illuminate propositions concerning the causal efficacy of jealousy to produce hostile behavior and, in so doing, will provide a clear test of the proposed model of jealousy.

Method Participants Forty-three undergraduates (30 female, 13 male) at Northeastern University participated in this experiment in partial fulfillment of a course requirement. Participants were randomly assigned to either the jealousy or the control condition.

Manipulations and Measures With the exception of the aggression measure, all manipulations and measures were identical to those used in Study 1. Given that this sample contained members of both genders, the confederate playing the partner was always of the opposite gender to the participant; the confederate playing the rival was always of the same gender. To assess aggression, we used a paradigm slightly modified from that developed by Lieberman, McGregor, and colleagues (Lieberman, Solomon, Greenberg, & McGregor, 1999; McGregor et al., 1998) in which participants were given the opportunity to inflict pain on others through deciding on the amount of hot sauce, a substance known to be potentially painful and disliked by the target others, that would be placed in the others’ mouths. The primary modification involved changes necessary for assessing aggressive behavior toward two individuals as opposed to one. The details of this measure are explained in the procedure section below. The amount of hot sauce was measured in grams by using preweighed containers on an Ohaus Adventurer Pro digital scale (Model AV212, Pinebrook, NJ) with a maximum weight capacity of 210 g and precision of 0.01 g.

Procedure With the exception of two changes necessary for implementing the aggression measure, the procedure was identical to that used in Study 1. One change, involving a ruse needed to collect taste preferences for the aggression measure occurred before the beginning of the Study 1 procedure. The second, involving the opportunity to engage in aggression, occurred at the conclusion of the Study 1 procedure. Between these two events, the Study 1 procedure unfolded as previously described. The two changes are detailed below. At the start the experiment, the participant and confederate playing the role of partner entered the room. After waiting to no avail for the arrival of the “other participants,” they were informed that they would be taking part in two unrelated studies: one investigating the effects of working alone or in pairs on task performance and one involving the relation of personality to taste preferences and acuity. At the start of the experiment, they would first complete a brief personality measure (i.e., a bogus 5-item measure

633

asking about hours per day spent watching television, enjoying outdoor pursuits, working on academic endeavors, etc.) and then a questionnaire that was designed to assess their degree of liking for several tastes: sweet, sour, creamy, salty, spicy, and fruity. Liking was assessed by using a 21-point scale ranging from 1 (don’t like at all) to 21 (extremely like). Participants were then informed that later in the session, each of them would be randomly assigned to provide a taste sample for other participants of a single item from a smaller subset of these categories. In order to provide an explanation for why participants would be making the taste sample, they were told that this procedure would allow the experimenter to remain blind to certain aspects of the experiment. After completing the personality and taste preference measures, the remainder of the procedure unfolded as in Study 1. The only modification in this section involved the rival completing the personality and taste preference measures before joining the participant and partner in the word scramble task. After completing the remainder of the Study 1 paradigm, the participant and the remaining confederate or confederates (both the rival and the partner in the jealousy condition or only the rival in the control condition) were told that it was now time to complete the taste preferences study. In the control condition, the participant and confederate were told that the partner had agreed to return to complete the taste preference study after his or her medical appointment. The experimenter then handed each of them a box that contained three food items, two empty sample containers, information regarding the food item they were to prepare for each of the other participants (i.e., the partner and the rival), the taste preference questionnaires of the two other people in the session, and the food category that they (i.e., each participant) were assigned to taste. Participants were informed that they would be allowed to see the others’ preferences as people are often curious about what others’ taste preferences might be. At this point, the experimenter told them that each person would go to a separate room to place the samples in the containers. After everyone had finished, the experimenter would return to each of them, place the designated samples as produced by the other participants into their mouths and ask them to fill out questionnaires regarding their attitudes toward these food items. Of importance, they were told that the entire contents of the sample containers would be placed in each of their mouths. At this point, the experimenter told the real participant that he or she would create the sample in the current room and offered the food item box. The other confederate(s) was then led out to a supposed different location. Upon opening the box, the participant saw the two food preference questionnaires from the partner and rival. They had no names indicated on them, but were identifiable by the first item: the gender of the person. Both questionnaires indicated a liking of 3 on the 21-point scale for spicy foods (on which 1 indicated no liking at all). After removing these questionnaires, the participants saw a written set of instructions, two sample containers, and three labeled food items: sweet (chocolate syrup), fruity (fruit punch), and spicy (hot sauce bottle with a fiery label and “hotness” warnings). The instructions noted that each of the two other participants (confederates) had been randomly assigned to receive spicy samples. Participants were reminded that the others would not know who had prepared each sample and that the entire amount in each sample cup would be placed into each of the other’s respective mouths. Participants were then instructed to pour any amount of the hot sauce into two containers labeled male and female, respectively. They were then to place a cover on the sample containers, place them in the box, and return the box to the experimenter. At this point, the experiment ended and participants were fully debriefed.

Results and Discussion As in Study 1, the jealousy manipulation resulted in increased jealousy, t(41) ⫽ 2.46, p ⫽ .02, d ⫽ 0.78, and lowered self-esteem,

DESTENO, VALDESOLO, AND BARTLETT

634

Figure 3. errors.

Implicit self-esteem as a function of the jealousy manipulation in Study 2. Error bars depict standard

t(41) ⫽ 2.08, p ⫽ .04, d ⫽ 0.64.8 Participants reported more jealousy when the partner left for a rival (M ⫽ 1.57, SD ⫽ 0.61) than for a medical appointment (M ⫽ 1.19, SD ⫽ 0.35).9 Similarly, they also evidenced lower ISE in response to the partner leaving for a rival (MD ⫽ 0.53, SD ⫽ 0.47) than leaving for an appointment (MD ⫽ 0.80, SD ⫽ 0.37). For ease of interpretability, Figure 3 depicts the response latencies in the ms metric for the IAT Me ⫹ Good and Me ⫹ Bad blocks as a function of jealousy condition.10 The degree to which Me ⫹ Bad response times exceeds Me ⫹ Good in the control as compared with the jealousy condition is 141 ms, thereby indicating, as was the case in Study 1, a lowered association of the self with positivity in the jealousy condition. Once again, no differences were evident on the explicit self-esteem measure. In turning to an examination of the aggression measure, a 2 (jealousy condition) ⫻ 2 (gender) ⫻ 2 (target: partner vs. rival) mixed ANOVA provided clear evidence for the predicted main effect. Participants aggressed toward the partner and the rival to a much greater degree in the jealousy condition (M ⫽ 3.41 g) than in the control condition (M ⫽ 1.44 g), F(1, 39) ⫽ 8.60, p ⬍ .01, d ⫽ 0.77. No differences emerged as a function of the target of the aggression; hostility was aimed equally at the partner and the rival. A main effect of gender also emerged; hot sauce samples produced by men were larger on average (M ⫽ 4.24 g) than were those produced by women (M ⫽ 1.67 g) across conditions, F(1, 39) ⫽ 8.31, p ⬍ .01, d ⫽ 0.87. This effect, though not explicitly predicted, may reflect either a stable gender difference in taste preference or portion allotment, or a more general tendency among men to act more aggressively irrespective of provocation (Eagly & Steffen, 1986).11 No other reliable effects emerged. The findings involving jealousy and ISE closely mirror those of Study 1 and, in so doing, provide strong support for the proposed role of self-esteem. Moreover, the demonstrated differences in aggression provide the first experimental evidence documenting a link between jealousy and aggressive behavior. Two important and intertwined issues, nonetheless, remained. These involved a repeated demonstration of the mediational role played by self-esteem in jealousy intensity and, more importantly, an examination of whether jealousy resulting from threatened self-esteem would mediate aggression aimed at the partner and rival. To examine these issues, we specified the recursive path model depicted in Figure 4. In this model, each of the respective potentially causal variables

(i.e., the jealousy manipulation, ISE, and jealousy intensity) is allowed to influence all downstream variables. That is, we allowed direct causal paths from the jealousy manipulation to ISE, jealousy intensity, and aggression; from ISE to jealousy intensity and aggression; and from jealousy intensity to aggression. In this way, the potential causal influence of each variable on those that are subsequent to it in the causal sequence can be assessed, thereby allowing us to test the viability of the proposed causal model. In addition, given the influence of gender on aggressive behavior irrespective of the jealousy manipulation, we also specified a direct path capturing this relation. Aggression here was defined as the mean level directed against partners and rivals. AMOS (Version 5.0; Arbuckle, 2003) was used to generate parameter estimates with a maximum likelihood algorithm. The resulting model fit the data quite well, ␹2exact fit(3, N ⫽ 43) ⫽ 2.66, p ⫽ .45; root-mean-square error of approximation ⬍ .01.12 As can be seen in Figure 4, the causal chain that emerged matched the predicted model: A partner leaving for a rival led to lowered self-esteem, which led to greater jealousy, which led to increased aggression aimed at the partner and rival. These findings are of great import, as they not only replicate Study 1’s demonstration of the mediating role played by threatened self-esteem in the evocation of jealousy but also provide empirical evidence of the mediating role of jealousy in eliciting aggression. Threatened selfesteem did not directly lead to hostility aimed at the sources of the threat but rather engendered an aversive emotional state that, in 8 Although we did not expect participant gender to have any individual or interactive effects on jealousy and threatened self-esteem, we also submitted the data to a factorial ANOVA with gender as a second predictor to examine this possibility. Gender did not differentially influence either variable. 9 This difference was also significant using a t test assuming heterogeneity in group variances (t ⫽ 2.49, p ⫽ .02). 10 A mixed ANOVA treating the IAT blocks as a repeated factor also produced a significant Condition ⫻ Block interaction ( p ⬍ .05). 11 A similar gender difference with the hot sauce measure was reported by Evers, Fischer, Mosquera, and Manstead (2005). 12 Supporting this view, constraining the nonsignificant paths to zero also resulted in a well-fitting model, ␹2exact fit(6, N ⫽ 43) ⫽ 9.86, p ⫽ .13, and a negligible decrement in fit, ⌬␹2(3, N ⫽ 43) ⫽ 7.20, ns.

JEALOUSY AND THE THREATENED SELF

Figure 4. Recursive path model specifying linkages among each predictor and all downstream variables. Black paths and coefficients indicate parameters that reliably differ from zero at p ⬍ .05. Grayscale paths represent nonsignificant relations. Jealousy condition and gender are dummy coded (control ⫽ 0, jealousy ⫽ 1; men ⫽ 0, women ⫽ 1).

turn, led individuals to inflict pain on those responsible for the threat.

General Discussion The findings of these two studies provide evidence for a theory of jealousy based on threatened self-esteem. In each case, a decrease in self-esteem occurred in real time as a function of a partner showing favor for a rival, and, of central import, this decrease directly mediated the intensity of jealousy experienced. Jealousy, moreover, was shown to mediate actual aggression aimed at partners and rivals, thereby providing construct validation for our jealousy induction procedure and identifying, for the first time, a direct causal link between jealousy and aggressive behavior. As we noted earlier, previous work has revealed a link between threatened self-esteem and aggression (Baumeister et al., 1996). This relation, we believe, supports the identification of threatened self-esteem as a principal mediator of jealousy, and in light of the current findings, points to the important role that may be played by emotion in mediating such outcomes. It is also worth reiterating that our use of an implicit measure of self-esteem is not meant to imply that individuals will not possess a conscious awareness of lowered self-evaluation or a feeling of inferiority. Our primary reason for the use of an implicit measure to assess alterations in self-evaluation involves its sensitivity to rapid context-induced changes with respect to currently salient features of self. In essence, it provides a measure of self-esteem that is responsive to whichever features of self constitute the working self-concept at a given moment (cf. Crocker & Wolfe, 2001; DeSteno & Salovey, 1997). It is possible that a carefully designed and validated explicit self-esteem measure targeted to a self-contingency directly related to the individual–partner–rival interaction might produce parallel findings. However, one would also need to consider the influence on this measure of strategic attempts to obscure evidence of feelings of threat.

Jealousy and Social Rejection The present findings may indicate an important role for jealousy in the study of social rejection. We believe that jealousy represents a specific emotional response to a specific form of social rejection: the actual or looming rejection by a partner in favor of a rival. In accord with our expectations, the present findings demonstrate the occurrence of this specific emotion and associated hostile behav-

635

iors.13 These findings raise the question of how jealousy relates to the expanding literature examining the phenomena of rejection and ostracism (e.g., Leary & Baumeister, 2000; Twenge et al., 2001; Williams, 1997; Williams et al., 2000). For example, work by Baumeister and colleagues (Baumeister, DeWall, Nathan, & Twenge, 2005; Twenge et al., 2001) has demonstrated that rejection often leads to several negative behavioral outcomes, including aggression. It is interesting that findings demonstrating emotional responses to rejection and their role in mediating subsequent behavior have been more mixed. Research by Baumeister and colleagues has repeatedly found little if any increases in selfreported negative affect in response to manipulations of rejection. Given that a motivation to engage in social relationships appears to be a fundamental drive (Baumeister & Leary, 1995; S. T. Fiske, 2004), the lack of strong evidence for an emotional response to relationship threats is somewhat surprising. However, recent work by Eisenberger, Lieberman, and Williams (2003) has produced evidence suggesting a link between rejection and a negative emotional response. Participants in their study who experienced rejection from a social triad demonstrated heightened activation of brain centers associated with the experience of pain. One reason underlying these divergent findings may involve the different methods commonly used to induce rejection. For example, false feedback that one is likely to be lonely in the future may hold different affective consequences as compared with active exclusion by social beings. Another reason may involve a reliance on measures of negative affect (e.g., global negative mood scales that assess general negativity or dysphoria) to assess participants’ emotional states as opposed to more discrete negative experiences such as jealousy. Active exclusion, for example, might well lead to feelings of anger but not to feelings of anxiety or sadness. Measures of global negativity might therefore be less sensitive to intensity differences in specific negative states (see Tracy & Robins, 2004, for a similar argument). To our mind, the type of emotional response that is elicited depends greatly upon the exact form of social rejection that is experienced. All social rejection is threatening and to be avoided; however, the adaptive responses and associated emotional states that are required vary depending on the specific nature of the threat. Similar mixed findings occur with respect to the effect of rejection on self-esteem. Work by Twenge and colleagues (Twenge et al., 2001; see also Baumeister et al., 2005) found little evidence that rejection influences self-esteem. Yet, work by Williams suggests that self-esteem is lowered in response to rejection (Williams, Cheung, & Choi, 2000; see also Leary et al., 1995). Here again, we suspect that some of these differences may depend 13 Given that our composite jealousy measure did contain descriptors of feeling items that might be relevant to any type of social rejection (i.e., hurt, anger), we also undertook an item-by-item analysis to be certain that the reported jealousy differences were not stemming solely from differences with respect to these feeling states. Through the use of Stouffer’s meta-analytic test on data from both studies, it is clear that the emotional states experienced by participants were characterized by jealousy. The jealousy manipulation produced significant differences on the items jealous (Z ⫽ 2.30, p ⫽ .02) and betrayed (Z ⫽ 3.21, p ⬍ .01) and marginal differences on the items angry (Z ⫽ 1.68, p ⫽ .10) and hurt (Z ⫽ 1.77, p ⫽ .08).

636

DESTENO, VALDESOLO, AND BARTLETT

on the manipulations used to induce rejection as well as on the nature of the measurement of self-esteem. The present research offers a unique perspective on these issues. To our knowledge, there has been little research in this area that has simultaneously investigated the causal links between the three constructs examined here: self-esteem threat, emotional response, and aggression. Rather, research has focused on subsets of linkages between these constructs. Some evidence supports each link: (a) self-esteem has been shown to be threatened by rejection, (b) rejection has been shown to produce aggression, and (c) rejection has been shown to be emotionally painful. Several different causal models specifying the relations among these variables could be proposed on the basis of these findings. In fact, excluding the present findings, there is no extant evidence that, as we would propose, rejection leads to a decrement in self-evaluation which leads to jealousy (in the case of rejection from an existing relationship in favor of a rival), which leads to aggression. Indeed, the potential for a negative feeling state to occur and mediate aggression in response to social exclusion was examined only in Twenge et al. (2001), in which little support was provided for this view. In that study, however, negative affect was assessed by using a global as opposed to a discrete measure. Jealousy, we believe, may be a linchpin that holds many of these phenomena together. The current findings suggest that it may function as the warning and impetus to protect valued relationships from being usurped. As self-esteem fluctuates in response to a partner showing greater interest in a rival, it serves as a proxy to assess the adaptive challenges resulting from threats to the relationship and, in turn, spurs a highly aversive emotional state designed to shunt thought and action toward preservation of the relationship. Jealousy, of course, is not the only negative emotion that may result from rejection. The presence of an existing relationship is required for its evocation, but, as noted, one may experience rejection in other ways as well. Other situations (e.g., refusal of admission into a social group) would, in all probability, lead to the experience of aversive states but not jealousy per se. The functional purpose of jealousy is intrinsically tied to behaviors designed to protect the integrity of a relationship (e.g., derogation of rivals, aggression toward partners and rivals). Given that rejection from an existing relationship due to the presence of a rival stands as one of the canonical sources of rejection in human life, the present findings suggest that jealousy may play a fundamental role in linking self-esteem threats from interpersonal rejection to aggression. Threats of loss not due to a rival may be expected to engender other emotion-mediated behaviors aimed at maintaining the relationship (e.g., greater attempts at attraction, tears or other signs of a need for succor).

Integrating a Disparate Literature At the outset of this article, we noted that the jealousy literature lacks consensus with respect to a broad theoretical framework. Although researchers agree on jealousy’s phenomenology, consensus regarding its underlying causes and mechanisms has been much more difficult to find. To date, jealousy’s causal mechanisms have been posited to depend on evolved sex-specific modules (Buss et al., 1992; Buunk et al., 1996; Wiederman & Allgeier, 1993), stable idiographic traits (Bringle, 1991), correlates of the attachment system (Collins & Read,

1990; Sharpsteen & Kirkpatrick, 1997), self-evaluation maintenance processes (DeSteno & Salovey, 1996b; Salovey & Rodin, 1984), and culturally learned syndromes (Hupka, 1991; Hupka & Ryan, 1990). Although not constituting an overarching theory, findings associated with each perspective clearly document variability in jealousy as a function of individual and cultural differences. In the current studies, although we provided evidence that threatened self-esteem mediates jealousy, we investigated none of these other factors (e.g., cultural membership, attachment style), leading questions to arise concerning whether these disparate findings can be integrated by using the proposed framework. In considering this issue, it is important to note that any mechanisms that underlie jealousy must evidence a high degree of flexibility. Individuals’ relationship partners, whether they be lovers, friends, parents, or coworkers, regularly interact with scores of people in myriad ways. Spouses socialize with business associates, parents play with multiple siblings, and friends have dinner with other friends. Sometimes such events evoke jealousy; sometimes they do not. Some individuals habitually react jealously; others often display a confident security. Events that cause jealousy among members of a certain culture are of no concern to members of another. Therefore, although the interest of a partner in a rival stands as the most basic factor in the elicitation of jealousy, many influences may function to modulate the resulting emotional experience. Thus, in the face of such countless and seemingly irreconcilable variants (see Salovey, 1991, for a comprehensive overview), contextual plasticity must be the essence of any potential mediator. It is our contention that a model of jealousy based on threatened self-esteem readily provides a mechanism to allow for the incorporation of universal, dispositional, and cultural influences in the determination of what types of actions by one’s partner evoke jealousy and, thereby, has the ability to explain many extant findings in the literature. For example, significant cultural and subcultural variability exists with respect to the types of behavior that evoke jealousy (Buunk & Hupka, 1987). Among the Todas of India or the “swinger” subculture in Europe and the United States, for instance, extra-dyadic interactions of certain types are accepted practice (Buunk, 1991; Rivers, 1906). It is interesting that jealousy is often staved off in such cultures through affirmations that a partner’s extradyadic liaisons reflect recreational needs and not a devaluation of the nonparticipating partner’s worth (Buunk, 1991). Moreover, cultures of honor regularly view certain types of extradyadic liaisons by one’s partner as resulting in one’s loss of honor and self-esteem and, correspondingly, also accept and expect more frequent aggression aimed at partners and rivals in such situations (Vandello & Cohen, 2003). In light of these and other related findings, a logical argument for the mediating role of self-esteem may be made. For example, individuals reared in a culture of honor acquire a heightened sensitivity to events that might result in a challenge to their status or public standing (Nisbett & Cohen, 1996). Consequently, one might readily expect their self-esteem to be more threatened by a partner showing interest in another, resulting in the more intense jealousy and associated aggression seen in members of this cultural group (cf. Vandello & Cohen, 2003). Similar actions by a partner, though certainly not pleasant

JEALOUSY AND THE THREATENED SELF

to anyone, might be expected to result in a less intense threat to an individual raised in a culture of law. The case is similar with known idiographic effects on jealousy. For example, variation in attachment styles has been linked with differential jealousy (Buunk, 1997; Sharpsteen & Kirkpatrick, 1997) and with differential self-esteem (Bartholomew & Horowitz, 1991; Brennan & Bosson, 1998; Collins & Read, 1990). More specifically, individuals who are not securely attached exhibit more intense or frequent jealousy (Guerrero, 1998; Sharpsteen & Kirkpatrick, 1997), lower self-esteem (Bartholomew & Horowitz, 1991; Collins & Read, 1990), and increased odds of aggressive behavior toward partners (Dutton, Saunders, Starzomski, & Bartholomew, 1994). Given that attachment styles can be conceptualized as mental models of the self in relation to significant others and that they have been found to shape interactions with valued relationship partners at all stages of life (Bartholomew & Horowitz, 1991; Bowlby, 1973; Fraley, 2002; Hazan & Shaver, 1987; Pietromonaco & Barrett, 1997), it seems likely that such models may color the interpretations of a partner’s interactions with a rival. Individuals whose attachment is characterized by more anxiety may be more likely to believe given interactions of their partner with potential rivals signal a greater valuation of such rivals vis-a-vis themselves. Thus, threatened self-esteem might play a role in the relation between jealousy and aggression in less securely attached individuals. A similar argument may pertain to individual differences in rejection sensitivity. Rejection sensitivity refers to a dispositional tendency to expect, readily perceive, and anxiously or angrily react to rejection (Downey & Feldman, 1996). In line with this tendency, heightened rejection sensitivity has been associated with heightened jealousy and aggression aimed at relationship partners (Downey & Feldman, 1996; Downey, Feldman, & Ayduk, 2000). Here again, a dispositional tendency toward rejection sensitivity may exert its influence on jealousy through increasing the likelihood that interactions by one’s partner with a potential rival are interpreted as threats to one’s self-esteem; rejection, here, implies a sense of inferiority to the partner’s other options. Given the complexity and variability of human social systems, we believe that threat assessment must be based on inputs from multiple systems. That is, self-esteem threats must be assessed with respect to universal, idiographic, and cultural determinants. For example, becoming aware that one’s partner is holding the hand of his or her sibling presents a very different threat possibility than becoming aware of his or her holding the hand of an unrelated individual. Similarly, specific interactions of men and women in one culture may portend distinctly different consequences than in another culture. Accordingly, self-esteem threat assessments may be derived from appraisal systems shaped by universal signals (e.g., nonverbal cues emitted by a partner and rival), idiographic factors (e.g., attachment style, rejection sensitivity), and cultural expectations (e.g., proscribed types of contact). In short, threats to self-esteem may be jointly determined by systems working in a synergistic or oppositional dynamic. Herein lies the benefit of using self-esteem as a proxy to assess threat. Although the motive to protect self-esteem (i.e., valuation by others) most likely stands as a biological universal, tuning of the system with respect to the factors that imply threat remains open for much input through idiographic social learning and acculturation. Put simply, social experience functions to fine tune interpretations of threat based on

637

one’s environs, resulting in greater efficiency at predicting and preventing a problem of significant adaptive consequence.

Future Priorities and Directions The present studies open wide avenues for examining jealousy and its behavioral sequelae. As just noted, investigation of the interplay of universal, idiographic, and cultural factors on jealousy stands as an area of high import both for further validation of the advocated theory and for increased understanding of the ways in which such multilevel factors shape the experience of social emotions in general. Humans are a social species, but we are also one that shows large variation with respect to cultural ethos. Social emotions, therefore, can be expected to be sensitive to these interwoven influences and the unique requirements they hold for adaptive functioning. These initial findings also call for replications involving relationships of a more long-standing nature. The ability to find jealousy within newly formed relationships is not surprising. In order for any relationship to become established, it must pass through initial formation stages. If a motive to protect such budding relationships did not exist, the benefits that are yet to come could not be realized. Consequently, jealousy aimed at guarding such relationships makes great sense, as these initial stages may represent one of the most vulnerable periods for filching by rivals. Nonetheless, it will be important to assess further the degree to which threatened self-esteem functions as the sole mediator of jealousy. One could argue that the complete, as opposed to the partial, mediation demonstrated for self-esteem in the current paradigm may stem from the use of novel relationships. That is, the loss of the partner to a rival possessed no risks beyond those to self-esteem. The relationship was not associated with other benefits or resources; there were no issues involving finances, mutual friends, or progeny. It is our expectation that more intense jealousy would occur with relationships of greater value and investment. However, we do not expect the causal efficacy of self-esteem to change. More intense jealousy most likely results from the greater weight one places on the views of well-loved partners in determining self-evaluation (cf. Murray et al., 2003). It may be quite true that financial or familial concerns with respect to relationship dissolution lead to intense negative emotions such as fear or sadness. These emotions, however, are likely to occur whether or not a relationship is threatened by a rival. Such concerns are relevant if a partner is ending a relationship for any reason; they are not uniquely dependent on the existence of a usurping rival. Jealousy, however, does require the presence of a rival. Consequently, we would anticipate that emotional experiences associated with threats to established relationships may be experienced as more aversive because of both greater jealousy resulting from threatened self-esteem and from the comorbidity of other loss-relevant emotions (e.g., fear, sadness). Indeed, work by Drigotas, Rusbult, Wieselquist, and Whitton (1999) has suggested that strongly valued partners exert a high degree of influence on the shaping of an individual’s self-concept through helping to define the nature of one’s ideal self. Consequently, any implied threats to the status of this ideal self may be quite painful and induce not only jealousy, but also feelings of dysphoria and depression (cf. Higgins, 1987). Of course, to the degree that relationships are characterized by strong mutual levels

DESTENO, VALDESOLO, AND BARTLETT

638

of commitment and investment, the modal level of well-being and the use of associated strategies (e.g., forgiveness) can be expected to limit the occurrences, but not the intensity, of jealousy (cf. Finkel, Rusbult, Kumashiro, & Hannon, 2002; Wieselquist, Rusbult, Foster, & Agnew, 1999). Consequently, we expect that an examination of threats to established relationships, though much more difficult to orchestrate, would result in a magnification of the findings presented here and, thereby, increase their generalizability. Finally, development of additional measures of jealousy intensity stands as an important goal for future research. In the present case, we relied on the use of self-report measures of emotion. Although selfreports of emotion have clearly been demonstrated to be a valid assessment tool (Barrett, 2004), they are, at times, subject to selfpresentational concerns. In the present case, such concerns may have led participants to underreport the intensities of jealousy they experienced because of the somewhat stigmatizing nature of this emotion. The development of alternative measures may be complicated by the probable fact that jealousy, like many more complex social emotions, is not associated with a specific, static facial expression (cf. Keltner & Buswell, 1997). Nonetheless, it may be possible to gauge its intensity through the coding of dynamic changes in expression and other nonverbal channels that, when taken together, comprise the blended phenomenological experience of jealousy (e.g., blends or sequences of anger and anxiety). Indeed, the development and use of a multiindicator assessment of jealousy that combines self-report and nonverbal measures with hormonal markers of emotional stress may provide a window into jealousy intensity that is less constrained by the methodological limitations associated with any of these strategies used in isolation.

Coda As noted in the preceding paragraphs, much work does remain to be done. At present, however, we feel that we have obtained a glimpse into the heart of the green-eyed monster. It is a heart built on two fundamental and interlinked motives. The first is the desire to feel good about the self; the second is the necessity to be engaged in beneficial relationships for which the first serves as a proxy. To sate these motives is to protect much that is important to social living at all stages of life. To threaten them is to signal possible problems of high consequence to well-being and, therefore, to whet the retributive appetite of Shakespeare’s monster.

References Arbuckle, J. L. (2003). AMOS, Version 5.0 [Computer software]. Chicago: Small Waters Corp. Barrett, L. (2004). Feelings or words? Understanding the content in selfreport ratings of experienced emotion. Journal of Personality and Social Psychology, 8, 266 –281. Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles among young adults: A test of a four-category model. Journal of Personality and Social Psychology, 61, 226 –244. Bartlett, M. Y., & DeSteno, D. (2006). Gratitude and prosocial behavior: Helping when it costs you. Psychological Science, 17, 319 –325. Baumeister, R. F., Bushman, B. J., & Campbell, W. K. (2000). Self-esteem, narcissism, and aggression: Does violence result from low self-esteem or from threatened egotism? Current Directions in Psychological Science, 9, 26 –29.

Baumeister, R. F., DeWall, C. N., Ciarocco, N. J., & Twenge, J. M. (2005). Social exclusion impairs self-regulation. Journal of Personality and Social Psychology, 88, 589 – 604. Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117, 497–529. Baumeister, R. F., Smart, L., & Boden, J. M. (1996). Relation of threatened egotism to violence and aggression: The dark side of high self-esteem. Psychological Review, 103, 5–33. Beer, J. S., Heerey, E. A., Keltner, D., Scabini, D., & Knight, R. T. (2003). The regulatory function of self-conscious emotion: Insights from patients with orbitofrontal damage. Journal of Personality and Social Psychology, 85, 594 – 604. Berkman, L. F., Vaccarino, V., & Seeman, T. (1993). Annals of Behavioral Medicine, 15, 112–118. Berridge, K. C. (2003). Comparing the emotional brains of humans and other animals. In R. J. Davidson, K. R. Scherer, & H. H. Goldsmith (Eds.), Handbook of affective sciences (pp. 25–51). New York: Oxford University Press. Berscheid, E., & Reis, H. T. (1998). Attraction and close relationships. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (4th ed., pp. 193–281). Boston: McGraw-Hill. Booth, R. J., & Pennebaker, J. W. (2000). Emotions and immunity. In M. Lewis & J. M. Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 558 –570). New York: Guilford Press. Bosson, J. K., Swann, W. B., & Pennebaker, J. W. (2000). Stalking the perfect measure of implicit self-esteem: The blind men and the elephant revisited? Journal of Personality and Social Psychology, 79, 631– 643. Bowlby, J. (1973). Separation: Anxiey and anger. Volume 2. Attachment and loss. New York: Basic Books Brennan, K. A., & Bosson, J. K. (1998). Attachment-style differences in attitudes toward and reactions to feedback from romantic partners: An exploration of the relational bases of self-esteem. Personality and Social Psychology Bulletin, 24, 699 –714. Bringle, R. G. (1991). Psychosocial aspects of jealousy: A transactional model. In P. Salovey (Ed.), The psychology of jealousy and envy (pp. 103–131). New York: Guilford Press. Bryson, J. B. (1991). Modes of response to jealousy-evoking situations. In P. Salovey (Ed), The psychology of jealousy and envy (pp. 178 –207). New York: Guilford Press. Buck, R. (1999). The biological affects: A typology. Psychological Review, 106, 301–336. Buunk, B. P. (1997). Personality, birth order, and attachment styles as related to various types of jealousy. Personality and Individual Differences, 23, 997–1006. Buunk, B. P., & Hupka, R. B. (1987). Cross-cultural differences in the elicitation of sexual jealousy. Journal of Sex Research, 23, 12–22. Buss, D. M., Larsen, R., Westen, D., & Semmelroth, J. (1992). Sex differences in jealousy: Evolution, physiology, and psychology. Psychological Science, 3, 251–255. Buunk, B. P. (1991). Jealousy in close relationships: An exchangetheoretical perspective. In P. Salovey (Ed.), The psychology of jealousy and envy (pp. 148 –177). New York: Guilford Press. Buunk, B. P., Angleitner, A., Oubaid, V., & Buss, D. M. (1996). Sex differences in jealousy in evolutionary and cultural perspective: Tests from the Netherlands, Germany, and the United States. Psychological Science, 7, 359 –363. Cacioppo, J. T., Hawkley, L. C., Crawford, E., Ernst, J. M., Burleson, M. H., Kowalewski, R. B., et al. (2000). Loneliness and health: Potential mechanisms. Psychosomatic Medicine, 64, 407– 417. Collins, N. L., & Read, S. J. (1990). Adult attachment, working models, and relationship quality in dating couples. Journal of Personality and Social Psychology, 58, 644 – 663.

JEALOUSY AND THE THREATENED SELF Cooley, C. H. (1902/1956). Human nature and the social order. New York: Schocken. Cosmides, L., & Tooby, J. (1992). Cognitive adaptations for social exchange. In J. Barkow, L. Cosmides, & J. Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press. Crocker, J., & Wolfe, C. T. (2001). Contingencies of self-worth. Psychological Review, 108, 593– 623. Damasio, A. (1994). Descarte’s errors. New York: Avon Books. Darwin, C. (1998). The expression of the emotions in man and animals (3rd ed.). New York: Oxford University Press. (Original work published 1872) Dasgupta, N., & Greenwald, A. G. (2001). On the malleability of automatic attitudes: Combating automatic prejudice with images of admired and disliked individuals. Journal of Personality and Social Psychology, 81, 800 – 814. Dasgupta, N., McGhee, D. E., Greenwald, A. G., & Banaji, M. R. (2000). Automatic preference for White Americans: Ruling out the familiarity effect. Journal of Experimental Social Psychology, 36, 316 –328. DeSteno, D. A. (2004, May). New perspectives on jealousy: An integrative view of the most social of social emotions. Paper presented at the meeting of the American Psychological Society, Chicago, IL. DeSteno, D. A., Bartlett, M. Y., Braverman, J., & Salovey, P. (2002). Sex differences in jealousy: Evolutionary mechanism or artifact of measurement? Journal of Personality and Social Psychology, 83, 1103–1116. DeSteno, D. A., Bartlett, M. Y., & Salovey, P. (in press). Constraining accommodative homunculi in evolutionary explorations of jealousy: A reply to Barrett et al. (2006). Journal of Personality and Social Psychology. DeSteno, D. A., Dasgupta, N., Bartlett, M. Y., & Cajdric, A. (2004). Prejudice from thin air: The effect of emotion on automatic intergroup attitudes. Psychological Science, 15, 319 –324. DeSteno, D. A., & Salovey, P. (1995). Jealousy and envy. In A. S. R. Manstead, M. Hewstone, S. T. Fiske, M. A. Hogg, H. T. Reis, & G. R. Semin (Eds.), The Blackwell encylopedia of social psychology (pp. 342–343). Oxford, MA: Blackwell. DeSteno, D. A., & Salovey, P. (1996a). Evolutionary origins of sex differences in jealousy? Questioning the “fitness” of the model. Psychological Science, 7, 367–372. DeSteno, D. A., & Salovey, P. (1996b). Jealousy and the characteristics of one’s rival: A self-evaluation maintenance perspective. Personality and Social Psychology Bulletin, 22, 920 –932. DeSteno, D. A., & Salovey, P. (1997). Structural dynamism in the concept of self: A flexible model for a malleable concept. Review of General Psychology, 1, 389 – 409. De Weerth, C., & Kalma, A. P. (1993). Female aggression as a response to sexual jealousy: A sex role reversal? Aggressive Behavior, 19, 265–279. Diener, E. (1984). Subjective well-being. Psychological Bulletin, 95, 542– 575. Downey, G., & Feldman, S. I. (1996). Implications of rejection sensitivity for intimate relationships. Journal of Personality and Social Psychology, 70, 1327–1343. Downey, G., Feldman, S. I., & Ayduk, O. (2000). Rejection sensitivity and male violence in romantic relationships. Personal Relationships, 7, 45– 61. Drigotas, S. M., Rusbult, C. E., Wieselquist, J., & Whitton, S. W. (1999). Close partner as sculptor of the ideal self: Behavioral affirmation and the Michelangelo phenomenon. Journal of Personality and Social Psychology, 77, 293–323. Dunn, J. (2003). Emotional development in early childhood: A social relationship perspective. In R. J. Davidson, K. R. Scherer, & H. H. Goldsmith (Eds.), Handbook of affective sciences (pp. 332–346). New York: Oxford University Press. Dutton, D. G., Saunders, K., Starzomski, A., & Bartholomew, K. (1994). Intimacy-anger and insecure attachment as precursors of abuse in inti-

639

mate relationships. Journal of Applied Social Psychology, 24, 1367– 1386. Eagly, A. H., & Steffen, V. J. (1986). Gender and aggressive behavior: A meta-analytic review of the social psychological literature. Psychological Bulletin, 100, 309 –330. Eisenberger, N. I., Lieberman, M. D., & Williams, K. D. (2003). Does rejection hurt? An fMRI study of social exclusion. Science, 302, 290 –292. Ellsworth, P. C., & Scherer, K. R. (2003). Appraisal processes in emotion. In R. J. Davidson, K. R. Scherer, & H. H. Goldsmith (Eds.), Handbook of affective sciences (pp. 572–595). New York: Oxford University Press. Evers, C., Fischer, A. H., Mosquera, P. M. R., & Manstead, A. S. R. (2005). Anger and social appraisal: A “spicy” sex difference? Emotion, 5, 258 –266. Finkel, E., J., Rusbult, C. E., Kumashiro, M., & Hannon, P. A. (2002). Dealing with betrayal in close relationships: Does commitment promote forgiveness? Journal of Personality and Social Psychology, 82, 956 –974. Fiske, A. P., Kitayama, S., Markus, H. R., & Nisbett, R. E. (1998). The cultural matrix of social psychology. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (4th ed., pp. 915– 981). Boston: McGraw-Hill. Fiske, S. T. (2004). Social beings: A core motives approach to social psychology. Hoboken, NJ: Wiley. Flavell, J. H. (2004). Theory-of-mind development: Retrospect and prospect. Merrill-Palmer Quarterly, 50, 274 –290. Fraley, R. C. (2002). Attachment stability from infancy to adulthood: Meta-analysis and dynamic modeling of developmental mechanisms. Personality and Social Psychology Review, 6, 123–151. Freud, S. (1955). Some neurotic mechanisms in jealousy, paranoia, and homosexuality. In J. Strachey (Ed. and Trans.), The standard edition of the complete works of Sigmund Freud (Vol. 18, pp. 223–232). London: Hogarth Press. (Original work published 1922) Frijda, N. H. (1986). The emotions. Cambridge, England: Cambridge University Press. Frijda, N. H. (2000). The psychologists’ point of view. In M. Lewis & J. M. Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 59 –74). New York: Guilford Press. Frith, C. D., & Frith, U. (1999). Interacting minds: A biological basis. Science, 286, 1692–1695. Frith, U., & Frith, C. (2001). The biological basis of social interaction. Current Directions in Psychological Science, 10, 151–155. Gallagher, H. L., & Frith, C. D. (2003). Functional imaging of ‘theory of mind.’ Trends in Cognitive Sciences, 7, 77– 83. Geary, D. C., Rumsey, M., Bow-Thomas, C. C., & Hoard, M. K. (1995). Sexual jealousy as a facultative trait: Evidence from the pattern of sex differences in adults from China and the United States. Ethology & Sociobiology, 16, 355–383. Gemar, M. C., Segal, Z. V., Sagrati, S., & Kennedy, S. J. (2001). Moodinduced changes on the Implicit Association Test in recovered depressed patients. Journal of Abnormal Psychology, 110, 282–289. Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J., & Wheatley, T. P. (1998). Immune neglect: A source of durability bias in affective forecasting. Journal of Personality and Social Psychology, 75, 617– 638. Glanz, K., & Lerman, C. (1992). Psychosocial impact of breast cancer: A critical review. Annals of Behavioral Medicine, 14, 204 –212. Goffman, E. (1959). The presentation of self in everyday life. Oxford, England: Doubleday. Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102, 4 –27. Greenwald, A. G., & Farnham, S. D. (2000). Using the Implicit Association Test to measure self-esteem and self-concept. Journal of Personality and Social Psychology, 79, 1022–1038. Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring

640

DESTENO, VALDESOLO, AND BARTLETT

individual differences in implicit cognition: The Implicit Association Task. Journal of Personality and Social Psychology, 74, 1464 –1480. Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and using the Implicit Association Test: I. An improved scoring algorithm. Journal of Personality and Social Psychology, 85, 197–216. Guerrero, L. K. (1998). Attachment-style differences in the experience and expression of romantic jealousy. Personal Relationships, 5, 273–291. Harris, C. R. (2003). A review of sex differences in sexual jealousy, including self-report data, psychophysiological responses, interpersonal violence, and morbid jealousy. Personality and Social Psychology Review, 7, 102–128. Harris, C. R., & Christenfeld, N. (1996). Gender, jealousy, and reason. Psychological Science, 7, 364 –366. Hazan, C., & Shaver, P. (1987). Romantic love conceptualized as an attachment process. Journal of Personality and Social Psychology, 52, 511–524. Heatherton, T. F., & Polivy, J. (1991). Development and validation of a scale for measuring state self-esteem. Journal of Personality and Social Psychology, 60, 895–910. Heerey, E. A., Keltner, D., & Capps, L. M. (2003). Making sense of self-conscious emotion: Linking theory of mind and emotion in children with autism. Emotion, 3, 394 – 400. Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94, 319 –340. Hupka, R. B. (1984). Jealousy: Compound emotion or label for a particular situation? Motivation & Emotion, 8, 141–155. Hupka, R. B. (1991). The motive for the arousal of romantic jealousy: Its cultural origin. In P. Salovey (Ed.), The psychology of jealousy and envy (pp. 252–270). New York: Guilford Press. Hupka, R. B., Buunk, B., Falus, G., Fulgosi, A., Ortega, E., Swain, R., & Tarabrina, N. V. (1985). Romantic jealousy and romantic envy: A seven-nation study. Journal of Cross-Cultural Psychology, 16, 423– 446. Hupka, R. B., & Ryan, J. M. (1990). The cultural contribution to jealousy: Cross-cultural aggression in sexual jealousy situations. Behavior Science Research, 24, 51–71. Jarvis, W. B. G. (2004). MediaLab 2004 [Computer software]. New York: Empirisoft. Jordan, C. H., Spencer, S. J., Zanna, M. P., Hoshino-Browne, E., & Correll, J. (2003). Secure and defensive high self-esteem. Journal of Personality and Social Psychology, 85, 969 –978. Karpinski, A. (2004). Measuring self-esteem using the Implicit Association Test: The role of the other. Personality and Social Psychology Bulletin, 30, 22–34. Kelley, W. M., Macrae, C. N., Wyland, C. L., Caglar, S., Inati, S., & Heatherton, T. F. (2002). Finding the self?: An event-related fMRI study. Journal of Cognitive Neuroscience, 14, 785–794. Keltner, D., & Busswell, B. N. (1997). Embarrassment: Its distinct form and appeasement functions. Psychological Bulletin, 122, 250 –270. Keltner, D., & Gross, J. J. (1999). Functional accounts of emotion. Cognition and Emotion, 13, 467– 480. Keltner, D., & Haidt, J. (1999). Social functions of emotion at four levels of analysis. Cognition & Emotion, 13, 505–521. Kennedy, S., Kiecolt-Glaser, J. K., & Glaser, R. (1990). Immunological consequences of acute and chronic stressors: Mediating role of interpersonal relationships. British Journal of Medical Psychology, 6, 77– 85. Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social psychology. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (4th ed., pp. 233–265). Boston: McGrawHill. Kiecolt-Glaser, J. K. (1999). Stress, personal relationships, and immune function: Health implications. Brain, Behavior & Immunity, 13, 61–72. Koole, S. L., Dijksterhuis, A., & Knippenberg, A. (2001). What’s in a name: Implicit self-esteem and the automatic self. Journal of Personality and Social Psychology, 81, 669 – 685.

Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford University Press. Leary, M. R. (2003). The self and emotion: The role of self-reflection in the generation and regulation of affective experience. In R. J. Davidson, K. R. Scherer, & H. H. Goldsmith (Eds.), Handbook of affective sciences (pp. 773–786). New York: Oxford University Press. Leary, M. R., & Baumeister, R. F. (2000). The nature and function of self-esteem: Sociometer theory. In M. Zanna (Ed.), Advances in experimental social psychology (pp. 1– 62). San Diego, CA: Academic Press. Leary, M. R., Koch, E. J., & Hechenbleikner, N. R. (2001). Emotional responses to interpersonal rejection. In M. R. Leary (Ed.), Interpersonal rejection (pp. 145–166). London: Oxford University Press. Leary, M. R., Tambor, E. S., Terdal, S. K., & Downs, D. L. (1995). Self-esteem as an interpersonal monitor: The sociometer hypothesis. Journal of Personality and Social Psychology, 68, 518 –530. LeDoux, J. E., & Phelps, E. A. (2000). Emotional newtworks in the brain. In M. Lewis & J. M. Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 157–172). New York: Guilford Press. Lewis, M. (2000). Self-conscious emotions: Embarrassment, pride, shame, and guilt. In M. Lewis & J. M. Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 623– 636). New York: Guilford Press. Lieberman, J. D., Solomon, S., Greenberg, J., & McGregor, H. A. (1999). A hot new way to measure aggression: Hot sauce allocation. Aggressive Behavior, 25, 331348. Lowery, B. S., Hardin, C. D., & Sinclair, S. (2001). Social influence effects on automatic racial prejudice. Journal of Personality and Social Psychology, 81, 842– 855. Macrae, C. N., Moran, J. M., Heatherton, T. F., Banfield, J. F. & Kelley, W. M. (2004). Medial prefrontal activity predicts memory for self. Cerebral Cortex, 14, 647– 654. Masciuch, S., & Kienapple, K. (1993). The emergence of jealousy in children 4 months to 7 years of age. Journal of Social & Personal Relationships, 10, 421– 435. Mathes, E. W., Roster, P. M., & Joerger, S. M. (1982). A convergent validity study of six jealousy scales. Psychological Reports, 50, 1143– 1147. McGregor, I., & Marigold, D. C. (2003). Defensive zeal and the uncertain self: What makes you so sure? Journal of Personality and Social Psychology, 85, 838 – 852. McGregor, H. A., Lieberman, J. D., Greenberg, J., Solomon, S., Arndt, J., Simon, L., & Pyszczynski, T. (1998). Terror management and aggression: Evidence that mortality salience motivates aggression against worldview-threatening others. Journal of Personality and Social Psychology, 74, 590 – 605. Mierke, J., & Klauer, K. C. (2003). Method-Specific variance in the Implicit Association Test. Journal of Personality and Social Psychology, 85, 1180 –1192. Mullen, P. E. (1993). The crime of passion and the changing cultural construction of jealousy. Criminal Behaviour & Mental Health, 3, 1–11. Mullen, P. E. (1996). Jealousy and the emergence of violent and intimidating behaviours. Criminal Behaviour & Mental Health, 6, 199 –205. Murray, S. L., Griffin, D. W., Rose, P., & Bellavia, G. M. (2003). Calibrating the sociometer: The relational contingencies of self-esteem. Journal of Personality and Social Psychology, 85, 63– 84. Myers, D., & Diener, E. (1995). Who is happy? Psychological Science, 6, 10 –19. Nisbett, R. E., & Cohen, D. (1996). Culture of honor: The psychology of violence in the south. Boulder, CO: Westview Press. ¨ hman, A. (2002). Automaticity and the amygdala: Nonconscious reO sponses to emotional faces. Current Directions in Psychological Science, 11, 62– 66. ¨ hman, A., & Wiens, S. (2003). On the automaticity of autonomic reO sponses in emotion: An evolutionary perspective. In R. J. Davidson,

JEALOUSY AND THE THREATENED SELF K. R. Scherer, & H. H. Goldsmith (Eds.), Handbook of affective sciences (pp. 256 –275). New York: Oxford University Press. Parker, J. G., Low, C. M., Walker, A. R., & Gamm, B. K. (2005). Friendship jealousy in young adolescents: Individual differences and links to sex, self-esteem, aggression, and social adjustment. Developmental Psychology, 41, 235–250. Parrott, W. G. (1991). The emotional experience of envy and jealousy. In P. Salovey (Ed.), The psychology of jealousy and envy (pp. 3–30). New York: Guilford Press. Parrott, W. G., & Smith, R. H. (1993). Distinguishing the experiences of envy and jealousy. Journal of Personality and Social Psychology, 64, 906 –920. Paul, L., Foss, M. A., & Galloway, J. (1993). Sexual jealousy in young women and men: Aggressive responsiveness to partner and rival. Aggressive Behavior, 19, 401– 420. Pietromonaco, P., & Barrett, L. F. (1997). Working models of attachment and daily social interactions. Journal of Personality and Social Psychology, 73, 1409 –1423. Rivers, W. H. R. (1906). The Todas. London: Macmillan. Sabini, J., & Green, M. C. (2004). Emotional responses to sexual and emotional infidelity: Constants and differences across genders, samples, and methods. Personality and Social Psychology Bulletin, 30, 1375–1388. Salovey, P. (1991). The psychology of jealousy and envy. New York: Guilford Press. Salovey, P., & Rodin, J. (1984). Some antecedents and consequences of social-comparison jealousy. Journal of Personality and Social Psychology, 47, 780 –792. Saxe, R., Carey, S., & Kanwisher, N. (2004). Understanding other minds: Linking developmental psychology and functional neuroimaging. Annual Review of Psychology, 55, 87–124. Schackelford, T. K. (2001). Self-esteem in marriage. Personality and Individual Differences, 30, 371–390. Schwarz, N., & Clore, G. L. (1996). Feelings and phenomenal experiences. In E. T. Higgins & A. W. Kruglanski (Eds), Social psychology: Handbook of basic principles (pp. 433– 465). New York: Guilford Press. Sharpsteen, D. J. (1991). The organization of jealousy knowledge: Romantic jealousy as a blended emotion. In P. Salovey (Ed.), The psychology of jealousy and envy (pp. 31–51). New York: Guilford Press. Sharpsteen, D. J., & Kirkpatrick. L. A. (1997). Romantic jealousy and adult romantic attachment. Journal of Personality and Social Psychology, 72, 627– 640. Smith, R. H. (1991). Envy and the sense of injustice. In P. Salovey (Ed.), The psychology of jealousy and envy (pp. 79 –99). New York: Guilford Press. Tangney, J. P., & Fischer, K. W. (1995). Self-conscious emotions: The

641

psychology of shame, guilt, embarrassment, and pride. New York: Guilford Press. Tesser, A. (1988). Toward a self-evaluation maintenance model of social behavior. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 21, pp. 181–227). New York: Academic Press. Tracy, J. L., & Robins, R. W. (2004). Putting the self into self-conscious emotions: A theoretical model. Psychological Inquiry, 15, 103–125. Twenge, J. M., Baumeister, R. F., Tice, D. M., & Stucke, T. S. (2001). If you can’t join them, beat them: Effects of social exclusion on aggressive behavior. Journal of Personality and Social Psychology, 81, 1058 –1069. U.S. Department of Justice (2003). Sourcebook of criminal justice statistics. Washington, DC: U.S. Government Printing Office. Vandello, J. A., & Cohen, D. (2003). Male honor and female fidelity: Implicit cultural scripts that perpetuate domestic violence. Journal of Personality and Social Psychology, 84, 997–1010. Vecchio, R. P. (2000). Negative emotion in the workplace: Employee jealousy and envy. International Journal of Stress Management, 7, 161–179. Volling, B. L., McElwain, N. L., & Miller, A. L. (2002). Emotion regulation in context: The jealousy complex between young siblings and its relations with child and family characteristics. Child Development, 73, 581– 600. White, G. L. (1991). Self, relationship, friends, and family: Some applications of systems theory to romantic jealousy. In P. Salovey (Ed.), The psychology of jealousy and envy (pp. 231–251). New York: Guilford Press. Wiederman, M. W., Allgeier, E. R. (1993). Gender differences in sexual jealousy: Adaptionist or social learning explanation? Ethology & Sociobiology, 14, 115–140. Wiederman, M. W., Allgeier, E. R., & Raguas, D. M. (1995). Empirical investigation of the use of the term “jealousy” in survey research. Representative Research in Social Psychology, 20, 15–29. Wieselquist, J., Rusbult, C. E., Foster, C. A., & Agnew, C. R. (1999). Commitment, pro-relationship behavior, and trust in close relationships. Journal of Personality and Social Psychology, 77, 942–966. Williams, K. D. (1997). Social ostracism. In R. M. Kowalski (Ed.), Aversive interpersonal behaviors (pp. 133–170). New York: Plenum Press. Williams, K. D., Cheung, C. K. T., & Choi, W. (2000). Cyberostracism: Effects of being ignored over the Internet. Journal of Personality and Social Psychology, 79, 748 –762. Wilson, T. D., Wheatley, T., Meyers, J. M., Gilbert, D. T., & Axson, D. (2000). Focalism: A source of durability bias in affective forecasting. Journal of Personality and Social Psychology, 78, 821– 836. Wittenbrink, B., Judd, C. M., & Park, B. (2001). Spontaneous prejudice in context: Variability in automatically activated attitudes. Journal of Personality and Social Psychology, 81, 815– 827.

Appendix Stimuli for Implicit Self-Esteem Task Pleasant Words glory, gold, health, joy, kindness, lucky, peace, sunrise, truth, warmth

Unpleasant Words

Other-Relevant Items Pat, Carter, February 28, 1968, CLEVELAND, IDAHO, 92473, CANADA, ROMANIAN, 978-25-8826

agony, corpse, death, filth, killer, poison, slum, stink, torture, vomit

Self-Information Probes first name, last name, birthday, birth year, hometown, home state, zip code, home country, ethnicity, student ID number

Received March 13, 2005 Revision received Nov. 8, 2005 Accepted Nov. 19, 2005 䡲

INTERPERSONAL RELATIONS AND GROUP PROCESSES

Romantic Involvement Often Reduces Men’s Testosterone Levels—But Not Always: The Moderating Role of Extrapair Sexual Interest Matthew McIntyre

Steven W. Gangestad

Harvard University

University of New Mexico

Peter B. Gray

Judith Flynn Chapman

University of Nevada, Las Vegas

Harvard University

Terence C. Burnham

Mary T. O’Rourke

Harvard Business School

Harvard University

Randy Thornhill University of New Mexico Testosterone (T) appears to facilitate what biologists refer to as mating effort—the investment of time and energy into same-sex competition and mate-seeking behavior. Multiple studies show that men who are romantically involved (i.e., are paired) have lower T than single men, which may be due to a facultative adjustment by men of T levels in response to lower demands for mating effort. The authors proceeded on the basis of the idea that men who retain interests in sexual opportunities with women other than a primary partner continue to dedicate more time and energy to mating effort when romantically paired, and so they predicted that the association between relationship status and T depends on men’s extrapair sexual interests. Study 1 used the Sociosexual Orientation Inventory to measure extrapair sexual interests, whereas Study 2 used a broader measure to examine this interaction. Both studies found support for it. These results have implications for an understanding of the biosocial regulation of men’s behavior in romantic relationships. Keywords: testosterone, relationship status, sociosexuality, sexual selection

(Basaria et al., 2002; Bhasin, 2003; Schroeder et al., 2003). Male upper body musculature is thought to have facilitated success in competition between men among human ancestors and, hence, the sexual dimorphism in muscle mass is thought to be partly the result of sexual selection on men’s abilities to compete for mates (through direct competition with each other and display of intrasexual competitive abilities to women; Ellison, 2001, pp. 273–274; Gaulin & Sailer, 1984; Martin, 1980).1 Furthermore, T supports psychological and behavioral outcomes that appear to encourage success in male–male competition and sexual behavior. A recent review of the literature on the association between T and sexual motivation suggests a threshold effect of T on libido (Bancroft, 2002). T levels below a certain level appear to be positively related to libido. Above this level, however, the mar-

The steroid hormone testosterone (T) plays a role in facilitating what biologists refer to as mating effort—the investment of time and energy into same-sex competition and mate-seeking behavior (Ellison, 2001, pp. 274 –280). In both men and women, T promotes the maintenance and growth of skeletal muscle, particularly the sexually dimorphic mass in the chest, upper arms, and shoulders

Matthew McIntyre, Department of Epidemiology, Harvard School of Public Health, Harvard University; Steven W. Gangestad, Department of Psychology, University of New Mexico; Peter B. Gray, Department of Anthropology and Ethnic Studies, University of Nevada, Las Vegas; Judith Flynn Chapman and Mary T. O’Rourke, Department of Anthropology, Harvard University; Terence C. Burnham, Harvard Business School; Randy Thornhill, Department of Biology, University of New Mexico. Correspondence concerning this article should be addressed to Matthew McIntyre, Department of Epidemiology, Channing Laboratory, Harvard University, 181 Longwood Avenue, 3rd Floor, Boston, MA 02115. E-mail: [email protected]

1 A competing theory states that the sexual dimorphism in muscular strength in humans has been maintained by a sexual division of labor rather than sexual selection. See Ellison (2001, pp. 272–273) for a discussion.

Journal of Personality and Social Psychology, 2006, Vol. 91, No. 4, 642– 651 Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.642

642

MATING, TESTOSTERONE, AND EXTRAPAIR INTEREST

ginal increase in libido associated with T is minimal. In another review of both human and rhesus macaque findings, Wallen (2001) discussed the difficulty of identifying changes in sexual motivation by using behavioral measures. Extrinsic social conditions constrain the ability of an individual to act on sexual feelings and may do so differently for different individuals. Relationships between sexual behaviors and T are stronger when social conditions are experimentally controlled. Even when not involved directly in increasing libido, T may function to facilitate male pursuit of female interest (i.e., mating effort). Wild male chimpanzees have increased T when parous females exhibit maximally tumescent sexual swellings but not in response to maximal swelling in nulliparous females, whom males do not compete as intensively over. As males copulate with nulliparous and parous females at approximately the same rates, these findings suggest that these changes in T facilitate male pursuit of females rather than sexual performance per se (Muller & Wrangham, 2004). A recent study found that men’s T increased after they interacted with an attractive woman and particularly so when the woman thought they were trying to impress her (Roney, Mahler, & Maestripieri, 2003). In related findings, T also appears to be associated with aspects of social assertiveness (Ellison, 2001, p. 265) or dominance seeking (Mazur & Booth, 1998). In chimpanzees, males become more aggressive when parous, but not nulliparous, females exhibit sexual swellings, which parallels changes in T (Muller & Wrangham, 2004). Associations between T and overt intrasexual aggressiveness in humans are weak, though meta-analysis reveals that they are reliable (Archer, Birring, & Wu, 1998; Book, Starzyk, & Quinsey, 2001). Social conditions constrain aggressive behavior just as they constrain sexual behavior. Dominance seeking may or may not entail aggressive behavior depending on social conditions and other moderating factors. For example, once an informal dominance hierarchy in a primate group is established, aggressive behavior may be minimal (and instead dominance relations may be established through stare downs). Furthermore, one should not expect physical aggressivity between two individuals who are able to accurately assess each other’s abilities to win a contest (de Waal, 1986; de Waal & Hoekstra, 1980). (In the human case, legal/moral strictures [i.e., the threat of institutional retaliation] are apropos.) Dominance seeking may be expressed, among other behaviors, in greater selective attention to angry faces (van Honk et al., 2000; van Honk et al., 1999), in less pronounced smiling (Dabbs, 1997), or in more visual attention toward interaction partners (Dabbs, Bernieri, Strong, Campo, & Milun, 2001).

T and Romantic Bonding If T is associated with mating effort in men, then one might expect men who are in committed romantic relationships to have lower levels of T. In fact, much evidence supports this prediction. Married men have lower T than unmarried men (Booth & Dabbs, 1993; Gray, Campbell, Marlowe, Lipson, & Ellison, 2004; Gray, Kahlenberg, Barrett, Lipson, & Ellison, 2002; Mazur & Michalek, 1998). Unmarried men in committed, romantic relationships have lower T than unpaired men (Burnham et al., 2003). In an important longitudinal study, Mazur and Michalek (1998) found that divorced men’s T dropped if they remarried. The T of men who were married and then divorced was particularly high right around the

643

period of the divorce. Together, these findings suggest that T increases when men (presumably) are searching for romantic partners and decreases when men (presumably) are not searching for romantic partners. In an evolutionary framework, responsiveness of T production to mating effort implies that there is a cost to men of maintaining high T when they are mated. In fact, T has such costs. It channels energy that could be used otherwise into, for instance, maintaining muscle mass. Perhaps for this reason, T may impair immune functioning (Campbell, Lukas, & Campbell, 2001; Klein, 2000; but see Granger, Booth, & Johnson, 2000). T may also encourage risk-taking behavior that, although potentially beneficial when men are competing for mates, entails costs not worth undertaking when men are in committed relationships. Finally, T may interfere with affiliative and nurturing behaviors that are important in species characterized by long-term social bonds between relationship partners and parental care by both sexes. In one such species, marmosets, T is reduced when males become fathers (Nunes, Fite, & French, 2000; Nunes, Fite, Patera, & French, 2001). Three studies have shown that T is similarly low in men who have recently become fathers (Berg & Wynne-Edwards, 2001; Fleming, Corter, Stallings, & Steiner, 2002; Storey, Walsh, Quinton, & Wynne-Edwards, 2000). Other examples of behaviors shown to affect T production in men may also be indirectly related to mating effort. The most important paradigm revealing responsive fluctuation in T has been the effects, in men, of winning or losing competitions. In anticipation of competition, T level rises. However, after the competition, T levels fall in losers but remain high in winners. Following the original findings in tennis matches and among recent recipients of the MD degree (Mazur & Lamb, 1980), and then among wrestlers (Elias, 1981), a similar difference has been reported among chess players (Mazur, Booth, & Dabbs, 1992), video game players (Mazur, Susman, & Edelbrock, 1997), and in fans (Bernhardt, Dabbs, Fielden, & Lutter, 1998). (A study of competitors in judo matches failed to replicate the effect; Suay et al., 1999.) T has also been found to increase following sexual intercourse (Dabbs & Mohammed, 1992). As noted above, T increases after men interact with an attractive woman (Roney et al., 2003). Competition, sexual intercourse, and association with women could all be seen as important signals of opportunity for, or devotion to, finding new romantic partners as opposed to nonsexual pursuits, parenting, or maintaining existing romantic relationships that have important nonsexual value. Although it remains unclear whether short-term fluctuations in T that are affected by social experiences could also have shortterm influences on psychological states or behavior, it seems likely that social experiences might serve as a cue about conditions for mate seeking over a longer term, during which T could affect both relevant physical states, such as metabolism and physical strength, and also psychological states, such as mood, libido, and competitiveness. In addition to the evidence for the effects of T already discussed, clinical trials of T administration have shown clear effects on mood and aggressiveness (Pope, Kouri, & Hudson, 2000). Whereas sexual and romantic relationships in any particular contemporary society involve particular norms, physical and psychological states associated with power, elevated mood, and elevated libido (to the extent that they can be generalized across cultures and species) are closely associated with male

644

MCINTYRE ET AL.

mating activities in mammals as a whole and in humans taken as a species.

The Potential Moderating Role of Extrapair Sexual Interests Although longitudinal data indicate that T levels change as a result of changing relationship status (Mazur & Michalek, 1998), from a conceptual standpoint, it is not clear that all men’s T levels should change in the same way. Whereas the levels of many men may drop when they become involved in a committed relationship, the levels of others may remain relatively high. The current studies investigated potential moderators of the effects of relationship status on men’s T. Paired men may be faithful or unfaithful. Faithful men are relatively committed to the sexually exclusive nature of their relationships. Although they may find women other than their partners attractive, they do not engage in effort to attract these women—that is, they do not engage in extrapair mating effort. Unfaithful men remain relatively interested in and open to opportunities to engage in sex outside of a relationship. They hence do engage in extrapair mating effort, at least when opportunities arise. If men’s changing interests in seeking and attracting new mates account for the association between their relationship status and T, men’s faithfulness may moderate this effect. The T levels of men who tend not to be interested in seeking new mates when in a committed relationship should hence change as a function of their relationship status. By contrast, the T levels of men relatively open to pursuing short-term sexual relationships with women other than a primary long-term partner may remain relatively high and change less markedly as a function of their relationship status. Hence, their mating effort may remain relatively high even when in a committed relationship. It follows, from this reasoning, that men’s extrapair sexual interest should statistically interact with their relationship status to predict T.

Study 1: Sociosexual Orientation, Relationship Status, and T Men’s degree of interest in pursuing or being open to sex outside of a relationship may vary along a dimension. Sociosexual orientation refers to individual differences in the willingness to engage in sex outside of a committed, emotionally involved romantic relationship (Simpson & Gangestad, 1991). Individuals with a restricted sociosexual orientation claim to be uncomfortable having sex outside of a committed relationship in which they are emotionally involved and have little history of doing so. When in a committed relationship, they tend not to fantasize about sex with individuals other than their partners. By contrast, individuals with an unrestricted sociosexual orientation claim to be relatively comfortable engaging in “casual,” uncommitted sex and tend to have a history of doing so. When in committed relationships, they tend to fantasize about sex with individuals other than their partners. Sociosexual orientation can be measured with the Sociosexual Orientation Inventory (SOI), a short, seven-item questionnaire assessing past sexual history, sexual fantasies, and attitudes toward uncommitted, casual sex (Simpson & Gangestad, 1991). High scores reflect relatively unrestricted sociosexual orientation; low scores reflect a relatively restricted one. Research has provided evidence for convergent validation of the SOI and discriminant

validity evidence that sociosexual orientation does not merely reflect generalized sex drive. In one study, the SOI did not predict the frequency of sex occurring within committed relationships, though it did predict how soon after the beginning of a relationship sex occurred (Simpson & Gangestad, 1991). Conceptually, sociosexual orientation should relate to men’s willingness to engage in sex outside of a relationship. In Study 1, then, we tested the hypothesis that male SOI moderates the association between relationship status and T.

Method Participants. Participants were undergraduate students at Harvard University between the ages of 17 and 26 years (M ⫽ 20 years, SD ⫽ 1.51). Participants were recruited from an introductory lecture course in psychology and anthropology that satisfies general matriculation requirements (n ⫽ 67) and at a table outside the dining room of an undergraduate residence hall (n ⫽ 40). Of 107 participants, 5 were excluded from the analyses presented in this article because they completed fewer than 4 of 5 components of the SOI questionnaire. Procedure. Lecture course students who chose to participate provided saliva samples and completed questionnaires at the end of required discussion sessions (in groups of 12–17), at various days and times, over the course of 1 week. Residence hall students were asked to provide saliva samples and complete the questionnaire either (a) at the table or (b) if already late in the day, on waking the next day. The latter participants returned their materials to researchers the following day. Consent was obtained verbally, and all materials were confidential and anonymous. The questionnaires included the following: 1.

A brief introductory questionnaire containing items requesting demographic information (e.g., age), the time of day at which the participant awoke, the time at which the saliva sample was collected, relationship status (specifically, whether the participant was in a committed, romantic relationship), the duration of his relationship (if applicable), and the percentage of time the participant had been in a relationship during the past 3 years.

2.

SOI (Simpson & Gangestad, 1991), an eight-item questionnaire asking (a) the number of partners in the past year, (b) the number of partners estimated for the next 5 years, (c) the number of partners with whom the respondent has had sex once and only once, (d) the frequency of fantasy about sex with someone other than a current romantic partner, and (e) three items, forming one component, concerning attitudes toward uncommitted, casual sex (e.g., “I can imagine myself being comfortable and enjoying “casual” sex with different partners”). These five components were first z scored and then averaged to produce total scores (␣ ⫽ .81).

Before completing the questionnaire, participants expectorated at least 1 ml of saliva into a test tube treated with the preservative sodium azide. Trident Sugar-Free Gum, which negligibly interferes with the T assay, was used to stimulate saliva production. After completion of the session, participants were fully debriefed, given the opportunity to ask questions, and thanked for their participation. The study was approved by the Harvard University review board for human subjects research. Salivary T assays. Analysis of saliva samples followed published protocols (Granger, Schwartz, Booth, & Arentz, 1999). The time at which saliva samples were collected ranged between 0700 and 0110. Participants had been awake at the time of collection for between 0001 and 1630. T concentration declines over the course of the waking day (Nelson, 2000). Therefore, all analyses include time awake as a covariate. Saliva samples

MATING, TESTOSTERONE, AND EXTRAPAIR INTEREST

645

were assayed for T in the Reproductive Ecology Laboratory at Harvard University. The assay method used is a modified version of Granger et al. (1999) that was based on an application of the 125I double antibody kit produced by Diagnostic Systems Laboratories (Webster, TX). Sample and standard reactions were run in duplicate. Substrate (400 ␮l) was pipetted into borosilicate tubes, 200 ␮l of sample and 200 ␮l of buffered saline or, for the standard reactions, a 400 pg/ml standard concentration was added in volumes of 2, 5, 15, 50, 125, and 375 ␮l, with volumes of buffered saline adjusted to yield 400 ␮l total volume. Undiluted antiserum (20 ␮l) and tracer (50 ␮l) were added to sample and standard tubes. Reactions incubated overnight for at least 18 hr, after which precipitating reagent (500 ␮l) was added, and the tubes were centrifuged. After centrifugation, the supernatant was aspirated before the tubes were placed in a gamma counter. The assays were sensitive to 14 pmol/L T. Participants were allocated randomly into three lots. Interassay coefficients of variation were 12.0% for low pools and 15.3% for high pools. The intraassay coefficient of variation was 9.5%. T concentrations reported in this article are the natural-logged averages of duplicates. For these data, natural-logging T concentration (a) normalized the right-skewed concentration variable and (b) maximized the variance explained by time awake. Identification of relationship status groups. Differences in T levels between groups of individuals identified by dating status in this sample are reported elsewhere by Gray, Chapman, et al. (2004). As noted by Gray, Chapman, et al., the two groups of individuals who were not in a relationship could be distinguished on the basis of T. Of the 102 participants, 37 reported being in a romantic relationship. Of the 65 participants not currently in a relationship, 21 also reported not having been in a relationship in the past 3 years. The less experienced men were somewhat younger than their 44 more experienced counterparts not currently in a relationship (19.5 years old relative to 20.1 years old; Cohen’s d ⫽ 0.39, p ⫽ .13), had somewhat lower (though insignificantly so) SOI scores (Cohen’s d ⫽ 0.40, p ⫽ .12), and, most notably, had lower salivary T (Cohen’s d ⫽ 0.51, p ⬍ .05). In fact, men who reported no experience with being in a relationship had mean T levels slightly (though nonsignificantly) lower than men in a relationship. The low T levels of these men suggest that, despite their not currently being in a relationship, they were investing relatively little in overall mating effort. One possibility is that, because of their youth or developmental immaturity, they had not developed the same level of interest in attracting romantic partners as other men. A second possibility is that they were relatively unattractive as mates and hence invested less in mating effort. In any case, in light of their low T levels, our prediction that SOI moderates the effect of relationship status on T is not clearly tested with this group. Also, Roney et al. (2003) did not find behavioral effects on T among romantically less experienced men, supporting their treatment as a distinct group. We do not include this group in our main analyses, though we also report results with them included. In our main analyses to test our hypothesis, we contrasted the paired group with the single group that included only men with some experience (in the last 3 years) with relationships. We also tested the relationship between SOI and T in paired men only, among whom a positive relationship was expected. Treatment of SOI and relationship length. To be consistent with our treatment of T concentration, we examined the distribution of SOI scores. It too was strongly right skewed. Hence, we log-transformed it as well. We added two points to the original SOI composite before logging these values to ensure positive outputs. The resulting skew was near zero. SOI was zero centered prior to being entered into analyses to ensure orthogonality of main effects and its interaction with relationship status. In some analyses on paired men, we include the variable relationship length (in months). This variable was also right skewed and hence we log-transformed it as well.

(paired vs. single; excluding less experienced men) and three continuous predictors. Logged SOI score was entered as a continuous predictor of interest, along with its interaction with relationship status (except when only paired men were included). Time awake was also entered as a control for the diurnal pattern of T production and was, in our overall analysis, a significant predictor, F(1, 75) ⫽ 17.73, p ⬍ .001, but is not discussed further. Age was also entered as a continuous predictor because of its association with both T and relationship history. Predicted effects were assessed with directed tests that allocate a probability of .04 to a predicted “tail’s” region of rejection and .01 to a nonpredicted tail (Rice & Gaines, 1994). (This procedure enhances power to detect a predicted effect relative to two-tailed tests without the problem of excluding any possibility of a nonpredicted effect entailed by one-tailed tests.) Results revealed the predicted interaction between relationship status and SOI, F(1, 75) ⫽ 5.38, p ⬍ .02. The main effects for relationship status, F(1, 75) ⫽ 1.17, and SOI, F(1, 75) ⫽ 0.18, and age, F(1, 75) ⫽ 2.63, were not significant. Table 1 reports all effects. When less experienced men were included (with relationship experience included as a second factor), the interaction between relationship status and SOI score remained significant, F(1, 95) ⫽ 3.58, p ⬍ .04. Relationship experience also had a significant effect, with more experienced men having higher T, F(1, 95) ⫽ 6.39, p ⬍ .01, as did age, with older men having lower T, F(1, 95) ⫽ 3.43, p ⬍ .04. Again, relationship status, F(1, 95) ⫽ 1.29, and SOI, F(1, 95) ⫽ 1.22, did not have a significant effect. Are the observed interactions between relationship status and SOI in predicting T levels explained by an association between T and SOI among single men, paired men, or both? To answer this question, we performed similar GLM analyses on paired men only and on single men only. In paired men, SOI approached significant prediction of T levels, F(1, 33) ⫽ 2.68, p ⫽ .069. Higher SOI scores tended to be associated with higher T levels. In paired men, however, there exists one additional factor that can be controlled— relationship length, which has been discussed previously as a negative predictor of T, at least over the first several months of a relationship (Gray, Chapman, et al., 2004). In an analysis including relationship length as a predictor, both SOI, F(1, 32) ⫽ 3.69, p ⬍ .04, and relationship length, F(1, 32) ⫽ 3.43, p ⬍ .04, were indeed significant predictors (see Table 2). Consistent with our prediction, then, men’s SOI is positively associated with T in paired men with relationship length controlled.

Results

Note. N ⫽ 81. T and Sociosexual Orientation Inventory (SOI) are logged values. Relationship status: 0 ⫽ single, 1 ⫽ paired. For all variables, directed tests were used to obtain p values. * p ⬍ .05. ** p ⬍ .01.

Logged T levels were analyzed with SPSS 12.0 (GLM Univariate) in a two-level factor design based on relationship status

Table 1 Study 1 General Linear Model Analysis of Testosterone (T) Levels Variable

F(1, 75)

Partial r

Time awake Age Relationship status SOI SOI ⫻ Relationship Status

17.73** 2.63 1.58 0.18 5.38*

⫺.44 ⫺.18 ⫺.14 .04 .26

MCINTYRE ET AL.

646

Table 2 Study 1 General Linear Model Analysis of Testosterone (T) Levels in Paired and Single Men Paired men

Single men

Variable

F(1, 32)

Partial r

F(1, 40)

Partial r

Time awake Age Relationship length SOI

18.09** 0.59 3.43* 3.69*

⫺.60 ⫺.13 ⫺.31 .32

1.59 1.92

⫺.19 ⫺.21

2.35

⫺.24

Note. n ⫽ 37 and 44, respectively. T, relationship length, and Sociosexual Orientation Inventory (SOI) are logged values. For all variables, directed tests were used to obtain p values. * p ⬍ .05. ** p ⬍ .01.

In more experienced single men, SOI did not significantly predict T, F(1, 40) ⫽ 2.35, p ⫽ .133 (see Table 2). When less experienced men were included and experience added as an additional predictor, SOI similarly was not significantly associated with T, F(1, 60) ⫽ 0.50. In this analysis, we also included percentage of time men had spent in a relationship in the past 3 years as a predictor. It had negligible effect, F(1, 39) ⫽ 0.07, ns, and its inclusion did not alter other effects. Its lack of effect is consistent with the proposal that relationship status affects T and not consistent with the proposal that T affects men’s tendency to be in relationships.

Discussion On the basis of the idea that romantically paired men with unrestricted sociosexual orientation continue to dedicate more time and energy to mating effort, the allocation of which is partly modulated by T, whereas men with relatively restricted sociosexual orientation engage in substantially less mating effort when romantically paired, we predicted that the association between relationship status and T would depend on men’s sociosexual orientation. In a sample of undergraduates more experienced with being in relationships, this prediction was supported by two lines of evidence. First, we detected a significant interaction between relationship status and SOI. Second, we found that among paired men, but not among single men, SOI predicts T levels. A prediction following from an alternative conceptualization, that SOI would be associated with T independent of relationship status, was not supported. To replicate and extend this finding, we examined these associations in a second study. The second study differed from the first in one important respect and in two more minor respects. The first study included only the SOI as a measure of interest in mating effort when in a relationship. Conceptually, it makes sense that the SOI could effectively serve as such a measure, but other measures are also possible. The SOI was designed to assess individual differences in willingness to have sex outside of a committed relationship characterized by emotional closeness. In the present context, we are interested in individual differences in willingness and desire to have sex with a partner other than a primary partner with whom an individual is currently paired. We added two simple measures to tap these individual differences: a

question asking whether an individual could imagine themselves having sex outside of a relationship (an “affair”) and a question asking whether an individual actually had had sex with someone other than a primary partner while being involved in a committed relationship. To test the prediction that willingness to engage and to have interest in sex with someone other than a primary partner would moderate the effect of relationship status on T, we examined T levels as a function of these measures as well as the SOI. (In fact, as we later describe, we created a composite measure, as all three covaried in our sample.) In the first study, we controlled for relationship length. We did not ask about relationship length in Study 2. (Data were collected initially for other purposes, for which relationship length was not required.) We did, however, inquire about the nature of a relationship (e.g., whether individuals were married, engaged, living with their partner, or dating exclusively), and we controlled for these variations. Whereas participants of Study 1 were Harvard undergraduates, Study 2 was conducted at the University of New Mexico, an institution with a student body presumably more diverse along sociocultural and intellectual dimensions.

Study 2 Method Participants. Participants were 74 men who took part in a larger study on scents and attraction (see Thornhill & Gangestad, 1999). Individuals were recruited from Introductory Psychology or other classes and given either course credit for a research requirement or extra credit toward their grade for participating. All participants were asked their sexual orientation (heterosexual, homosexual, or bisexual). As our interest in this study was on heterosexual relationships, only the 69 men who reported to be heterosexual were retained for analysis. Of these men, the mean age was 20.3 years (range ⫽ 17–33, SD ⫽ 2.40). (One individual did not report his age and hence could not be included in analyses controlling for age.) Fifty-five percent reported themselves to be Caucasian, 27% Hispanic, 7% African American, 4% Native American, 3% Asian American, and 3% another ethnicity. Procedure. Participation involved two sessions. In a first session, after providing informed consent, participants were asked to fill out a series of questionnaires. These questionnaires included the following: 1.

A personal data sheet. This questionnaire asked individuals basic demographic information such as age, ethnic background, and sexual orientation. In addition, relationship status was assessed with a series of seven items. Each individual was asked to indicate with a check mark whether each of the following characteristics applied to him: married, married but separated, divorced, engaged to be married, not currently married but cohabiting with a partner, dating one person exclusively, dating multiple persons, and not currently dating. An individual was classified as paired (n ⫽ 26) if he was dating one person exclusively (n ⫽ 19), married (n ⫽ 4), cohabiting with a partner (n ⫽ 2), or engaged to be married (n ⫽ 1). He was classified as single (n ⫽ 43) if he was dating multiple persons (n ⫽ 15) or not dating (n ⫽ 29). (One individual claimed to be dating multiple persons and not dating.)

2.

SOI (Simpson & Gangestad, 1991). As in Study 1, the five components of this measure were z scored and averaged to create a composite (␣ ⫽ .75).

3.

Willingness to engage in extrapair sex. A single item asked

MATING, TESTOSTERONE, AND EXTRAPAIR INTEREST whether men would ever consider having extrapair sexual relations. Specifically, they were asked Would you ever consider having an “affair” (sex with a person other than a main, current relationship partner) behind the back of your relationship partner? (Here, consider not only your present partner [if you have one], but any partner you might have in the future.) Check one: A. No, I would never have sex outside of a relationship under any circumstances. B. I can imagine that I could possibly have sex outside of a relationship under certain circumstances. Of all men, 52% claimed that they would not; 48% said that they could imagine such circumstances. Within groups, 23% and 63% of paired and single men, respectively, claimed that they could, ␹2(1, N ⫽ 69) ⫽ 10.24, p ⫽ .001. 4.

History of extrapair sex. Men were asked whether they had ever engaged in sex with a partner other than a current partner while involved in a romantic relationship. Of all men, 29% claimed that they had, including 23% of paired men and 33% of single men, ␹2(1, N ⫽ 69) ⫽ 0.81, ns.

At the first session, men were provided a test tube, a stick of sugarless gum, and instructions for how to collect saliva. Each tube had a small amount of a preservative, sodium azide. Each individual brought a tube with approximately 1 ml of saliva to a second session. They were asked to collect the saliva on waking the morning of their session to control for diurnal variation in T. (For reasons unrelated to the current study but related to the study’s focus on scent and attraction, individuals were asked to refrain from eating pungent foods and from drinking alcohol, smoking, having sex, or sleeping with someone the two days and nights prior to collection of the saliva.) Following completion of the study, individuals were fully debriefed and given an opportunity to ask questions. Salivary T assay. T concentrations were estimated with a modification of a commercially available fluoroimmunoassay kit (DELFIA testosterone, Wallac, Turku, Finland; Wallac is now part of Perkin-Elmer, Boston, MA) in the Reproductive Ecology Laboratory at Harvard University. One milliliter of sample was extracted in diethyl ether, dried under nitrogen, and reconstituted in 100 ul of assay buffer; that is, the volume increased by a factor of 10. Duplicate 25 ul aliquots were assayed for each sample. High T and low T quality control pools were run with each assay (coefficient of variation 4.6% high pool, 18.1% low pool), and in all cases where there were two samples from an individual both samples were run in the same assay and averaged. Concentrations for the standard curve were from 0.5 to 50.0 nmol/L. Values for all samples were above the lower limit of the curve. As in Study 1, T concentrations were log-transformed.

647

averaged these z scores for each individual. (One individual did not report on his history of affairs and hence his score was the average of the other two components.) We interpret and refer to this composite as extrapair sexual interest. Group comparison on T: Paired versus single men. Before examining the moderating role of extrapair sexual interest, we examined whether T levels of paired and single men differed. Consistent with other work, T did vary as a function of relationship status, t(67) ⫽ 2.61, p ⬍ .01. Single men (mean logged T ⫽ 6.16, SD ⫽ 0.30) had higher levels of T than paired men (M ⫽ 5.94, SD ⫽ 0.39). T as a function of relationship status and extrapair sexual interest. T levels were analyzed with SPSS 12.0 (GLM Univariate) in a two group (relationship status: paired vs. single) design with two continuous predictors, extrapair sexual interest and age. Directed tests (Rice & Gaines, 1994) were used to assess the predicted effects of relationship status and the extrapair sexual interest by relationship status interaction. Results revealed that extrapair sexual interest significantly interacted with relationship status, F(1, 63) ⫽ 4.32, p ⬍ .03 (see Table 3). The effect of relationship status fell short of statistical significance, F(1, 63) ⫽ 2.55, p ⫽ .072, as did the effect of age, F(1, 63) ⫽ 1.80, ns. To interpret the interaction effect, we ran separate analyses on paired and single men. In paired men, the effect of extrapair sexual interest approached significance, F(1, 23) ⫽ 2.44, p ⫽ .083, as in Study 1. In Study 1, we also controlled for relationship length. In this study, we did not measure relationship length. We did, however, ask men whether they were married or engaged, living with their partner, or simply dating one person. These categories undoubtedly reflect differences in level of commitment to a relationship as well as, to some extent, relationship length. To control for overall level of commitment (or type) of relationships, we categorized men into three groups (married/ engaged, living with, and dating only) and included this variable in the analysis. As predicted, extrapair sexual interest significantly predicted T, F(1, 21) ⫽ 5.49, p ⬍ .02 (see Table 4). This result conceptually replicates our finding for paired men in Study 1. The effects of age, F(1, 21) ⫽ 2.42, p ⫽ .084, and relationship type, F(1, 21) ⫽ 2.21, p ⫽ .13, fell short of statistical significance. In single men, neither extrapair sexual interest, F(1, 39) ⫽ 1.30, ns, nor age, F(1, 39) ⫽ 0.80, ns, predicted T (see Table 4). In a follow-up analysis, we also included whether men said they were dating multiple women (vs. not currently dating someone). The effects of extrapair sexual interest, F(1, 38) ⫽ 0.71, and age, F(1,

Results Associations between sociosexual orientation, willingness to have extrapair sex, and history of extrapair sex. We first examined the correlations between our three measures tapping individual differences in interest in extrapair or casual sex. Not surprisingly, all three measures correlated significantly: SOI and willingness to have extrapair sex, r ⫽ .47, p ⬍ .001; SOI and history of extrapair sex, r ⫽ .37, p ⬍ .001; and willingness to have extrapair sex and history of extrapair sex, r ⫽ .30, p ⬍ .02. At the same time, the measures were not so highly correlated so as to be completely redundant with one another. A first principal component running through these variables accounted for 60% of their total variance (loadings range from .73 to .83). To tap this component with a single measure, we z scored all three variables and

Table 3 Study 2 General Linear Model Analysis of Testosterone (T) Levels Variable

F(1, 63)

Partial r

Age Relationship status EPSI EPSI ⫻ Relationship Status

1.80 2.55 0.81 4.32*

⫺.17 ⫺.20 .11 .25

Note. N ⫽ 68. T is logged. Relationship status: 0 ⫽ single, 1 ⫽ paired. For all variables, directed tests were used to obtain p values. EPSI ⫽ extrapair sexual interests. * p ⬍ .05.

MCINTYRE ET AL.

648

Table 4 Study 2 General Linear Model Analysis of Testosterone (T) Levels in Paired and Single Men Paired men

suggests that characteristics that reflect continued mating effort (e.g., diminished honesty, unwillingness to sacrifice self-interests for partners) may indeed be associated with higher T amongst men in relationships.

Single men

Variable

F(1, 21)

Partial r

F(1, 39)

Partial r

Age Relationship type Extrapair interests

2.42 2.21a 5.49*

⫺.32 .42a .45

0.80

⫺.14

1.30

⫺.18

Note. n ⫽ 37 and 44, respectively. T is logged. For all variables except relationship type, directed tests were used to obtain p values. a The F testing relationship type has degrees of freedom ⫽ 2, 21. The effect size for relationship type is a partial multiple r. * p ⬍ .05.

38) ⫽ 1.43, remained nonsignificant. The effect of dating multiple women fell just short of significance, F(1, 37) ⫽ 3.82, p ⫽ .058. Perhaps interestingly, men who reported dating multiple women tended to have lower T levels than men who reported not currently dating.2

General Discussion We found in both Study 1 and Study 2 that involvement in a romantic relationship and interest in extrapair sexual interest interact to predict T level among male college students. We interpret these results as suggesting that the relationship between romantic relationship status and T production is psychologically moderated by commitment to the sexually exclusive nature of the relationship and/or sexual interests outside the relationship. Men in relationships but who nonetheless are interested in new sexual encounters maintain high T levels despite being paired. These effects persist even when relationship length (Study 1) or depth of long-term commitment to the relationship (marital and living status; Study 2) are controlled. These additional results further bolster the interpretation that it is sexual interest in outside partners, not long-term commitment to the primary relationship, that accounts for the effects we observed. Were it the case that relationship duration or long-term commitment mediated the effects, the associations between measures of extrapair sexual interests and T would have weakened when duration and relationship type were controlled (Baron & Kenny, 1986). They showed no hint of doing so; indeed, the associations tended to strengthen in our samples. This pattern of results suggests that factors that contribute to relationship duration and commitment and those that contribute to extrapair sexual interests have at least partially independent effects on the T of men in relationships. Of course, these results do not imply that there are no differences in the qualities of the relationships of men who are relatively open to extrapair sexual relationships and those who are not, even with relationship duration controlled, or that correlates do not also predict T in paired men. We suspect there may be differences (e.g., with respect to men’s desire for intimacy, willingness to sacrifice self-interests for partners, honesty with their partners, or desire to spend time with their partners). Possibly, these characteristics relate to T in paired men in a fashion similar to men’s interests in extrapair sex. Future research may explore the more general set of correlates of T in paired men. The framework we have used

The Social Modulation of T These findings are consistent with a growing literature that indicates the importance of the understanding that men’s T levels are modulated by social circumstances and the motivations that they induce. As noted earlier, a variety of studies indicate that T levels are affected by outcomes of social competition (e.g., Bernhardt et al., 1998; Mazur et al., 1992, 1997). In addition, a recent study showed that men who interacted with an attractive woman experienced increased T production (Roney et al., 2003). T varies as a function of relationship status (e.g., Booth & Dabbs, 1993; Burnham et al., 2003; Gray et al., 2002; Mazur & Michalek, 1998). The current study suggests that these latter effects are at least partly due to the modulation of T by social circumstances: Although the T levels of men may, on average, drop when they enter relationships, the T levels of men who retain extrapair sexual interests even when in a relationship show no evidence of such a drop. Were this association due solely to the effects of T on men’s tendency to be in relationships, there would have been little reason to expect an interaction between men’s extrapair sexual interests and their T levels. Men who were less likely to be in committed relationships (e.g., perhaps men with higher SOI scores) would be expected to have higher T independent of whether they were currently in a committed relationship. Possibly, though men with higher T are less likely to be in relationships, when they do get into relationships, their higher T predisposes them to be interested in extrapair sex. Perhaps when single men are asked about their willingness to have extrapair sex, their answers are relatively meaningless because they have no specific partner on whom they would cheat, leading T not to be associated with their responses (and to the interaction we observed). This view implies no differential effect of relationship status on T across different sorts of men. Inconsistent with the view, however, are findings on the SOI. The content of the SOI largely concerns sexual attitudes and sexual history. Were it simply the case that men who have high T are more likely to cheat when in relationships, their T should covary with the SOI when they are single, but we did not observe that pattern of results. This view also suggests that more experienced, single men with low T should be more likely to have been in relationships than those with high T, but we did not find that the percentage of time more experienced, single men have been in a relationship the past 3 years predicts their T levels. Further support for the moderating effect of men’s sociosexual interests on the influence of relation2 We followed up our main analyses with parallel analyses by using the individual components of extrapair sexual interest. The two-way interaction between extrapair sexual interest and relationship status was most strongly carried by willingness to have extrapair sex, F(1, 63) ⫽ 5.95, p ⬍ .02; this interaction fell short of significance for both the SOI and history of extrapair sex. The variation across measures could possibly merely be sampling variability, however, and hence we focus on the main analysis by using the composite measure and do not interpret differences in results across measures.

MATING, TESTOSTERONE, AND EXTRAPAIR INTEREST

ship status on T could come from a longitudinal study of men’s T as they enter and depart from relationships, of the sort performed by Mazur and Michalek (1998), but which also includes a measure of sociosexual interests. If T is modulated by social circumstances, there may well be a functional reason for it to be modulated. T production presumably facilitates performance demanded by the circumstances that promote it (or, conversely, hinders performance demanded by the circumstances that lead to reduced T levels). As noted at the outset, T is widely thought to modulate allocation of mating effort, with increased T levels promoting male mating effort (social competition between men, particularly in sociosexual circumstances) and diminished T levels promoting other forms of effort (e.g., parental effort) at the expense of mating effort. The long-term effects of T on muscle growth illustrate these effects. Over shorter intervals, T probably affects performance in more subtle ways through psychological pathways (e.g., motivation, attention, allocation of cognitive effort). A recent study showing that men who had lower T levels responded with greater sympathy and alertness to infant cries (Fleming et al., 2002) may illustrate these effects. Similarly, lower T levels may promote investment in and attention to relationships and satisfying partner needs, and higher T levels may promote greater attention to alternative mates and thereby lower relationship investment.

Implications for an Understanding of Human Mating More Broadly These results may have broader implications for an understanding of the evolution of human mating. In recent years, anthropologists and psychologists have debated the nature of the mating system and selection pressures that forged human mating adaptations. Much debate concerns the role of male parental effort and its implications for sexual selection processes. Hawkes and her colleagues have proposed that men have not evolved to invest in offspring for the sake of parental investment (i.e., because of benefits to offspring). Rather, apparent paternal care (e.g., hunting and provisioning of women and offspring) evolved through its benefits in mate acquisition, not offspring quality per se, and hence qualifies as mating effort, not parental effort (Hawkes & Bird, 2002; Hawkes, O’Connell, & Jones, 2001; O’Connell, Hawkes, Lupo, & Jones, 2002). According to this view, then, sexual selection processes (selection on individuals’ ability to acquire mates) are fundamental to understanding any aspect of human male mating adaptation, including ones in which T is involved. Others have argued that substantial paternal effort is key to understanding adaptations involved in human reproduction (e.g., Kaplan, Hill, Lancaster, & Hurtado, 2000; see also Ellison, 2001). In this view, sexual selection may have played some role in forging male sex-specific adaptations, but many sex-specific adaptations may be due to the nature of sex-specific parental investments. For instance, men’s greater musculature (as facilitated by T) may have been maintained at least partly because men have evolved to specialize in provisioning through hunting, given the constraints women experience because of childbearing and child care. Wood and Eagly (2002) proposed a model “that does not assume that any sexual selection pressures that contributed to physical dimorphism between the sexes are major influences on sex-typed psychological attributes” (p. 702). Instead, they propose, behavioral sex differ-

649

ences are the result of sexual divisions of labor imposed by physical differences (e.g., women’s role in childbearing).3 Relatedly, Miller and Fishkin (1997) argued that although short-term mating strategies may be a fall-out from a failure of humans to interface with their adapted-for environments, seeking a long-term mate for a close and enduring relationship is based on universal design features (i.e., part of our evolutionary heritage). (pp. 228 –229)

Yet others have argued that, though male parental care has been important in human history, both men and women exhibit “mixed” evolved strategies that involve seeking long-term mateships as well as opportunistically seeking short-term (e.g., extrapair) relationships contingent on circumstances (e.g., Buss & Schmitt, 1993; Gangestad & Simpson, 2000; Trivers, 1972). A primary source of information about ancestral selection processes is contained within the nature of the organism they have shaped: Ancestral selection pressures leave telltale “footprints” in the very nature of organismic adaptive design they shape (e.g., Andrews, Gangestad, & Matthews, 2002; Thornhill, 1997). The current findings, in conjunction with previous research, may contribute to our understanding of adaptations that regulate T. As shown previously, men’s T levels are reduced when they are mated and even more so when they have children. In conjunction with evidence that lower T facilitates parental care (Fleming et al., 2002), these findings suggest that human men, like some birds, do indeed have special biological adaptations that promote parental investment by diverting reproductive efforts away from mate seeking (Gray, Chapman, et al., 2004). The current research adds a potentially important caveat: Men who are paired do not always have lower T. When men continue to be interested in pursuing mates even when mated, they do not have lower T levels. These findings thereby suggest that men also have adaptations that promote contingent, opportunistic short-term mating. Overall, then, the pattern of findings points to roles for both paternal care and mixed mating strategies in the shaping of male adaptations affecting T. Though current findings are suggestive, no doubt more research is needed to fully assess the nature and design of the adaptations that regulate male T levels. Interesting research questions include the following: (a) How are T levels of mated men affected by the sex ratio in the population, which affects men’s mating opportunities? (b) Does extrapair sexual interest affect the T levels even of men who become fathers, or does it only affect T levels in less committed relationships? (c) Do T levels in mated men change as a function of their perceived mating opportunities and success? For example, as mated men find they have fewer opportunities because of altered circumstances or attractiveness, do their T levels accordingly drop? 3 It should be noted that Wood and Eagly (2002) did not argue this point on the basis of a view that human men have evolved to provision offspring but rather from a view that men and women have different societal roles in a division of labor. Indeed, they cite Hawkes’ work showing that men have little control over what they provide their own families—which is somewhat perplexing given the highly contrasting views these authors have concerning the role of sexual selection in shaping male mating adaptations.

MCINTYRE ET AL.

650 Summary

That relationship status is associated with male T, at least in North American populations, is now fairly well established. This association may be due to the fact that men reduce T and thereby allocate effort away from mate seeking when they are mated. On the basis of the idea that some men may engage in mate seeking even when mated, the current research explored whether this association is moderated by male interest in pursuing mates, even when paired. The fact that this appears to be the case suggests complexity in the biosocial regulation of male T. The adaptive nature of this complexity may be fundamental to an understanding of human romantic and sexual relationships at both proximate and ultimate levels of explanation.

References Andrews, P. A., Gangestad, S. W., & Matthews, D. (2002). Adaptationism—How to carry out an exaptationist program. Behavioral and Brain Sciences, 25, 489 –504. Archer, J., Birring, S. S., & Wu, F. C. W. (1998). The association between testosterone and aggression among young men: Empirical findings and a meta-analysis. Aggressive Behavior, 24, 411– 420. Bancroft, J. (2002). Biological factors in human sexuality. Journal of Sex Research, 39, 15–21. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Basaria, S., Leib, J., Tang, A. M., DeWeese, T. Carducci, M., Eisenberger, M., & Dobs, A. S. (2002). Long-term effects of androgen deprivation therapy in prostate cancer patients. Clinical Endocrinology, 56, 779 – 786. Berg, S. J., & Wynne-Edwards, K. E. (2001). Changes in testosterone, cortisol, and estradiol levels in men becoming fathers. Mayo Clinic Proceedings, 76, 582–592. Bernhardt, P. C., Dabbs, J. M., Fielden, J. A., & Lutter, C. D. (1998). Testosterone changes during vicarious experiences of winning and losing among fans at sporting events. Physiology and Behavior, 65, 59 – 62. Bhasin, S. (2003). Regulation of body composition by androgens. Journal of Endocrinological Investigation, 26, 814 – 822. Book, A. S., Starzyk, K. B., & Quinsey, V. L. (2001). The relationship between testosterone and aggression: A meta-analysis. Aggression and Violent Behavior, 6, 579 –599. Booth, A., & Dabbs, J. M. (1993). Testosterone and mens’ marriages. Social Forces, 72, 463– 477. Burnham, T. C., Chapman, J. F., Gray, P. B., McIntyre, M. H., Lipson, S. F., & Ellison, P. T. (2003). Men in committed, romantic relationships have lower testosterone. Hormones and Behavior, 44, 119 –122. Buss, D. M., & Schmitt, D. P. (1993). Sexual Strategies Theory: A contextual evolutionary analysis of human mating. Psychological Review, 100, 204 –232. Campbell, B. C., Lukas, W. D., & Campbell, K. L. (2001). The reproductive ecology of male immune function. In P. T. Ellison (Ed.), Reproductive ecology and human evolution (pp. 159 –178). New York: Aldine de Gruyter. Dabbs, J. M. (1997). Testosterone, smiling, and facial appearance. Journal of Nonverbal Behavior, 21, 45–55. Dabbs, J. M., Bernieri, F. J., Strong, R. K., Campo, R., & Milun, R. (2001). Going on stage: Testosterone in greetings and meetings. Journal of Research in Personality, 35, 27– 40. Dabbs, J. M., & Mohammed, S. (1992). Male and female salivary testosterone concentrations before and after sexual activity. Physiology and Behavior, 52, 195–197.

de Waal, F. B. M. (1986). The integration of dominance and social bonding in primates. Quarterly Review of Biology, 61, 459 – 479. de Waal, F. B. M., & Hoekstra, J. A. (1980). Contexts and predictability of aggression in chimpanzees. Animal Behaviour, 28, 929 –937. Elias, M. (1981). Serum cortisol, testosterone, and testosterone-binding globulin responses to competitive fighting in human males. Aggressive Behavior, 7, 215–224. Ellison, P. T. (2001). On fertile ground: A natural history of reproduction. Cambridge, MA: Harvard University Press. Fleming, A. S., Corter, C., Stallings, J., & Steiner, M. (2002). Testosterone and prolactin are associated with emotional responses to infant cries in new fathers. Hormones and Behavior, 42, 399 – 413. Gangestad, S. W., & Simpson, J. A. (2000). The evolution of human mating: Trade-offs and strategic pluralism. Behavioral and Brain Sciences, 23, 675– 687. Gaulin, S. J., & Sailer, L. D. (1984). Sexual dimorphism in weight among the primates: The relative impact of allometry and sexual selection. International Journal of Primatology, 5, 515–535. Granger, D. A., Booth, A., & Johnson, D. R. (2000). Human aggression and enumerative measures of immunity. Psychosomatic Medicine, 62, 583–590. Granger, D. A., Schwartz, E. B., Booth, A., & Arentz, M. (1999). Salivary testosterone determination in studies of child health and development. Hormones and Behavior, 35, 18 –27. Gray, P. B., Campbell, B. C., Marlowe, F. W., Lipson, S. F., & Ellison, P. T. (2004). Social variables predict between-subject but not day-to-day variation in the testosterone of US men. Psychoneuroendocrinology, 29, 1153–1162. Gray, P. B., Chapman, J. F., Burnham, T. C., McIntyre, M. H., Lipson, S. F., & Ellison, P. T. (2004). Human male pair bonding and testosterone. Human Nature—An Interdisciplinary Biosocial Perspective, 15, 119 –131. Gray, P. B., Kahlenberg, S. M., Barrett, E. S., Lipson, S. F., & Ellison, P. T. (2002). Marriage and fatherhood are associated with lower testosterone in males. Evolution and Human Behavior, 23, 193–201. Hawkes, K., & Bird, R. B. (2002). Showing off, handicap signaling, and the evolution of men’s work. Evolutionary Anthropology, 11, 58 – 67. Hawkes, K., O’Connell, J. F., & Jones, N. G. B. (2001). Hadza meat sharing. Evolution and Human Behavior, 22, 113–142. Kaplan, H., Hill, K., Lancaster, J., & Hurtado, A. M. (2000). A theory of human life history evolution: Diet, intelligence, and longevity. Evolutionary Anthropology, 9, 156 –185. Klein, S. L. (2000). The effects of hormones on sex differences in infection: From genes to behavior. Neuroscience and Biobehavioral Reviews, 24, 627– 638. Martin, R. D. (1980). Sexual dimorphism and the evolution of higher primates. Nature, 287, 273–275. Mazur, A., & Booth, A. (1998). Testosterone and dominance in men. Behavioral and Brain Sciences, 21, 353–397. Mazur, A., Booth, A., & Dabbs, J. M. (1992). Testosterone and chess competition. Social Psychology Quarterly, 55, 70 –77. Mazur, A., & Lamb, T. A. (1980). Testosterone, status, and mood in human males. Hormones and Behavior, 14, 236 –246. Mazur, A., & Michalek, J. (1998). Marriage, divorce, and male testosterone. Social Forces, 77, 315–330. Mazur, A., Susman, E. J., & Edelbrock, S. (1997). Sex difference in testosterone response to a video game contest. Evolution and Human Behavior, 18, 317–326. Miller, L. C., & Fishkin, S. A. (1997). On the dynamics of human bonding and reproductive success: Seeking windows on the adapted-for humanenvironmental interface. In J. A. Simpson & D. T. Kenrick (Eds.), Evolutionary social psychology (pp. 197–235). Mahwah, NJ: Erlbaum. Muller, M. N., & Wrangham, R. W. (2004). Dominance, aggression and

MATING, TESTOSTERONE, AND EXTRAPAIR INTEREST testosterone in wild chimpanzees: A test of the “challenge hypothesis.” Animal Behaviour, 67, 113–123. Nelson, R. J. (2000). An introduction to behavioral endocrinology. Sunderland, MA: Sinauer. Nunes, S., Fite, J. E., & French, J. A. (2000). Variation in steroid hormones associated with infant care behaviour and experience in male marmosets (Callithrix kuhlii). Animal Behaviour, 60, 1–9. Nunes, S., Fite, J. E., Patera, K. J., & French, J. A. (2001). Interactions among paternal behavior, steroid hormones, and parental experience in male marmosets (Callithrix kuhlii). Hormones and Behavior, 39, 70 – 82. O’Connell, J. F., Hawkes, K., Lupo, K. D., & Jones, N. G. B. (2002). Male strategies and Plio-Pleistocene archaeology. Journal of Human Evolution, 43, 831– 872. Pope, H. G., Kouri, E. M., & Hudson, J. I. (2000). Effects of supraphysiologic doses of testosterone on mood and aggression in normal men. Archives of General Psychiatry, 57, 133–140. Rice, W. R., & Gaines, S. D. (1994). ‘Heads I win, tails you lose’: Testing directional alternative hypotheses in ecological and evolutionary research. Trends in Ecology and Evolution, 9, 235–237. Roney, J. R., Mahler, S. V., & Maestripieri, D. (2003). Behavioral and hormonal responses of men to brief interactions with women. Evolution and Human Behavior, 24, 365–375. Schroeder, E. T., Singh, A., Bhasin, S., Storer, T. W., Azen, C., Davidson, T., et al. (2003). Effects of an oral androgen on muscle and metabolism in older, community-dwelling men. American Journal of Physiology— Endocrinology and Metabolism, 284, E120 –E128. Simpson, J. A., & Gangestad, S. W. (1991). Individual differences in sociosexuality: Evidence for convergent and discriminant validity. Journal of Personality and Social Psychology, 60, 870 – 883. Storey, A. E., Walsh, C. J., Quinton, R. L., & Wynne-Edwards, K. E. (2000). Hormonal correlates of paternal responsiveness in new and expectant fathers. Evolution and Human Behavior, 21, 79 –95.

651

Suay, F., Salvador, A., Gonzalez-Bono, E., Sanchis, C., Martinez, M., Martinez-Sanchis, S., et al. (1999). Effects of competition and its outcome on serum testosterone, cortisol and prolactin. Psychoneuroendocrinology, 24, 551–566. Thornhill, R. (1997). The concept of an evolved adaptation. In M. Daly (Ed.), Characterizing human psychological adaptations (pp. 4 –13). London: Wiley. Thornhill, R., & Gangestad, S. W. (1999). The scent of symmetry: A human sex pheromone that signals fitness? Evolution and Human Behavior, 20, 175–201. Trivers, R. L. (1972). Parental investment and sexual selection. In B. Campbell (Ed.), Sexual selection and the descent of man (pp. 136 –179). Hawthorne, NY: Aldine de Gruyter. van Honk, J., Tuiten, A., van den Hout, M., Koppeschaar, H., Thijssen, J., de Haan, E., & Verbaten, R. (2000). Conscious and preconscious selective attention to social threat: Different neuroendocrine response patterns. Psychoneuroendocrinology, 25, 577–591. van Honk, J., Tuiten, A., Verbaten, R., van den Hout, M., Koppeschaar, H., Thijssen, J., & de Haan, E. (1999). Correlations among salivary testosterone, mood, and selective attention to threat in humans. Hormones and Behavior, 36, 17–24. Wallen, K. (2001). Sex and context: Hormones and primate sexual motivation. Hormones and Behavior, 40, 339 –357. Wood, W., & Eagly, A. H. (2002). A cross-cultural analysis of the behavior of men and women: Implications of the origins of sex differences. Psychological Bulletin, 128, 699 –727.

Received September 10, 2004 Revision received December 8, 2005 Accepted December 17, 2005 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 652– 661

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.652

Stereotyping and Evaluation in Implicit Race Bias: Evidence for Independent Constructs and Unique Effects on Behavior David M. Amodio

Patricia G. Devine

New York University

University of Wisconsin—Madison

Implicit stereotyping and prejudice often appear as a single process in behavior, yet functional neuroanatomy suggests that they arise from fundamentally distinct substrates associated with semantic versus affective memory systems. On the basis of this research, the authors propose that implicit stereotyping reflects cognitive processes and should predict instrumental behaviors such as judgments and impression formation, whereas implicit evaluation reflects affective processes and should predict consummatory behaviors, such as interpersonal preferences and social distance. Study 1 showed the independence of participants’ levels of implicit stereotyping and evaluation. Studies 2 and 3 showed the unique effects of implicit stereotyping and evaluation on self-reported and behavioral responses to African Americans using double-dissociation designs. Implications for construct validity, theory development, and research design are discussed. Keywords: prejudice, stereotyping, implicit evaluation, affect, cognition

affective and semantic neural circuits are most pronounced in more basic, implicit levels of processing, theories of implicit race bias have much to gain by considering the alternative roles of affect and cognition. In the present research, we examined the roles of affect and cognition in implicit race bias and their effects on behavior. On the basis of theory and research from social psychology and neuroscience, we argue for a conceptual distinction between implicit stereotyping and implicit evaluative race bias and propose that these two forms of implicit race bias are predictive of different types of discriminatory responses.

The distinction between affect and cognition in the human psyche dates back to the earliest philosophers of the mind and continues to be a major feature of modern psychology and neuroscience. Indeed, contemporary theorists have argued that the affective– cognitive distinction is essential for understanding the mind, brain, and behavior (Cacioppo, Gardner, & Berntson, 1999; Damasio, 1994; Zajonc, 1980), and neuroscientists have delineated distinct neural pathways for basic affective and cognitive systems of learning and memory (Davis & Whalen, 2001; Squire & Zola, 1996). In the intergroup relations literature, affect and cognition traditionally correspond to two key components of race bias: prejudice and stereotyping (Allport, 1954; Devine, 1989; Dovidio, Brigham, Johnson, & Gaertner, 1996; Fiske, 1998; Mackie & Smith, 1998). Whereas the term prejudice refers to negative affective responses toward outgroup members (McConahay & Hough, 1976), the term stereotype refers to cognitive representations of culturally held beliefs about outgroup members (Hamilton, 1981). In research on traditional, explicit race biases, the conceptual distinction between prejudice and stereotyping has provided a useful framework for examining their respective contributions to different forms of discrimination (Dovidio, Esses, Beach, & Gaertner, 2004; Park & Judd, 2005). By contrast, in research on more automatic, or implicit, forms of race bias, little attention has been given to the affective– cognitive distinction or the important implications it may have for understanding the relationship between implicit biases and behavior. Because the distinction between

Relationship Between Implicit Stereotyping and Evaluation A survey of the implicit race bias literature reveals that very few studies have directly examined the relation between affective and cognitive aspects of implicit bias (for reviews, see Blair, 2001; Fazio & Olson, 2003), and none have sought to obtain independent measures of implicit stereotyping versus evaluation.1 Indeed, most expressions of race bias reflect a combination of affective and cognitive processes, and the most commonly reported African American stereotypes are negative in valence (e.g., unintelligent, hostile, poor, lazy, and dishonest; Devine & Elliot, 1995). Yet despite the common concurrence of negative valence and stereotypes of stigmatized groups, underlying distinctions between affective and cognitive components may be important for understanding mechanisms of implicit race biases and their effects on behavior.

David M. Amodio, Department of Psychology, New York University; Patricia G. Devine, Department of Psychology, University of Wisconsin— Madison. We gratefully acknowledge Carolyn Stahlhut, Ryan Beld, and Marissa Langhoff for their assistance in data collection. Correspondence concerning this article should be addressed to David M. Amodio, Department of Psychology, New York University, 6 Washington Place, New York, NY 10003. E-mail: [email protected]

1 Throughout this article, we use the term implicit evaluation rather than implicit prejudice as a more precise label to refer to automatic evaluative associations. By using implicit evaluation, we avoid invoking unintended connotations associated with the complicated construct of prejudice, such as consciously endorsed racist attitudes and beliefs (Amodio et al., 2003; Devine et al., 2002).

652

IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS

Although the distinction is seldom made, past research has featured measures that may be characterized as assessing either implicit stereotyping (e.g., Lepore & Brown, 1997; Spencer, Fein, Wolfe, Fong, & Dunn, 1998), implicit evaluation (e.g., Amodio, Harmon-Jones, & Devine, 2003; Dovidio, Kawakami, Johnson, Johnson, & Howard, 1997; Fazio, Jackson, Dunton, & Williams, 1995; Greenwald, McGhee, & Schwartz, 1998), or some combination of stereotyping and evaluation (e.g., Dovidio, Evans, & Tyler, 1986; Kawakami, Dion, & Dovidio, 1998; Rudman, Ashmore, & Gary, 2001; Wittenbrink, Judd, & Park, 1997, 2001). The use of such measures suggests that both are valid constructs that have been studied somewhat independently and that both forms of implicit bias are prevalent among European Americans, such that African Americans are typically associated with negative evaluations and with the culturally defined stereotype content (Blair, 2001). However, although the distinction between implicit stereotyping and implicit evaluation has been acknowledged in past work (e.g., Greenwald & Banaji, 1995; Greenwald et al., 2002), theorizing has not been advanced to directly address the relation between cognitive and affective mechanisms underlying these two forms of implicit race bias.

Distinct Neural Substrates for Basic Affective and Semantic Associations In the neuroscience literature, neural substrates of affective forms of learning and memory have been distinguished from semantic forms, and this distinction has implications for the present set of issues. Results from decades of research on animals and humans suggest that the amygdala and its associated subcortical circuits are central to affective learning and memory (Lang, Bradley, & Cuthbert, 1990; LeDoux, 2000). This body of work has shown that affective associations are learned quickly, often after a single presentation of an unconditioned stimulus in a fear-learning paradigm. Once learned, such associations extinguish slowly, and subsequent reconditioning to the stimulus is facilitated (Bouton, 1994). It is important to note that amygdala-based learning does not depend on semantic associations; for example, mice easily learn affective associations despite their inability to process semantic information. By comparison, semantic learning and memory (e.g., conceptual priming) are embedded in mechanisms for language, believed to be supported by a phylogenetically newer network of neocortical structures that are significantly expanded among humans compared with those of other species (Gabrieli, 1998; Rissman, Eliassen, & Blumstein, 2003; Squire & Zola, 1996). Semantic associations may be learned in the absence of affective content, such that patients with a damaged amygdala retain normal semantic associations despite the loss of conditioned physiological responses in a fear-conditioning paradigm (Bechara, Damasio, & Damasio, 2003). An examination of anatomical and neurochemical connectivity of the amygdala and surrounding structures reveals strong direct links with neural regions associated with modulating behavior on the basis of reward and punishment cues (e.g., basal ganglia, motor cortex, orbital frontal cortex; Davis & Whalen, 2001; Park & Judd, 2005) and for mobilizing fight or flight responses (e.g., via the hypothalamic–pituitary–adrenal axis; Feldman, Conforti, & Weidenfeld, 1995). By contrast, neocortical regions associated with semantic associations appear to have few, if any, direct connections to these systems. Rather, semantic associations are likely

653

embedded in distributed networks in association cortex and thus may influence social cognition by biasing higher order information processing, such as when inferring the beliefs and intentions of another person (Amodio & Frith, 2006). Although systems for affect- and semantic-based associations typically function in concert, and thus tend to appear blended in outward verbal and behavioral responses, a consideration of their distinct operations is critical for understanding the behavioral effects of implicit stereotyping and evaluation.

Relationship Between Implicit Stereotyping and Implicit Evaluative Race Bias On the basis of social psychological and neuroscientific theorizing, we proposed that implicit stereotyping and evaluation should represent independent constructs. Although past theorizing has pointed to this distinction (e.g., Greenwald & Banaji, 1995; Greenwald et al., 2002), few studies have explored it directly (cf. Dovidio et al., 1986; Kawakami et al., 1998; Rudman, Greenwald, & McGhee, 2001; Wittenbrink et al., 1997, 2001), and none has examined the respective implications of implicit stereotyping versus implicit evaluation for behavior. A limiting factor in this line of inquiry is that in previous research, independent assessments of implicit stereotyping and evaluative race bias have not been obtained, and thus it has not been possible to examine the conceptual relationship of implicit stereotyping and evaluative race bias and their potentially unique effects on behavior. Hence, the first major goal in the present work was to obtain independent measures of implicit stereotyping and evaluation that would permit a fair test of the independence hypothesis.

Differential Effects of Implicit Evaluative Race Bias and Stereotyping on Behavior If implicit stereotyping and evaluation reflect independent cognitive and affective systems, then they may be uniquely associated with different types of discriminatory responses. Consistent with this hypothesis, Millar and Tesser (1986, 1989) argued that instrumental behaviors (e.g., forming judgments and goals) are driven primarily by cognitive processes, whereas consummatory behaviors (e.g., appetitive or aversive behaviors) are driven primarily by affective– evaluative processes. On the basis of this theorizing, Dovidio and his colleagues (1996, 2004; Esses & Dovidio, 2002; see also Stangor, Sullivan, & Ford, 1991) proposed that by considering the match between the affective versus cognitive nature of race-bias measures and forms of discriminatory outcomes, greater correspondences between assessments of race bias and behavior may be attained. In a meta-analysis focusing on explicit self-reports of stereotyping and prejudice, Dovidio et al. (2004) showed that affect-based measures of race bias tended to be predictive of basic approach/ avoidance responses (e.g., nonverbal behaviors and affective responses) toward African Americans, whereas cognition-based measures of race bias tended to be predictive of the endorsement of stereotypes and support for policies that disadvantage African Americans. Although Dovidio et al.’s meta-analysis focused on explicit measures of race bias, extant findings from the implicit race bias literature are generally consistent with these predictions (Ashburn-Nardo, Knowles, & Monteith, 2003; Dovidio et al.,

AMODIO AND DEVINE

654

1997; Dovidio, Kawakami, & Gaertner, 2002; Fazio et al., 1995; McConnell & Leibold, 2001; Wilson, Lindsey, & Schooler, 2000). For example, Fazio et al. (1995) found that implicit evaluative bias was predictive of less friendly behavior toward a Black experimenter but was not associated with participants’ views on the Rodney King verdict and ensuing riots. In other research, greater implicit evaluative bias was associated with more uncomfortable interactions (e.g., less eye contact, more blinking) with a Black confederate compared with those involving a White confederate (Dovidio et al., 1997, 2002) and more negative interactions with a Black experimenter on a host of indicators, including speech hesitations and errors and behavior judged to be abrupt, unfriendly, and uncomfortable (McConnell & Leibold, 2001). By contrast, researchers have not examined the unique effects of implicit stereotyping on behavior, although some previous findings bear on the topic. For example, Kawakami et al. (1998) found that higher levels of implicit stereotyping were predictive of the attribution of stereotypic traits to a larger proportion of African Americans (in addition to reporting more prejudiced attitudes). It is important to note that in previous research, the hypothesis that implicit stereotyping and implicit evaluation are uniquely predictive of alternative forms of race-biased behavior has not been directly tested. Hence, the second main goal of the present work was to test this hypothesis directly.

Overview of Present Research In the present research, we examined the relationship between implicit stereotyping and implicit evaluative race bias and their respective effects on instrumental versus consummatory forms of race-biased behavior. Although stereotyping and evaluation processes typically operate in concert, it was necessary for us to obtain relatively pure measures of implicit stereotyping and evaluation to examine their unique effects on behavior. To this end, we designed separate implicit association tests (IATs) to assess implicit stereotyping and implicit evaluative race bias. The IAT was chosen because it has been shown to be reliable (Greenwald, Nosek, & Banaji, 2003), and it has been widely used in the implicit race bias literature (Devine, 2001). In Study 1, we examined the degree to which measures of implicit stereotyping and evaluative race bias were independent (i.e., uncorrelated). In Studies 2 and 3, we examined the unique effects of implicit stereotyping and evaluation on instrumental and consummatory forms of behavior.

Study 1 Method

Black or White by pressing one of two keys on the computer keyboard. Stimuli consisted of pleasant and unpleasant words as used by Greenwald et al. (1998) and pictures of White and Black male faces displaying neutral expressions (Malpass, Lavigueur, & Weldon, 1973) as used by Devine, Plant, Amodio, Harmon-Jones, and Vance (2002; Study 3). Pleasant words included honor, lucky, diamond, loyal, freedom, rainbow, love, honest, peace, and heaven. Unpleasant words included evil, cancer, sickness, disaster, poverty, vomit, bomb, rotten, abuse, and murder. The IAT procedure comprised five blocks of trials (Greenwald et al., 1998). Stimuli were presented individually in the center of the computer monitor in randomized order. In Block 1, participants viewed 10 Black and 10 White faces and categorized Black faces by pressing the left response key (“a” on the alphabetic keyboard) and White faces by pressing the right response key (“5” on numeric keypad). In Block 2, participants viewed 10 pleasant and 10 unpleasant words, categorizing unpleasant words with the left response key and pleasant words with the right response key. In Block 3, stimuli included White faces, Black faces, pleasant words, and unpleasant words, and response mappings were combined such that participants categorized Black faces and unpleasant words by pressing the left response key and White faces and pleasant words by pressing the right response key. This block consisted of 40 trials and was referred to as the compatible block (Greenwald et al., 1998), given that response pairings of White with good and Black with bad are compatible with Whites’ tendency to prefer White faces over Black faces. In Block 4, participants viewed 10 Black and 10 White faces but this time categorized White faces with the left response key and Black faces with the right response key to counterbalance response mappings. In Block 5, categorizations were again combined such that participants categorized White faces and unpleasant words by pressing the left response key and Black faces and pleasant words by pressing the right response key. This block included 40 trials and was referred to as the incompatible block. Half of the participants completed the IAT as described above; half completed a version with reversed response mappings. Stereotyping IAT. We designed a new IAT in which participants viewed two classes of words associated with the positive characteristics of intelligence and athleticism/rhythmicity, and categorized them as mental or physical, respectively, in addition to the Black versus White face categorizations. Intelligence and athleticism/rhythmicity are central to the African American stereotype, such that African Americans are stereotyped as more athletic/rhythmic and less intelligent than European Americans (Devine & Elliot, 1995). Because the mental and physical categories were relatively neutral, the categorization of words relating to athleticism/rhythmicity and intelligence as mental or physical did not involve evaluative judgments.2 Target word stimuli used in the stereotyping IAT were selected on the basis of pretesting.3 Ten target words were selected for each category on the basis of category fit and stereotypicality. Mental words included math, brainy, aptitude, educated, scientist, smart, college, genius, book, and read. Physical words included athletic, boxing, basketball, run, agile, dance, jump, rhythmic, track, and football. The procedure for the stereotyping IAT was identical to that of the evaluative IAT, except that the pleasant and

Participants and Procedure One hundred fifty-one European American introductory psychology students (82 women, 69 men) participated in exchange for extra course credit. After providing informed consent, participants received instructions on completing separate IAT measures of stereotyping and prejudice administered on a PC using Inquisit software (Millisecond Software, Seattle, WA). IAT order was counterbalanced across participants. After completing the measures, participants were debriefed, thanked, and dismissed.

Materials Evaluative IAT. The IAT is a dual categorization task in which participants categorize words as pleasant or unpleasant and faces as either

2 We developed additional IATs for other common African American stereotypes. Using the method by which Rudman, Greenwald, and McGhee (2001) measured implicit gender stereotyping, we pretested sets of target words related to poor (vs. wealthy), hostile (vs. friendly), and lazy (vs. motivated). In each case, however, the stereotype was strongly related to evaluation (e.g., poor is negative and wealthy is positive), and therefore these were not suitable for examining the independence of implicit evaluation and implicit stereotyping. 3 Our lab group first generated separate lists of 22 words corresponding to the physical and mental categories. Sixty-one introductory psychology students then rated the fit of each word with its respective category and its degree of association with White and Black Americans on a scale of 1 (not

IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS unpleasant target words and category labels were replaced with intelligence- and athletics-related target words and the mental and physical category labels. Hence, the compatible block included Black/physical and White/mental categorizations and the incompatible block included Black/ mental and White/physical categorizations. IAT scoring. Responses to the evaluative and stereotyping IATs were scored using the “improved algorithm,” outlined by Greenwald et al. (2003, p. 214), which produced the D statistic.4 However, because the IAT used in Study 1 consisted of the original five-block version (Greenwald et al., 1998), steps involving practice blocks were omitted. Following the algorithm, responses with latencies greater than 10,000 ms were removed. Separate means were computed for correct raw response latencies on compatible and incompatible blocks. Error responses within each block were replaced by the mean correct reaction time for that block, plus a 600-ms error penalty. D was quantified as the difference between incompatible and compatible mean reaction times divided by the pooled standard deviation of reaction times on compatible and incompatible blocks. Data from two participants were excluded because of outlying scores (Student’s t scores differed significantly from mean, p ⬍ .05), and data from one participant were excluded because a high percentage of his responses (18%) on the stereotyping IAT were faster than 300 ms (Greenwald et al., 2003); results did not differ when outliers were included.

Results Evidence for implicit bias was examined using one-sample t tests of D scores (effect size r is presented for each t value). All tests were two-tailed. Evaluative IAT scores were significantly greater than zero (M ⫽ .51, SD ⫽ .42), t(147) ⫽ 14.60, p ⬍ .001, r ⫽ .77, suggesting a negative evaluative association with Black faces relative to White faces, replicating past work. Stereotyping IAT scores were also significantly greater than zero (M ⫽ .17,

at all) to 9 (extremely). Pretest ratings of category fit for physical and mental target words exceeded the scale midpoint, ps ⬍ .001, indicating that target words were good exemplars of their respective categories, and fit scores for the mental and physical target words did not differ, t(60) ⫽ .40, p ⫽ .69, r ⫽ .05. Physical target words were rated as more stereotypical of Black people (M ⫽ 7.68, SD ⫽ 0.98) than of White people (M ⫽ 5.19, SD ⫽ 1.26), t(60) ⫽ 13.12, p ⬍ .001, r ⫽ .86, whereas mental target words were rated as more stereotypical of White people (M ⫽ 7.23, SD ⫽1.42) than of Black people (M ⫽ 4.30, SD ⫽ 1.31), t(60) ⫽ 13.31, p ⬍.001, r ⫽ .86. A separate sample of 39 participants rated the favorability of words associated with the mental/physical and pleasant/unpleasant IATs on a scale from 1 (extremely unfavorable) to 9 (extremely favorable). Pleasant words (M ⫽ 7.93, SD ⫽ 0.54) were rated much more favorably than unpleasant words (M ⫽ 1.75, SD ⫽ 0.58), t(38) ⫽ 38.83, p ⬍ .001, r ⫽ .99. Unexpectedly, mental words (M ⫽ 7.06, SD ⫽ 0.78) were rated as more favorable than were physical words (M ⫽ 6.22, SD ⫽ .77), t(38) ⫽ 7.07, p ⬍ .001, r ⫽ .75, although both mental and physical word lists were rated significantly above the neutral midpoint of the scale, ps ⬍ .001, and both were rated as more favorable than unpleasant words, ps ⬍ .001, and less favorable than pleasant words, ps ⬍ .001. Although mental words were rated more favorably than physical words, this difference was much smaller than the difference in ratings between pleasant and unpleasant words, t(38) ⫽ 28.77, p ⬍ .001, r ⫽ .98. We used covariate analyses in our hierarchical regressions to ensure that effects of stereotyping IAT scores were not driven by evaluative associations (and vice versa) because any shared variance was statistically controlled. If anything, the valence effect found among the stereotyping IAT words would enhance the relationship between implicit stereotyping and evaluation, thereby working against our hypotheses and rendering more conservative tests.

655

SD ⫽ .43), t(147) ⫽ 4.72, p ⬍ .001, r ⫽ .36, such that participants exhibited a pattern of stereotypic trait associations with Black and White faces. No effects were found for sex or IAT order, Fs ⬍ 1. Next, we tested our primary hypothesis that levels of implicit stereotyping and implicit evaluation should be independent by examining their correlation. Participants’ evaluative and stereotyping IAT scores were not significantly correlated, r(147) ⫽ .06, p ⫽ .47, supporting our hypothesis.

Discussion The results of Study 1 showed that participants possessed significant levels of implicit evaluative and stereotyping biases but that their levels of each bias were uncorrelated, suggesting conceptual independence. It is noteworthy that although athleticism, rhythmicity, and (un)intelligence represent a subset of commonly observed African American stereotypes, they are among the most central to the stereotype. Indeed, these three attributes were the most frequently cited by participants instructed to freely list traits associated with African Americans (Devine & Elliot, 1995). Because our stereotyping IAT focused on the three most central traits of the African American stereotype, and given previous findings that the activation of a central stereotype typically activates the constellation of African American stereotypes (Devine, 1989; Lepore & Brown, 1997), it is likely that our measure of implicit stereotyping reflected associations with the general African American stereotype. Nevertheless, it would be important to show that stereotyping IAT scores were predictive of responses to an African American target, reflecting stereotypic content that reached beyond traits of (un)intelligence, athleticism, and rhythmicity. Studies 2 and 3 were designed with two goals in mind: to replicate Study 1 findings and to test the hypothesis that implicit stereotyping and evaluation are uniquely predictive of different forms of race-biased behavioral outcomes. The behavioral effects of implicit stereotyping and evaluation in Studies 2 and 3 were examined using double-dissociation designs constructed to isolate unique effects of predictors on specific outcome variables. Here, we tested the hypothesis that implicit stereotyping would be associated with instrumental but not with consummatory forms of race-biased behavior, whereas implicit evaluative race bias would be associated with consummatory but not with instrumental forms of race-biased behavior.

Study 2 In Study 2, we examined the degree to which participants’ levels of implicit stereotyping and evaluation influenced their impressions of an African American student. To measure instrumental forms of behavior, we assessed participants’ use of stereotypes as they formed an impression of the African American student on the basis of the student’s writing sample (Moreno & Bodenhausen, 2001). To measure basic approach/avoidance responses associated with consummatory behaviors, we examined participants’ preference for the writer as a potential friend. We also collected participants’ affective ratings of various ethnic groups, including African Americans, using a feelings thermometer measure. We 4

IAT effects based on difference scores (e.g., Greenwald et al., 1998) replicated results reported for the D statistic in all studies. Analyses of difference scores are available from the authors.

AMODIO AND DEVINE

656

hypothesized that implicit stereotyping but not implicit evaluation would be related to more stereotypic trait ratings of the African American student, whereas implicit evaluation but not stereotyping would relate to a greater desire to befriend the writer and more negative affective responses toward African Americans.

for African Americans, European Americans, Asian Americans, and Latino Americans.

Method

As in Study 1, participants exhibited significant levels of implicit evaluation (M ⫽ 0.32, SD ⫽ 0.17), t(31) ⫽ 10.96, p ⬍ .001, r ⫽ .89, and implicit stereotyping (M ⫽ 0.29, SD ⫽ 0.23), t(31) ⫽ 7.24, p ⬍ .001, r ⫽ .79, yet IAT scores were not significantly correlated, r(30) ⫽ .16, p ⫽ .37. No significant effects emerged for sex or IAT order, Fs ⬍ 2.04, ps ⬎ .16.

Participants and Procedure Thirty-six European American introductory psychology students (15 men, 21 women) participated in exchange for extra course credit. After providing consent, participants were told that the study consisted of two parts. The experimenter explained that the first part examined people’s ability to form impressions of others on the basis of short writing samples. Participants were shown a set of 10 file folders containing different writing samples and were asked to choose one at random (although all folders contained identical materials). Participants were given the chosen folder, which contained the writer’s demographic information, a copy of the essay, and a set of forms to record their ratings. The demographic information included the writer’s name, age, sex, ethnicity, year in college, and hometown, indicating that he was a 19-year-old male African American sophomore from Milwaukee, WI. Participants transferred this demographic information onto the evaluation form; read the essay, which contained some grammatical and spelling errors; and then provided their ratings of the essay and the writer. As the second part of the study, participants completed the evaluative and stereotyping IATs, in counterbalanced order, and the feelings thermometer measure. The essay ratings, IATs, and the feelings thermometer measure were administered in this order to prioritize the more covert measures as a means of minimizing participants’ suspicions. Lastly, participants were probed for suspicion regarding the cover story and hypotheses, debriefed, thanked, and dismissed. Five participants’ data were excluded because their scores on one or more measures differed significantly from the mean ( p ⬍ .05) in a Student’s t distribution and were considered outliers. Although inclusion of outliers inflated standard errors and thus weakened effect sizes, it did not change the pattern of effects.

Materials Evaluative and stereotyping IATs. The evaluative and stereotyping IATs consisted of the same stimuli described in Study 1 but were administered using DirectRT software (Empirisoft, New York) and included sets of 20 practice trials before the compatible and incompatible blocks. The D statistic was computed as in Study 1, with the additional incorporation of responses from the practice blocks (Greenwald et al., 2003). Essay evaluation materials. The essay evaluation form included items for rating (a) the general quality and style of the essay (included to bolster the cover story), (b) the trait attributes of the writer, and (c) participants’ liking of and perceived similarity with the writer. Trait ratings of the writer were made using a scale ranging from 1 (not at all) to 10 (very much) on a list of adjectives known to be highly associated with the Black stereotype (lazy, dishonest, unintelligent, and trustworthy; Devine & Elliot, 1995) intermixed with filler traits that were relatively neutral and not typically associated with the stereotype (modest, assertive, and thoughtful). Ratings were averaged to form separate indices of stereotypic ratings (␣ ⫽ .68, with trustworthy reverse-scored) and neutral filler ratings (␣ ⫽ .53). Liking ratings were made for five items (e.g., “The writer seems like the type of person I would like to get to know better”; “The writer and I have a lot of things in common”) on a scale of 1 (strongly disagree) to 10 (strongly agree; mean ratings: ␣ ⫽ .73). Feelings thermometer. The feelings thermometer questionnaire consisted of a scale along which a range of “degrees” were depicted, from 0° (extremely unfavorable) to 100° (extremely favorable), with 50° labeled neither favorable nor unfavorable. Ratings were provided

Results IAT Effects

IAT Effects on Behavioral Responses To test our main hypotheses regarding double dissociations of the stereotyping and evaluation IATs, we used hierarchical linear regressions. First, the D score for the IAT that was not hypothesized to predict the outcome was entered as a covariate in Step 1. In Step 2, D for the hypothesized predictor was added to the regression model. We could then obtain evidence for a double dissociation by examining the simultaneous effects of the two predictors in Step 2. The semipartial r (sr) is reported as an effect size estimate.

Stereotype Ratings Evaluative IAT scores, entered in Step 1, did not predict stereotype ratings of the African American essay writer, ␤ ⫽ ⫺.17, t(29) ⫽ ⫺0.90, p ⫽ .37, sr ⫽ ⫺.17. However, higher stereotyping IAT scores were associated with more stereotypic ratings of the African American essay writer in Step 2, ␤ ⫽ .39, t(28) ⫽ 2.70, p ⫽ .03, sr ⫽ .39, whereas the effect of evaluative IAT scores remained nonsignificant, ␤ ⫽ ⫺.23, t(28) ⫽ ⫺1.33, p ⫽ .20, sr ⫽ ⫺.23. Ratings of nonstereotypic traits were not associated with scores on the stereotyping IAT, ␤ ⫽ ⫺.01, t(29) ⫽ ⫺0.04, p ⫽ .97, sr ⫽ ⫺.01, or the evaluative IAT, ␤ ⫽ ⫺.02, t(28) ⫽ ⫺0.11, p ⫽ .92, sr ⫽ ⫺.02. Finally, when nonstereotypic ratings were included as a covariate in Step 1, stereotyping IAT scores continued to predict stereotypic ratings, ␤ ⫽ .39, t(27) ⫽ 2.73, p ⬍ .01, sr ⫽ .38, whereas evaluative IAT scores did not, ␤ ⫽ ⫺.18, t(28) ⫽ ⫺1.13, p ⫽ .26, sr ⫽ ⫺.18.

Affective Responses In analyses of preference for the writer, stereotyping IAT scores entered in Step 1 were not predictive of preferences, ␤ ⫽ .06, t(29) ⫽ 0.33, p ⫽ .75, sr ⫽ .06. In Step 2, higher evaluative IAT scores were associated with less desire to befriend the essay writer, ␤ ⫽ ⫺.32, t(28) ⫽ ⫺1.79, p ⫽ .08, sr ⫽ ⫺.32, whereas the effect for stereotyping IAT scores remained nonsignificant, ␤ ⫽ .01, t(28) ⫽ .04, p ⫽ .97, sr ⫽ .01, supporting our hypothesis. Participants’ feelings thermometer ratings provided an additional index of consummatory response toward African Americans that could be used to corroborate the marginally significant effect on preference for the writer. Average thermometer ratings for Whites, Asians, and Latinos were entered in the first regression step as a baseline covariate, followed by stereotyping IAT scores in Step 2 and evaluative IAT scores in Step 3. The effect for baseline thermometer ratings was significant, ␤ ⫽ .90, t(29) ⫽ 11.40, p ⬍

IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS

.001, sr ⫽ .90, which reflected individual differences in scale usage, but the effect for stereotyping IAT scores was not significant, ␤ ⫽ ⫺.10, t(28) ⫽ ⫺1.25, p ⫽ .22, sr ⫽ ⫺.10. Notably, higher evaluative IAT scores were predictive of more negative feelings toward African Americans, ␤ ⫽ ⫺.18, t(27) ⫽ ⫺2.04, p ⫽ .05, sr ⫽ ⫺.16, consistent with effects for writer preference. Additional analyses examining IAT effects on thermometer ratings of Whites, Asians, and Latinos produced no significant effects.

Discussion The results of Study 2 further supported the independence of implicit stereotyping and implicit evaluation, such that scores on the stereotyping and evaluative IATs were not significantly correlated. Furthermore, our regression analyses revealed the hypothesized double dissociation between implicit stereotyping and implicit evaluation effects. These results indicated that cognitive and affective forms of implicit race bias are uniquely associated with instrumental versus consummatory forms of race-biased behavior, respectively, and hence showed discriminant and predictive validity of the stereotyping and evaluative IATs. It is notable that although the stereotyping IAT focused on a subset of the African American stereotype (e.g., athleticism, rhythmicity, and lack of intelligence), it predicted a broader instantiation of the stereotype, including the descriptors of lazy, dishonest, and (un)trustworthy, consistent with research evidencing strong links between subcomponents of the stereotype (e.g., Devine, 1989). Although Study 2 provided good support for our hypotheses using conventional social psychological measures, it may have been limited in some respects. For instance, the procedure of Study 2 did not provide a good model of how implicit race biases would predict a White person’s responses in anticipation of a real-life interaction with an African American. A second potential limitation was that the predictor and outcome variables were collected in the same experimental session, precluding causal inference and raising the possibility that the outcome measures might have influenced IAT scores. These limitations were addressed in Study 3, in which participants completed measures of implicit stereotyping and evaluation several weeks before being recruited for a purportedly separate experiment in which they expected to interact with an African American participant.

Method Participants In the first phase of this study, participants were 43 introductory psychology students, 23 of whom were successfully recruited later in the semester for what they believed was an unrelated study. Evaluative IAT data from 2 participants were missing because of a computer malfunction, leaving 21 participants (13 women, 8 men) with valid data from both sessions. IAT scores of participants who did versus those who did not return for Session 2 did not differ, ps ⬎ .23.

Procedure Session 1. Participants completed stereotyping and evaluative IATs in one of two counterbalanced orders, and the IATs were scored to yield D scores, as in Study 2. Session 2. Participants were told the study would involve pairs of participants. At the scheduled experiment time, the experimenter entered the waiting room and called out the names of the participant and the (imaginary) partner. The partner’s name alternated between “Darnell Stewart” and “Tyrone Washington” to suggest African American ethnicity (Greenwald et al., 1998). Noting that the partner had not yet arrived, the experimenter escorted the participant to the experiment room to get started. After providing consent, the participant was told the following: We’re studying peoples’ ability to cooperate with another person on some tasks assessing different types of general knowledge. You and a partner are going to complete a set of tasks, and then your combined score on these tasks will be compared with other teams who are in this study. You should try your best on these tasks, because the teams with the top five combined scores will be entered into a drawing for $40. Participants were then asked to rate their abilities in various subject areas, including their mathematic and verbal skills and their knowledge of sports and cultural trivia. The experimenter then left momentarily, purportedly to check for the arrival of the partner. After a few minutes, the experimenter returned to explain that the other participant had arrived and was filling out initial questionnaires in another room. The participant was then shown the one-page participant information form identical to that used in Study 2. The top half was already filled in by the partner so that the participant would see he was African American. The participant completed the bottom half of the form. Next, the experimenter noted they were running behind schedule and gave the following explanation: To save time, I’m going to have you decide which tasks you’ll do and which your partner will do. Then we’ll all go to the main testing room. Remember, you want to choose tasks for yourself and your partner that will give you the best combined score, not just so that only you or he will do well. There are four different tests: one has questions from the math SAT, another has questions from the verbal SAT, and the other two have questions about sports and popular culture.

Study 3 Study 3 comprised two sessions. In the first session, participants completed IAT measures of stereotyping and evaluative race bias. In the second, ostensibly unrelated session, participants were led to believe that they would interact with an African American partner on various tasks involving tests of academic (verbal and mathematic) and nonacademic (sports and popular culture) knowledge. Participants rated how well they thought that they and their partner would perform on each of these tasks (Ashburn-Nardo et al., 2003) as an index of instrumental behavior. To assess consummatory behavior, we measured the distance participants chose to sit from the partner’s belongings in a row of chairs just prior to their interactions (Macrae, Bodenhausen, Milne, & Jetten, 1994). We hypothesized that implicit stereotyping but not evaluation would predict stereotype-consistent performance expectations, whereas implicit evaluation but not stereotyping would predict seating distance from the partner.

657

Participants indicated which tasks they chose for themselves and their partners and then rated their perceptions of how well they and their partners would perform on each of the tasks.5 After leaving briefly to check up on the supposed partner, the experimenter explained that the participant and the partner would now meet together in another room to complete their tasks. The experimenter led the participant out of the experiment room and,

5

Participants were rather egalitarian in their assignments of the academic tasks, with 20 of 21 participants assigning one of the SAT tasks to themselves and the other to the partner. This pattern restricted the variance of task assignments, and thus it was an insensitive measure of stereotypeconsistent behavior.

AMODIO AND DEVINE

658

explaining that the partner had left momentarily to use the bathroom, directed the participant to sit in one of a row of chairs to wait. Eight identical chairs were arranged in a line, equally spaced approximately 4 in. (10.16 cm) apart along the hallway. A coat and backpack putatively belonging to the partner were placed on the chair nearest to the experiment room doorway. After the participant chose a seat, the experimenter surreptitiously recorded the participant’s seating position and then left momentarily to photocopy the participant’s information sheet. After returning, the experimenter explained that the session would have to end early and led the participant back into the experiment room. The experimenter then probed the participant for suspicion regarding the cover story and the connection between Sessions 1 and 2, provided a debriefing and full explanation of the procedures, and then thanked and dismissed the participant. Two participants expressed some suspicion but were unable to identify key aspects of the cover story, the connection between Sessions 1 and 2, or the hypotheses.

Materials Participants rated how well they thought that they would perform on the tests of SAT mathematic and verbal skills, sports trivia, and popular culture on a scale ranging from 1 (very poorly) to 9 (very well). Ratings of expected enjoyment on each task were also made on a scale ranging from 1 (not at all) to 9 (very much). Next, participants rated their expectations of their partner’s performance and enjoyment on the same tasks, using the same scales.

Results IAT Scores Participants exhibited significant levels of implicit evaluation (M ⫽ 0.38, SD ⫽ 0.29), t(20) ⫽ 5.93, p ⬍ .001, r ⫽ .80, and implicit stereotyping (M ⫽ 0.15, SD ⫽ 0.18), t(20) ⫽ 3.70, p ⬍ .001, r ⫽ .64. Evaluative and stereotyping IAT scores were uncorrelated, r(19) ⫽ .02, p ⫽ .93, replicating the findings of Studies 1 and 2.

Ratings of Partner Abilities and Enjoyment An index was created to represent the extent to which the partner was expected to perform poorly on academic tasks but to excel on nonacademic tasks, relative to participants’ own expected performance. Participants’ self-expectation ratings on each task were subtracted from their partner-expectation ratings. These scores were standardized, with ratings of counter-stereotype skills (mathematic and verbal) reverse-scored, and averaged, such that higher scores represented more stereotype-consistent expectations of the partner’s performance, relative to expectations of the self. The hypothesized double-dissociation effects were tested using hierarchical regressions as in Study 2. In Step 1, evaluative IAT scores were not significantly associated with expectations of the partner’s performance, ␤ ⫽ ⫺.24, t(19) ⫽ ⫺1.08, p ⫽ .29, sr ⫽ –.24. However, in Step 2, higher stereotyping IAT scores significantly predicted more stereotype-consistent expectations for the partner’s performance, ␤ ⫽ .47, t(18) ⫽ 2.32, p ⫽ .03, sr ⫽ .47, whereas the effect of evaluative IAT scores remained nonsignificant, ␤ ⫽ ⫺.25, t(28) ⫽ ⫺1.24, p ⫽ .23, sr ⫽ ⫺.25.6 When participant sex was included in Step 1 as a covariate, effects for sex, ␤ ⫽ ⫺.18, t(18) ⫽ ⫺0.76, p ⫽ .46, sr ⫽ ⫺.17, and evaluative IAT scores, ␤ ⫽ ⫺.19, t(18) ⫽ ⫺0.79, p ⫽ .44, sr ⫽ ⫺.18, were not significant, whereas the effect for stereotyping IAT scores remained significant, ␤ ⫽ .50, t(17) ⫽ 2.48, p ⫽ .02, sr ⫽ .49.

Similarly, ratings of expected partner enjoyment on more stereotype-consistent tasks were not associated with evaluative IAT scores in Step 1, ␤ ⫽ ⫺.06, t(19) ⫽ ⫺0.25, p ⫽ .81, sr ⫽ ⫺.06, but were significantly associated with stereotyping IAT scores in Step 2, ␤ ⫽ .44, t(18) ⫽ 2.07, p ⫽ .05, sr ⫽ .44.

Seating Distance From Partner On average, participants sat 1.7 (SD ⫽ .78) chairs away from the partner’s belongings. Stereotyping IAT scores, included in Step 1, were not associated with seating distance, ␤ ⫽ ⫺.09, t(19) ⫽ ⫺0.37, p ⫽ .71, sr ⫽ ⫺.09. However, as revealed in Step 2, participants with higher evaluative IAT scores chose to sit further from the partner’s belongings, ␤ ⫽ .44, t(18) ⫽ 2.10, p ⫽ .05, sr ⫽ .44, whereas the effect of stereotyping IAT scores remained nonsignificant, ␤ ⫽ ⫺.09, t(28) ⫽ ⫺0.45, p ⫽ .66, sr ⫽ ⫺.09, supporting our hypothesis.

Discussion The results of Study 3 corroborated and extended the findings of Study 2. Greater implicit stereotyping scores uniquely predicted more stereotype-consistent expectations for the partner’s performance, whereas greater implicit evaluation scores uniquely predicted greater seating distance from the African American partner’s belongings. These findings provided additional support for our double-dissociation hypothesis of implicit stereotyping versus evaluation, whereby implicit stereotyping is rooted in semantic processes and is uniquely predictive of discrimination associated with instrumental responses, whereas implicit evaluation is rooted in affective processes and is uniquely predictive of discrimination associated with consummatory responses. The Study 3 findings allayed concerns over some potential limitations of Study 2. First, the differential effects of implicit stereotyping and evaluation of Study 2 were replicated in a more realistic, ecologically valid context. Second, the two-session procedure used in Study 3 alleviated concerns regarding the order in which measures were administered in Study 2. Moreover, because IAT scores collected in the initial session were predictive of behaviors weeks later, our results suggest the effects of implicit bias are stable over time.

General Discussion The present research produced two major findings. First, results suggest that implicit stereotyping and evaluative race biases represent conceptually independent constructs. Despite exhibiting significant levels of bias on both implicit measures across studies, participants’ scores on these two measures were not significantly correlated, consistent with evidence for independent mechanisms of basic cognitive and affective processes (Cacioppo et al., 1999; Squire & Zola, 1996; Zajonc, 1980). Second, the results showed that implicit stereotyping and implicit evaluation have unique effects on alternative forms of race-biased behavior. The results of Study 2 showed that implicit stereotyping but not evaluation was 6 IAT scores were not associated with absolute ratings of expected performance for the self, ps ⬎ .37, or the partner, ps ⬎ .30, indicating that implicit stereotyping effects were observable only when partner ratings were anchored by participants’ self-reference.

IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS

predictive of stereotype-consistent trait ratings of a Black student that were based on a short writing sample. In contrast, implicit evaluative race bias but not stereotyping was predictive of participants’ belief that they would get along with the student as a friend. Study 3 extended these findings by focusing on participants’ behavior as they prepared to interact with an African American partner. In this study, implicit stereotyping, but not evaluation, predicted stereotype-consistent expectations of how well the African American partner would perform on a series of tasks. On the other hand, implicit evaluative race bias, but not stereotyping, predicted how far participants chose to sit from the African American partner’s belongings in a row of chairs. Although the samples used in Studies 2 and 3 were relatively small, the use of doubledissociation designs ensured that null effects were always interpreted in the context of a complementary significant effect, and therefore low statistical power cannot account for the pattern of results. Taken together, these findings support the overarching hypothesis that implicit stereotyping processes are predictive of instrumental forms of race-biased behavior, whereas implicit evaluative processes are predictive of consummatory forms of racebiased behavior.

Implications for Theory and Research on Implicit Race Bias Clarifying the Construct of Implicit Race Bias In recent years, social psychologists have grappled with the meaning of implicit race biases in an effort to understand what they represent, how they function, and what they may predict (cf. Devine, 2001; Fazio & Olson, 2003). Against a backdrop of mixed findings regarding the effects of implicit race bias on behavior (Blair, 2001), our theorizing and results suggest that significant effects of implicit race bias on behavior may be observed when their underlying affective versus cognitive processes are taken into consideration and are matched with classes of behavior associated with consummatory versus instrumental responses (cf. Ajzen & Fishbein, 1977). On the basis of neuroscientific research, implicit evaluation is supported by subcortical mechanisms and is most directly expressed in basic approach/withdrawal behaviors. By contrast, implicit stereotyping is supported by neocortical networks and is most directly expressed in biased cognitive processing. This analysis provides a theoretical basis for conceiving of implicit stereotyping and evaluation as independent constructs and suggests refined definitions of these constructs that are rooted in neural mechanisms of learning and memory. Furthermore, it suggests that the effects of implicit stereotyping versus evaluation are likely to be expressed to different degrees in different situations and on different assessments (Livingston & Brewer, 2002; Macrae, Bodenhausen, Milne, Thorn, & Castelli, 1997). Although findings to date regarding the effect of implicit race bias on behavior are notoriously mixed, many null effects reported in the literature may have resulted from a mismatch between forms of implicit bias with outcome measures of discrimination.

Implications for Theory To date, theories of implicit race bias have not addressed the possibility that implicit forms of stereotyping and evaluation may arise from distinct underlying processes and may affect behavior

659

via alternative routes of processing (cf. Greenwald et al., 2002). Granted, stereotypes and affective responses are typically congruent and work together to facilitate a coordinated response (e.g., racial discrimination). Nevertheless, the predictive utility of a theory depends on whether it can be used to discern underlying processes and their respective effects on behavior. Future models of implicit race bias will benefit from the conceptual distinction presented here in several ways. First, a consideration of alternative forms of implicit bias will enhance predictive validity by permitting more refined hypotheses for how different forms of implicit bias should affect behavior. Second, our analysis links implicit stereotyping and evaluative bias to physiological models of the brain and behavior, permitting integration with other theoretical approaches and suggesting appropriate physiological indicators for different forms of bias. Indeed, previous research has associated indices of amygdala activity with implicit evaluation (e.g., Amodio et al., 2003; Phelps et al., 2000). Although neural correlates of implicit racial stereotyping have not yet been determined, eventrelated potential research on stereotype-based expectancy violation is consistent with a neocortical (versus subcortical) substrate (e.g., Bartholow, Fabiana, Gratton, & Bettencourt, 2001). If implicit stereotyping and evaluation arise from distinct neural substrates, as we proposed, it follows that they are learned and unlearned via different mechanisms. One may refine theories of implicit race bias malleability and change by considering the respective dynamics of classical (fear) conditioning versus semantic associative learning. For example, human and animal models of learning and memory suggest that implicit evaluations may be learned more quickly and unlearned more slowly than implicit stereotypes. They also suggest that claims that implicit prejudice can be extinguished following a single experimental manipulation may be implausible and that other interpretations should be considered (e.g., the manipulation inhibited the initial activation of bias or elicited preconscious forms of regulation).

Implications for Study Design It follows from the theoretical implications listed above that future research will benefit from a careful selection of measures and response contexts when examining the effects of implicit bias on behavior. The results of the present work suggest that implicit evaluation corresponds most directly with consummatory responses involving basic behavioral approach/withdrawal and that these effects should be strongest when behaviors involve minimal controlled processing. By contrast, implicit stereotyping affects behavior by biasing cognitive processing and thus should be most evident on measures that involve a higher degree of cognitive processing, provided that participants are unaware of the potentially biasing effects.

Future Directions Joint Effects of Implicit and Explicit Race Bias Although we went to great lengths to distinguish effects of implicit stereotyping from those of implicit evaluation in the present work, these two forms of bias typically operate in concert. An important new theoretical issue concerns the interplay of implicit stereotyping and implicit evaluation: When and how do they operate in concert? For example, behaviors that combine elements of consummatory and

AMODIO AND DEVINE

660

instrumental responses may be best predicted by the joint effects of implicit stereotyping and prejudice. Additionally, there are many situations in which explicit measures of prejudice and stereotyping may be better predictors of behavior. Finally, although levels of implicit stereotyping and evaluation were not correlated in our samples, these two forms of implicit bias may be more strongly correlated among some groups of individuals (e.g., highly biased individuals) than others. Future research is needed to explore how the full range of discriminatory behavior may be explained by complex interactions among implicit and explicit forms of prejudice and stereotyping for different groups of people.

Regulatory Mechanisms for Implicit Stereotyping Versus Implicit Evaluative Race Bias Our findings raise new questions as to whether the behavioral effects of implicit stereotyping and evaluation may be regulated via different processes and whether either form of implicit bias is more difficult to regulate. It is likely that regulation occurs at several different levels. For example, the spreading activation of automatic stereotypes within a semantic network could be inhibited via lateral inhibition (Bodenhausen & Macrae, 1998). Alternatively, the effects of implicit stereotypes may be inhibited in behavioral channels, such that a stereotype-congruent response tendency is overridden by a deliberative unbiased response (Amodio et al., 2004). Implicit evaluation associated with amygdala activation may be inhibited by the countervailing activation of reward structures in the brain, or its effect on behavior may be overridden via controlled processes as a behavioral response is formed. The inhibition of implicit stereotyping and evaluation effects at the response-formation level likely rely on the same frontal cortical mechanisms of control (Amodio et al., 2004; Amodio, Kubota, Harmon-Jones, & Devine, 2006). On the other hand, the inhibition of stereotypes within a neocortical semantic network and evaluations within a subcortical affective network rely on different mechanisms, and thus the parameters of regulation may vary considerably. The present theoretical analysis highlights some previously unexplored complexities regarding mechanisms for regulating the effects of implicit race bias.

Conclusion Affect and cognition represent two fundamental processes of the human mind, and the distinction between affective and cognitive processes is critical for the understanding of a wide range of psychological functions (Cacioppo et al., 1999). On the basis of past social psychological and neuroscientific theories, we showed that cognitive and affective components of implicit race bias are conceptually independent and are uniquely predictive of instrumental and consummatory forms of race-biased behaviors, respectively. The present work is also notable in that we took a social neuroscientific approach: We applied neurocognitive models of learning and memory to elucidate social psychological conceptions of implicit processes that had been poorly defined. Our findings suggest that greater conceptual clarity in implicit race bias research may be achieved by considering the differential effects of implicit stereotyping and evaluation when interpreting extant findings, developing new theories, and designing future research.

References Ajzen, I., & Fishbein, M. (1977). Attitude– behavior relations: A theoretical analysis and review of empirical research. Psychological Bulletin, 84, 888 –918. Allport, G. W. (1954). The nature of prejudice. Reading, MA: AddisonWesley. Amodio, D. M., & Frith, C. D. (2006). Meeting of minds: The role of the medial frontal cortex in social cognition. Nature Reviews Neuroscience, 1, 268 –277. Amodio, D. M., Harmon-Jones, E., & Devine, P. G. (2003). Individual differences in the activation and control of affective race bias as assessed by startle eye blink responses and self-report. Journal of Personality and Social Psychology, 84, 738 –753. Amodio, D. M., Harmon-Jones, E., Devine, P. G., Curtin, J. J., Hartley, S. L., & Covert, A. E. (2004). Neural signals for the detection of unintentional race bias. Psychological Science, 15, 88 –93. Amodio, D. M., Kubota, J. T., Harmon-Jones, E., & Devine, P. G. (2006). Alternative mechanisms for regulating racial responses according to internal vs. external cues. Social Cognitive and Affective Neuroscience, 1, 26 –36. Ashburn-Nardo, L., Knowles, M. L., & Monteith, M. J. (2003). Black Americans’ implicit racial associations and their implications for intergroup judgment. Social Cognition, 21, 61– 87. Bartholow, B. D., Fabiani, M., Gratton, G., & Bettencourt, B. A. (2001). A psychophysiological analysis of cognitive processing of and affective responses to social expectancy violations. Psychological Science, 12, 197–204. Bechara, A., Damasio, H., & Damasio, A. R. (2003). Role of the amygdala in decision-making. Annual Review of Neuroscience, 985, 356 –369. Blair, I. (2001). Implicit stereotypes and prejudice. In G. Moskowitz (Ed.), Cognitive social psychology: On the tenure and future of social cognition (pp. 359 –374). Mahwah, NJ: Erlbaum. Bodenhausen, G. V., & Macrae, C. N. (1998). Stereotype activation and inhibition. In R. S. Wyer, Jr. (Ed.), Advances in social cognition (Vol. 11, pp. 1–52). Mahwah, NJ: Erlbaum. Bouton, M. E. (1994). Conditioning, remembering, and forgetting. Journal of Experimental Psychology: Animal Behavior Processes, 20, 219 –231. Cacioppo, J. T., Gardner, W. L., & Berntson, G. G. (1999). The affect system has parallel and integrative processing components: Form follows function. Journal of Personality and Social Psychology, 76, 839 – 855. Damasio, A. D. (1994). Descartes’ error: Emotion, reason, and the human brain. New York: Avon. Davis, M., & Whalen, P. J. (2001). The amygdala: Vigilance and emotion. Molecular Psychiatry, 6, 13–34. Devine, P. G. (1989). Prejudice and stereotypes: Their automatic and controlled components. Journal of Personality and Social Psychology, 56, 5–18. Devine, P. G. (2001). Implicit prejudice and stereotyping: How automatic are they? Introduction to the special section. Journal of Personality and Social Psychology, 81, 757–759. Devine, P. G., & Elliot, A. J. (1995). Are racial stereotypes really fading? The Princeton Trilogy revisited. Personality and Social Psychology Bulletin, 21, 1139 –1150. Devine, P. G., Plant, E. A., Amodio, D. M., Harmon-Jones, E., & Vance, S. L. (2002). The regulation of explicit and implicit race bias: The role of motivations to respond without prejudice. Journal of Personality and Social Psychology, 82, 835– 848. Dovidio, J., Kawakami, K., Johnson, C., Johnson, B., & Howard, A. (1997). On the nature of prejudice: Automatic and controlled processes. Journal of Experimental Social Psychology, 33, 510 –540. Dovidio, J. F., Brigham, J. C., Johnson, B. T., & Gaertner, S. L. (1996). Stereotyping, prejudice and discrimination: Another look. In C. N. McCrae, C. Stangor, & M. Hewstone (Eds.), Stereotypes and stereotyping (pp. 276 –319). New York: Guilford. Dovidio, J. F., Esses, V. M., Beach, K. R., & Gaertner, S. L. (2004). The role of affect in determining intergroup behavior: The case of willing-

IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS ness to engage in intergroup affect. In D. M. Mackie & E. R. Smith (Eds.), From prejudice to intergroup emotions: Differentiated reactions to social groups (pp. 153–171). Philadelphia: Psychology Press. Dovidio, J. F., Evans, N., & Tyler, R. B. (1986). Racial stereotypes: The contents of their cognitive representations. Journal of Experimental Social Psychology, 22, 22–37. Dovidio, J. F., Kawakami, K., & Gaertner, S. L. (2002). Implicit and explicit prejudice and interracial interaction. Journal of Personality and Social Psychology, 82, 62– 68. Esses, V. M., & Dovidio, J. F. (2002). The role of emotions in determining willingness to engage in intergroup contact. Personality and Social Psychology Bulletin, 28, 1202–1214. Fazio, R., Jackson, J., Dunton, B., & Williams, C. (1995). Variability in automatic activation as an unobtrusive measure of racial attitudes: A bona fide pipeline? Journal of Personality and Social Psychology, 69, 1013–1027. Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition research: Their meaning and uses. Annual Review of Psychology, 54, 297–327. Feldman, S., Conforti, N., & Weidenfeld, J. (1995). Limbic pathways and hypothalamic neurotransmitters mediating adrenocortical responses to neural stimuli. Neuroscience and Biobehavioral Reviews, 19, 235–240. Fiske, S. T. (1998). Stereotyping, prejudice, and discrimination. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (Vol. 2, pp. 357– 411). New York: McGraw-Hill. Gabrieli, J. D. (1998). Cognitive neuroscience of human memory. Annual Review of Psychology, 49, 87–115. Greenwald, A., McGhee, D., & Schwartz, J. (1998). Measuring individual differences in implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology, 74, 1464 –1480. Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition. Psychological Review, 102, 4 –27. Greenwald, A. G., Banaji, M. R., Rudman, L. A., Farnham, S. D., Nosek, B. A., & Mellott, D. S. (2002). A unified theory of implicit attitudes, stereotypes, self-esteem, and self-concept. Psychological Review, 109, 3–25. Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and using the Implicit Association Test: I. An improved scoring algorithm. Journal of Personality and Social Psychology, 85, 197–216. Hamilton, D. L. (1981). Stereotyping and intergroup behavior: Some thoughts on the cognitive approach. In D. L. Hamilton (Ed.), Cognitive processes in stereotyping and intergroup behavior (pp. 333–353). Hillsdale, NJ: Erlbaum. Kawakami, K., Dion, K. L., & Dovidio, J. F. (1998). Racial prejudice and stereotype activation. Personality and Social Psychology Bulletin, 24, 407– 416. Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1990). Emotion, attention, and the startle reflex. Psychological Review, 97, 377–395. LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of Neuroscience, 23, 155–184. Lepore, L., & Brown, R. (1997). Category and stereotype activation: Is prejudice inevitable? Journal of Personality and Social Psychology, 72, 275–287. Livingston, R. W., & Brewer, M. B. (2002). What are we really priming? Cue-based versus category-based processing of facial stimuli. Journal of Personality and Social Psychology, 82, 5–18. Mackie, D. M., & Smith, E. R. (1998). Intergroup relations: Insights from a theoretically integrative approach. Psychological Review, 105, 499 –529. Macrae, C. N., Bodenhausen, G. V., Milne, A. B., & Jetten, J. (1994). Out of mind but back in sight: Stereotypes on the rebound. Journal of Personality and Social Psychology, 67, 808 – 817.

661

Macrae, C. N., Bodenhausen, G. V., Milne, A. B., Thorn, T. M. J., & Castelli, L. (1997). On the activation of social stereotypes: The moderating role of processing objectives. Journal of Experimental Social Psychology, 33, 471– 489. Malpass, R. S., Lavigueur, H., & Weldon, D. E. (1973). Verbal and visual training in face recognition. Perception and Psychophysics, 14, 285–292. McConahay, J. B., & Hough, J. C. (1976). Symbolic racism. Journal of Social Issues, 32, 23– 45. McConnell, A. R., & Leibold, J. M. (2001). Relations among the Implicit Association Test, discriminatory behavior, and explicit measures of racial attitudes. Journal of Experimental Social Psychology, 37, 435– 442. Millar, M. G., & Tesser, A. (1986). Effects of affective and cognitive focus on the attitude-behavior relation. Journal of Personality and Social Psychology, 51, 270 –276. Millar, M. G., & Tesser, A. (1989). The effects of affective-cognitive consistency and thought on the attitude-behavior relation. Journal of Experimental Social Psychology, 25, 189 –202. Moreno, K. N., & Bodenhausen, G. V. (2001). Intergroup affect and social judgment: Feelings as inadmissible information. Group Processes and Intergroup Relations, 4, 21–29. Park, B. & Judd, C. M. (2005). Rethinking the link between categorization and prejudice within the social cognition perspective. Personality and Social Psychology Review, 9, 108 –130. Phelps, E. A., O’Connor, K. J., Cunningham, W. A., Funayama, S., Gatenby, J. C., Gore, J. C., & Banaji, M. R. (2000). Performance on indirect measures of race evaluation predicts amygdala activation. Journal of Cognitive Neuroscience, 12, 729 –738. Rissman, J., Eliassen, J. C., & Blumstein, S. E. (2003). An event-related fMRI investigation of implicit semantic priming. Journal of Cognitive Neuroscience, 15, 1160 –1175. Rudman, L. A., Ashmore, R. D., & Gary, M. L. (2001). “Unlearning” automatic biases: The malleability of implicit stereotypes and prejudice. Journal of Personality and Social Psychology, 81, 856 – 868. Rudman, L. A., Greenwald, A. G., & McGhee, D. E. (2001). Implicit self-concept and evaluative implicit gender stereotypes: Self and ingroup share desirable traits. Personality and Social Psychology Bulletin, 27, 1164 –1178. Spencer, S. J., Fein, S., Wolfe, C. T., Fong, C., & Dunn, M. A. (1998). Automatic activation of stereotypes: The role of self-image threat. Personality and Social Psychology Bulletin, 24, 1139 –1152. Squire, L. R., & Zola, S. M. (1996). Structure and function of declarative and nondeclarative memory systems. Proceedings of the National Academy of Sciences of the USA, 93, 13515–13522. Stangor, C., Sullivan, L. A., & Ford, T. E. (1991). Affective and cognitive determinants of prejudice. Social Cognition, 9, 359 –380. Wilson, T., Lindsey, S., & Schooler, T. (2000). A model of dual attitudes. Psychological Review, 107, 101–126. Wittenbrink, B., Judd, C. M., & Park, B. (1997). Evidence for racial prejudice at the implicit level and its relationship with questionnaire measures. Journal of Personality and Social Psychology, 72, 262–274. Wittenbrink, B., Judd, C. M., & Park, B. (2001). Evaluative versus conceptual judgments in automatic attitude activation. Journal of Experimental Social Psychology, 37, 244 –252. Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psychologist, 35, 151–175.

Received April 29, 2004 Revision received January 9, 2006 Accepted January 27, 2006 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 662– 685

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.662

Regulation Processes in Intimate Relationships: The Role of Ideal Standards Nickola C. Overall and Garth J.O. Fletcher

Jeffry A. Simpson

University of Canterbury

University of Minnesota, Twin Cities Campus

This research investigated the consistency between partner perceptions and ideal standards (ideal– perception consistency) and the partner regulation attempts of 200 individuals involved in relationships (Study 1) and 62 heterosexual couples (Study 2). As predicted, greater regulation attempts were associated with lower ideal–perception consistency, and these links operated within 3 pivotal mateevaluation dimensions and were moderated by perceived regulation success. Ideal–perception consistency also mediated the relation between partner regulation and relationship quality, and cross-lagged analyses suggested that ideal consistency and regulation influenced each other over time. Finally, stronger partner regulation was generally associated with more negative self-evaluations and more self-regulation by the targeted partner. These novel results support and extend the Ideal Standards Model (J. A. Simpson, G. J. O. Fletcher, & L. Campbell, 2001). Keywords: partner regulation, relationship improvement, ideal standards, ideal discrepancy

dresses this gap by analyzing the consequences of partner regulation for partner and relationship evaluations.

Given the importance that relationships have for psychological and physical well-being, it is no mystery why people are motivated to maintain or improve their long-term intimate relationships (see Baumeister & Leary, 1995). Indeed, an impressive body of research over the past decade has produced a sizeable list of cognitive tactics that individuals use to sustain their relationships when aspects of their partner are less than ideal. For example, individuals commonly reform their expectations to more closely fit with the reality of their partner (e.g., Fletcher, Simpson, & Thomas, 2000a), perceive their partner to more closely resemble their ideal than they actually do (e.g., Murray, Holmes, & Griffin, 1996), and enhance negative partner qualities by associating unfavorable attributes with more virtuous traits (e.g., Murray & Holmes, 1999). Such techniques are effective at maintaining relationship satisfaction and security throughout the inevitable ups and downs people experience in their intimate relationships. Despite the ubiquity and effectiveness of such cognitive maintenance strategies, however, individuals should also be motivated at times to try and change behaviors or characteristics of their partner. Surprisingly, however, partner improvement or regulation attempts have received remarkably little attention from relationship scientists, and we know next to nothing about the consequences of relationship regulation. Drawing upon established theory regarding the motivating conditions underlying regulation attempts, the present research ad-

Partner Regulation and the Ideal Standards Model The extensive theoretical and empirical work regarding selfregulation provides a beginning point for considering when individuals might be motivated to regulate their partner and the outcome of regulation attempts. The most common starting point for self-regulation models is the proposition that individuals compare the qualities of some feature of the self with a preexisting standard. The discrepancy between perceptions and relevant standards then drives emotions and cognitions and motivates behavior designed to reduce or resolve the discrepancy (see, e.g., Carver & Scheier, 1998; Higgins, 1987). Such regulation attempts, and their success, also feed back into and influence the level of consistency between perceptions and associated expectations or standards (see Carver & Scheier, 1998). A recent theory that applies some basic principles of regulation theory specifically to intimate relationships is the Ideal Standards Model proposed by Fletcher, Simpson, and colleagues (see Simpson, Fletcher, & Campbell, 2001, for an overview). However, as shall be seen, modeling regulation processes in relationships is far more complex than modeling regulation processes in the self, simply because two people are involved. Thus, for example, the consequences of partner regulation for the self include the way in which both the self and the partner perceive and respond to such attempts.1 The Ideal Standards Model first proposes that individuals possess chronically accessible mate and relationship ideal standards that predate specific relationships and are used to evaluate both potential mates and partners in existing relationships. The level of

Nickola C. Overall and Garth J.O. Fletcher, Department of Psychology, University of Canterbury, Christchurch, New Zealand; Jeffry A. Simpson, Department of Psychology, University of Minnesota, Twin Cities Campus. This research was supported by a Foundation of Research, Science and Technology New Zealand scholarship awarded to Nickola C. Overall. We thank the reviewers and the relationship research group at the University of Canterbury for their helpful comments. Correspondence concerning this article should be addressed to Nickola C. Overall, who is now at the Department of Psychology, University of Auckland, Private Bag 92019, Auckland, New Zealand. E-mail: [email protected]

1 See Simpson et al. (2001) for a detailed discussion of the similarities and differences between self-discrepancy and relationship-discrepancy models.

662

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS

consistency between ideal standards and accompanying perceptions (henceforth termed ideal–perception consistency) is then postulated to drive evaluative judgments about relationships and to signal any need for regulation. Thus, in ongoing relationships perceptions of the partner are constantly (often automatically) compared with the standards and needs of the perceiver. As already noted, researchers know that individuals are good at forming attributions to write off negative behavior (Fincham, 2001) or to rationalize their partners’ negative characteristics (Murray et al., 1996; Murray & Holmes, 1999). Nevertheless, if the partner consistently fails to meet specific standards or needs that are central to the individual, then such discrepancies are likely to be noticed, become difficult to rationalize away, become increasingly irksome, and finally motivate desires and strategies to change the partner and relationship in some way. What is the origin of the importance attached to particular standards that individuals use in their relationships? At the proximal level, there are likely to be a plethora of factors involved, including recent experiences (such as reading a self-help book, meeting an alternative potential partner, talking to one’s partner, undergoing a religious conversion) and self-perceptions of mate value. For example, if Mary perceives herself as attractive or ambitious, she is likely to set high standards for her partner on the same dimensions (Fletcher, 2002). It is also possible that the content of the key dimensions of mate evaluation has distal origins in terms of human evolution. Informed by Gangestad and Simpson’s (2000) strategic pluralism model of human mating, the Ideal Standards Model postulates the existence of three major dimensions that individuals consider when evaluating (and regulating) prospective or current partners: warmth/trustworthiness, attractiveness/vitality, and status/resources. Why are these three categories so important? Evaluating mates on these dimensions could have promoted the reproductive success of our ancestors via two distinct routes— either good investment and/or good genes. The possession of warmth and trustworthiness, for example, may signal the capacity to be a good mate and parent (i.e., the motivation for good investment), whereas either the actual possession of status and resources or the drive to obtain them might signal the ability to provide good investment. In addition, the possession of attractiveness and vitality is likely to be an indicator of good genes, signaling higher fertility and perhaps better long-term health (see Fletcher, 2002, for associated evidence). There is considerable evidence that, across many cultures, both men and women focus on these particular dimensions when looking for long-term mates (see Buss, 1999; Fletcher, 2002). Factor analytic studies of mate importance ratings also have revealed that most mate-evaluation items fall into these three categories and that this three-dimensional structure replicates well across gender, relationship status, and short-term versus long-term relationship contexts (Fletcher, Simpson, Thomas, & Giles, 1999; Fletcher, Tither, O’Loughlin, Friesen, & Overall, 2004). Because the original factor analytic research was based on items generated from open-ended protocols and responses, the results suggest that these three mate-selection categories are cognitively represented in lay schemas rather than merely existing in the minds and models of evolutionary psychologists (see Fletcher et al., 1999). To date, research testing the Ideal Standards Model has focused on the evaluation function of ideal standards, revealing (as pre-

663

dicted) that when perceptions of the current partner and relationship more closely match an individual’s ideal standards (high ideal–perception consistency), partners and relationships are evaluated more positively (Campbell, Simpson, Kashy, & Fletcher, 2001; Fletcher et al., 1999, Study 6), and breakup rates are lower (Fletcher et al., 2000a). However, the model’s second major postulate—that low consistency between perceptions and ideal standards (low ideal–perception consistency) should be associated with regulation attempts— has not been tested.

The Consequences of Partner Regulation A central question regarding partner regulation is what implications regulatory efforts have for partner perceptions and relationship evaluations. The kind of cognitive strategies we described previously help to maintain relationship satisfaction by ameliorating discrepancies between partner perceptions and ideal standards. Given that the purpose of regulation attempts is (presumably) to shift partner’s qualities closer to the ideal standard, one might plausibly suspect that partner regulation will also normally have positive consequences for the relationship. However, to preview our arguments, we predicted that attempts to change the partner would generally produce negative relationship outcomes for two main reasons. First, regulatory efforts directed toward the partner are likely to increase the salience of any discrepancy between partner attributes and ideal standards. Moreover, regulation attempts will only increase ideal–perception consistency if they are successful at bringing about change in the partner (which might be a difficult enterprise). However, unsuccessful regulation attempts, which may be commonplace, seem likely to amplify dissatisfaction and perceived partner discrepancies. Second, attempts to change the partner may well communicate lack of acceptance and, thus, have negative effects on the way targeted partners view both themselves and their relationship. We expand on these points and postulate psychological processes next by introducing a series of specific causal models concerning the associations between partner regulation, ideal– perception consistency, regulation success, and relationship quality (see Figure 1).

Model 1: The Impact of Regulation on Ideal–Perception Consistency Model 1 (Figure 1) suggests that regulation influences ideal– perception consistency. Both the Ideal Standards Model and prior regulation theories propose that regulation is motivated by a lack of consistency between actual perceptions and ideal standards. For example, if Mary places considerable importance on status and resources ideal standards, but perceives her partner John to have limited potential to be financially secure, she may encourage him to retrain or look for another job. However, the principal motivation behind regulation is to reduce any discrepancy between current perceptions and ideal standards (i.e., Mary wants to change John’s status and resources to more closely resemble her ideal). Hence, regulation attempts should feed back into and influence later judgments of ideal–perception consistency. This model exemplifies the feedback loop described in prior accounts of self-regulation (e.g., Carver & Scheier, 1998) in which individuals continually monitor the level of consistency

OVERALL, FLETCHER, AND SIMPSON

664 Model 1

Partner Ideal-Perception Consistency

Partner Regulation

Model 2 Perceived Regulation Success

Partner Ideal-Perception Consistency

Partner Regulation

Model 3

Partner Regulation

Partner Ideal-Perception Consistency

Perceived Relationship Quality

Partner Ideal-Perception Consistency

Perceived Relationship Quality

Model 4

Partner Regulation

(a) (b)

Figure 1. Models of the links between regulation, ideal–perception consistency, and perceived relationship quality.

between perceptions and ideal standards, including any changes in consistency arising from active regulation attempts. Thus, the primary consequence of regulatory behavior is a shift toward or away from a desired endpoint (i.e., goal or ideal). In the perfect relationship world, the path between partner regulation and ideal–perception consistency would be positive— stronger efforts to change the partner would produce higher ideal– perception consistency. However, changing the partner may not be an easy task. Indeed, consistent with self-perception theory (Bem, 1972), the presence of strong regulation behavior may serve to induce or maintain perceptions of low ideal–perception consistency by constantly priming the belief that the partner is not meeting expectations. Moreover, attributes such as trustworthiness, attractiveness, and ambitiousness are not easily or rapidly changed in either oneself or in others (Fletcher et al., 2000a). And it seems likely that unproductive partner regulation attempts may typically heighten the sense of dissatisfaction and produce a per-

ception that the partner is even further away from meeting important standards. Thus, it may well be the case that regulation attempts often have negative consequences for partner evaluations. Nevertheless, the size and direction of this path should (in part) depend on the effectiveness of regulation attempts, which brings us to Model 2.

Model 2: Regulation, Regulation Success, and Ideal–Perception Consistency Model 2 (Figure 1) suggests that the link between regulation and ideal–perception consistency should be moderated by the perceived success of regulation attempts. However, the moderating effect of regulation success is likely to be most marked when regulation efforts are strong. For example, Mary is likely to be more disappointed at a lack of change in John when she has repeatedly discussed with John his education and career options

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS

than she will be when she casually mentions the possibility of a night course (perhaps because her lack of success in the first example signals to Mary that John is disappointingly less ambitious than she initially assumed). Although we expected a main effect of regulation success regardless of level of regulation (i.e., individuals who are less successful should have lower ideal– perception consistency), the negative impact of low regulation success should be experienced more acutely by individuals who are working assiduously to change their partners. Therefore, we expected that individuals who had the lowest levels of ideal– perception consistency would be those who reported substantial efforts to change their partner but who perceived such efforts to be ineffective.

Model 3: Regulation, Ideal–Perception Consistency, and Relationship Quality Model 3 details how regulation and ideal–perception consistency are tied to perceptions of relationship quality. As described above, and consistent with the Ideal Standards Model, partners who possess lower consistency between their ideal standards and perceptions of their partners hold more negative judgments regarding the quality of their relationship (Campbell et al., 2001; Fletcher et al., 1999, 2000a). We surmised that stronger partner regulation attempts would also be associated with lower perceived relationship quality. Recall that Models 1 and 2 (see Figure 1) propose that a primary consequence of regulation attempts constitutes the extent to which regulation attempts modify ideal–perception consistency. Indeed, the Ideal Standards Model proposes that ideal–perception consistency judgments are key proximal-level drivers of relationship evaluations. Thus, we propose that partner regulation attempts should influence judgments of relationship quality to the extent that they influence consistency between partner perceptions and associated ideal standards. Accordingly, we predicted that ideal– perception consistency would mediate the relation between partner regulation and relationship quality (see Figure 1, Model 3).

Model 4: Longitudinal Links Between Regulation, Ideal–Perception Consistency, and Relationship Quality The fourth model (see Figure 1, Model 4) examines how ideal consistency and regulation might influence each other and relationship evaluation over time. The top path (Path a) running between partner regulation and ideal–perception consistency suggests that regulation attempts should influence perceptions of ideal consistency. We also tested the assumption that regulation is motivated by a lack of consistency between current perceptions and ideal standards (as hypothesized by the Ideal Standards Model and prior accounts of regulation; Path b). In summary, we predicted that, over time (a) greater regulation attempts should predict lower judgments of partner ideal–perception consistency and (b) lower partner ideal–perception consistency should predict greater partner regulation attempts. Finally, Model 4 also portrays how ideal–perception consistency and regulation should impact on later judgments of relationship quality. We predicted that the proximal predictor of relationship quality constitutes the extent to which the partner matches ideal standards (and not levels of partner regulation), consistent with the

665

assertion that both the primary cause and outcome of regulation is the consistency between perceptions and ideal standards, which in turn influences judgments of relationship quality (also see Model 3).

Impact of Regulation Attempts on the Partner The models outlined in Figure 1 describe the consequences that partner regulation is likely to have for the individual’s perceptions of his or her partner and relationship. A second route by which regulation might produce negative relationship outcomes is the impact that regulatory efforts actually have on the partner. Previous research indicates that lower partner ideal–perception consistency ratings not only predict more negative judgments of relationship quality (in Person a), but simultaneously predict more negative relationship evaluations harbored by Person a’s partner (Campbell et al., 2001). This is perhaps not surprising given the multiple ways (verbal and nonverbal, intentional and unintentional) in which individuals communicate their satisfaction or dissatisfaction regarding specific traits or behaviors of their partners. Indeed, one major way in which individuals might assess how they are evaluated by their partner is via their partner’s regulatory behavior. For example, if Mary tries to get John to communicate more sensitively, such efforts are likely to be noticed by John. Receiving regulation attempts from Mary might, in turn, cause John to have doubts about his standing on the warmth/trustworthiness dimension. Moreover, even if John stubbornly retains his positive view of himself, he is likely to realize that he does not conform very closely to Mary’s expectations. Thus, we predicted that being the target of partner’s regulation attempts should lower self-evaluations and/or produce more negative perceptions of how closely the self matches the partner’s ideal standards. These predicted effects are consistent with prior regulation theory and associated research illustrating that individuals do respond to the perceived appraisals and ideal standards of close others (e.g., Higgins, 1987; Leary, 2004; Moretti & Higgins, 1999). In addition, when failing to meet the expectations held by one’s partner (as revealed by their regulation attempts) individuals may be motivated to improve discrepancy-related attributes in order to boost their partner’s regard. Accordingly, we tested whether experiencing more regulation from the partner was associated with increases in self-regulation attempts.

Research Overview Across two studies, we tested a series of predictions derived from the Ideal Standards Model (and extensions) that (to our knowledge) have not been subject to prior empirical research. One major aim of these studies was to test our proposal that stronger partner regulation attempts would feed into more negative perceptions of ideal consistency (see Figure 1, Model 1). Although Studies 1 and 2a involved cross-sectional samples, gathering retrospective reports of regulation behavior over the previous 6 months allowed us to (a) test predictions regarding how past regulation is associated with current perceptions of idealconsistency (see Model 1) and (b) test various causal models regarding the impact that regulation and regulation success should

OVERALL, FLETCHER, AND SIMPSON

666

have on ideal–perception consistency and relationship evaluations (Models 2 and 3 of Figure 1). Study 1 tested these hypotheses with a sample of individuals who were currently involved in heterosexual romantic relationships. In Study 2a, our predictions were tested with a sample of couples, which allowed us to test whether the regulation attempts and ideal–perception consistency of individuals were systematically linked to the relationship evaluations (and other judgments) held by their partners. As described above, we expected that receiving strong regulation attempts from the partner would be associated with more negative self-perceptions and/or increased self-regulatory efforts. Finally, in an extension of Study 2, we collected longitudinal data to test our predictions that (a) stronger regulation attempts reduce perceptions of ideal–perception consistency over time and also that (b) lower ideal–perception consistency motivates increased regulation attempts (Model 4). In addition, we also assessed relationship quality across time to more rigorously test the proposition that the associations between regulation and relationship quality would be mediated by ideal–perception consistency, but not vice versa (see Models 3 and 4).

Study 1 In Study 1, individuals in heterosexual relationships rated their current partner perceptions, ideal standards, and partner ideal– perception consistency using the short forms of the Partner Ideal Scales developed by Fletcher et al. (1999). For each item, participants also indicated how much they had attempted to change their partner over the past 6 months and how successful any regulation attempts had been. There were several reasons why we chose to assess these variables in this manner. First, actual attempts to regulate the partner can only sensibly be reported over past periods of time (rather than the present). Thus, to avoid ambiguity we specified a 6-month period. Second, one of the aims of this research was to examine the impact of regulation and perceived regulation success. Evaluating the success of regulation attempts is likely to involve an examination of how qualities have changed over time. Accordingly, we measured self-reports of regulation attempts conducted over the past 6 months and current perceptions of ideal consistency. Consequently, for these data the causal path runs from regulation to ideal–perception consistency (Model 1, Figure 1), allowing tests of the moderating role of regulation success (see Model 2) and the mediation model outlined in Model 3. We tested three main hypotheses. First, we expected that there would be moderately large negative correlations between partner regulation attempts and ideal–perception consistency (see Model 1, Figure 1). In addition, we predicted these links would be domain specific. Greater regulation of attributes on one ideal dimension (e.g., warmth/trustworthiness) should influence perceptions of partner attributes relevant to that dimension (e.g., sensitive and caring) but not attributes associated with other ideal dimensions (e.g., characteristics related to attractiveness/vitality and/or status/resources). Second, the strength of the relation between regulation attempts and ideal–perception consistency would be moderated by the perceived success of regulation (Model 2). For example, if individuals have been unsuccessful in their regulation attempts over the past 6

months, their ideal–perception consistency should be lower. This tendency, however, should be most marked for individuals who have tried especially hard to change their partner. Third, stronger partner regulation should be associated with lower relationship quality, but this link should be mediated by ideal–perception consistency (Model 3). Specifically, stronger regulation attempts in the past 6 months should predict lower current ideal–perception consistency, which in turn should predict more negative perceptions of relationship quality.

Method Participants One hundred men and 100 women currently involved in a romantic relationship of at least 6 months duration were recruited through university laboratory classes or poster advertisements at the University of Canterbury, Christchurch, New Zealand. Participants ranged from 18 to 51 years of age, with a mean age of 23.22 years (SD ⫽ 6.10). Of the sample, 52 participants were living with their partner and 30 were married. Of the remaining participants, 78 reported their relationship as serious, 36 as steady, and 4 as casual. The mean length of relationships was 33.81 months (SD ⫽ 47.83 months).

Measurement Strategy and Psychometric Analyses All of the primary measures were constructed from the short forms of the Partner Ideal Scales. These scales have demonstrated good internal reliability, test–retest reliability, and convergent and predictive validity when used to assess the importance of partner ideal standards, and they comprise three distinct factors (Campbell et al., 2001; Fletcher et al., 1999, 2000a, 2004). The specific scale items for the three mate ideal dimensions were warmth/trustworthiness (“understanding,” “supportive,” “kind,” “good listener,” “sensitive,” and “considerate”), attractiveness/vitality (“sexy,” “nice body,” “attractive appearance,” “good lover,” “outgoing,” and “adventurous”), and status/resources (“successful,” “nice house,” “financially secure,” “dresses well,” and “good job”). The phrase “potential to achieve” was added to the items from the third ideal dimension (e.g., financially secure [or potential to achieve]). These same 172 partner characteristics were used to create the following five scales: (a) Partner Ideal Standards, (b) Perceptions of Actual Partner Qualities, (c) Consistency Between Partner Perceptions and Ideal Standards, (d) Actual Attempts to Change the Partner, and (e) Perceived Success of Partner Regulation Attempts. Tests of factorial structure using confirmatory factor analysis confirmed that the items for all scales formed three quasi-independent factors representing the three mate ideal dimensions (see Fletcher et al., 2004).3 The items for each ideal dimension were

2

A single intelligence item was also included in all scales. Consistent with previous research (Fletcher et al., 1999), the intelligence item loaded equally across all three ideal dimensions, and it was therefore analyzed separately. Because this item did not produce significant results when controlling for items on the other ideal dimensions, these results are not reported. 3 Using confirmatory factor analysis, a three-factor model (representing the three mate ideal dimensions) was tested and compared with a one-factor model. For all five scales, the three-factor model produced a good fit, ␹2(25, N ⫽ 200) ⫽ 50.45 to 70.83, ps ⬍ .05, comparative fit indices (CFIs) ⫽ .93 to .97, root-mean-square error of approximation (RMSEAs) ⫽ .07 to .10, and a significantly better fit, ⌬␹2(2, N ⫽ 200) ⫽ 120.61 to 415.10, ps ⬍ .001, than did the one-factor model, which uniformly fit the data poorly, ␹2(27, N ⫽ 200) ⫽ 191.44 to 470.05, ps ⬍ .001, CFIs ⫽ .53 to .75, RMSEAs ⫽ .17 to .29.

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS also internally consistent within each scale, with Cronbach’s alphas ranging from .63 to .88 (see Table 1). Consequently, for each of the following measures, items within each dimension were summed and averaged to provide single indexes for each ideal dimension. Ideal standards. Participants were asked to rate each attribute in terms of the importance that it assumed in describing their ideal partner in a close relationship (dating, living together, or married; 1 ⫽ very unimportant, 7 ⫽ very important). Higher scores reflect higher expectations for an individual’s ideal partner. Partner perceptions. Participants were asked to rate each attribute in terms of how accurately it described their current romantic partner (1 ⫽ not at all like my partner, 7 ⫽ very much like my partner), with higher scores revealing more positive partner perceptions. Ideal–perception consistency. Participants were asked to compare their current partner with their expectations regarding their ideal partner. Participants rated each attribute according to the degree to which their current romantic partner matched their ideal partner (1 ⫽ does not match my ideal at all, 7 ⫽ completely matches my ideal). Higher scores indicate greater consistency between an individual’s partner ideal standards and his or her partner perceptions. This methodology has produced valid and reliable results in prior research (e.g., Campbell et al., 2001).4 Regulation and perceived regulation success. For each attribute, participants also rated the extent to which they (a) had desired change in that aspect of their partner during the past 6 months (1 ⫽ no desire to change, 7 ⫽ strong desire to change), (b) had tried in some way to change that aspect of their partner during the past 6 months (1 ⫽ not tried at all to change, 7 ⫽ tried hard to change), and (c) were successful in any attempts to change that aspect of their partner (1 ⫽ attempts have not been successful, 7 ⫽ attempts have been successful). Higher scores represent stronger desires and efforts to change and higher perceived success in changing attributes or behaviors. If participants had not tried to change a particular aspect of their partner (i.e., they reported 1 for Question 2 above), they were instructed to report 1 for Question 3 regarding how successful regulation attempts have been. Individuals who reported zero regulation attempts for a particular ideal dimension were subsequently excluded from all of the analyses involving regulation success.5 Relationship quality. The short version of the Perceived Relationship Quality Components Inventory (PRQC; Fletcher, Simpson, & Thomas, 2000b) was used to assess relationship quality. This scale has good internal reliability and predictive validity (Fletcher et al., 2000a, 2000b). The short version consists of the seven items that most directly tap each component of relationship quality that the inventory was designed to measure: satisfaction, commitment, intimacy, trust, passion, love, and romance (e.g., “how satisfied are you with your relationship?”). Participants were asked to rate each item with reference to their current romantic relationship (1 ⫽ not at all, 7 ⫽ extremely). All items were summed and averaged to provide an overall index of relationship quality, with higher scores indicating greater perceived quality. This measure had good internal reliability (Cronbach’s ␣ ⫽ .83).

Procedure Participants completed the entire set of questionnaires individually or in same-sex groups of 2–3 people. Participants were first provided with general information about the study, assured of their anonymity and the confidentiality of all information, and informed they could withdraw from the study at any stage. Both written and verbal instructions were provided to ensure the accurate completion of all scales, and participants were instructed to complete the questionnaires in sequence without reviewing previous answers. Once consent was obtained, participants were asked to provide their gender, age, relationship status, and length of current relationship. Participants then completed the PRQC Inventory, followed by the

667

Table 1 Means, Standard Deviations, and Reliability Coefficients of All Scales (Study 1) Total Variable Partner ideal standards Warmth/trustworthiness Attractiveness/vitality Status/resources Partner perceptions Warmth/trustworthiness Attractiveness/vitality Status/resources Partner ideal–perception consistency Warmth/trustworthiness Attractiveness/vitality Status/resources Partner regulation Warmth/trustworthiness Attractiveness/vitality Status/resources Partner regulation successa Warmth/trustworthiness (N ⫽ 174) Attractiveness/vitality (N ⫽ 172) Status/resources (N ⫽ 155) Relationship quality

M

SD

IR

6.00 4.85 4.50

0.73 0.87 1.19

.85 .75 .84

5.60 5.35 5.29

0.94 0.85 1.07

.86 .72 .82

5.68 5.47 5.53

0.99 0.95 1.10

.88 .79 .86

2.46 2.24 2.30

1.22 1.10 1.30

.81 .72 .77

2.74 2.37 2.65 5.81

1.10 1.05 1.21 0.77

.70 .65 .63 .83

Note. Internal reliability (IR) was measured with Cronbach alphas. a Descriptive data regarding success of regulation attempts only include participants who reported actually attempting to change their partner on specific ideal dimensions. Thus, the sample sizes vary across dimensions for these analyses.

scales described above concerning ideal standards, perceptions, ideal– perception consistency, and regulation.6 The order in which these scales were presented was counterbalanced within each gender so that half of the sample answered the scales assessing ideal standards, perceptions and

4 Because participants completed separate scales rating the importance of ideal standards as well as ratings of actual partner perceptions, we were able to calculate a second indirect measure of ideal–perception consistency by regressing mean levels of perceptions on mean levels of ideal standards for each ideal dimension. The standardized residuals from this regression were then treated as an index of ideal–perception consistency, with more negative residuals representing a greater discrepancy between current perceptions and ideal standards relative to the sample. All of the analyses presented in Studies 1 and 2a were run with this indirect measure, producing very similar results to those reported here. (For an example of this approach, see Knee, Nanayakkara, Vietor, Neighbors, and Patrick, 2001.) 5 We have reported only the results with ratings of actual attempts to change. However, all of the analyses in Study 1 and Study 2 were run with desired change as the predictor variable (in place of regulation attempts) producing essentially identical results to those reported. However, participants consistently reported a stronger desire for change than actual attempts to change the partner, and desired change generally exhibited stronger links with ideal–perception consistency. This suggests that desire for change does not always generate efforts to improve the partner, but it nonetheless influences perceptions of ideal consistency and relationship quality. 6 At this point, participants also completed some additional questionnaires that are not germane to the current study. Hence, they are not described.

OVERALL, FLETCHER, AND SIMPSON

668

ideal–perception consistency before the scales assessing regulation and regulation success.7

Table 2 Regression Coefficients Between Partner Ideal–Perception Consistency and Partner Regulation Attempts

Results Means, standard deviations, and internal reliabilities for all scales are shown in Table 1. All scales showed adequate internal reliability and variances, and the means for partner ideal standards, perceptions, and ideal–perception consistency were similar to those reported in prior research (e.g., Campbell et al., 2001; Fletcher et al., 2000a). As is typical for such samples, participants reported high levels of relationship quality. Nonetheless, the key variable of regulation attempts revealed reasonable levels of variance, with the bulk of participants (98% of sample) reporting attempts to change some aspect of their partner in the past 6 months.

Links Between Regulation and Ideal–Perception Consistency As predicted, across the three ideal dimensions, individuals who reported more attempts to change their partner in the past 6 months perceived lower consistency between their partner perceptions and their ideal standards (rs ⫽ ⫺.35 to ⫺.59, ps ⬍ .05). However, the pattern of negative correlations could reflect overall evaluative or halo effects. To control for overall evaluation, we calculated each coefficient controlling for the ideal–perception consistency responses obtained for the other two ideal dimensions. The resulting regression coefficients are displayed in Table 2 (see Study 1 in Table 2). As can be seen, all of the within-dimension associations between regulation and ideal–perception consistency remained significant. These results illustrate that the links between regulation and ideal–perception consistency are funneled through specific ideal domains rather than via global, higher order perceptions of mate value. Thus, regulation of specific partner qualities is associated with evaluations of those particular qualities and not other attributes or global perceptions of the degree to which the partner is matching ideal standards. Another alternative explanation is that the association between regulation and ideal–perception consistency is simply produced by more negative perceptions of the partner within each dimension (as opposed to the discrepancy between perceptions and ideal standards). Recall that we collected ratings of actual partner perceptions across the three ideal dimensions. Thus, we were able to rule out this possible explanation by regressing regulation attempts on both ideal–perception consistency and actual perceptions simultaneously. For all three ideal dimensions, ideal–perception consistency remained a significant predictor of regulation attempts (␤s ⫽ ⫺.26 to ⫺.60, ps ⬍ .05). In contrast, only attractiveness/vitality partner perceptions remained associated with regulation when ideal–perception consistency was controlled (␤ ⫽ ⫺.24, p ⬍ .05) and perceptions failed to predict regulation on the other two ideal dimensions (␤s ⫽ .02 and ⫺.07 for warmth/trustworthiness and status/resources, respectively). Finally, consistent with prior theory and research advocating the motivating influence of goals, perhaps the associations shown in Table 2 are due to the importance attached to ideal standards within each dimension. For example, individuals who possess strong intimacy goals engage in relationship activities and conflict

Study 2a Ideal dimension

Study 1

Women

Men

Warmth/trustworthiness Attractiveness/vitality Status/resources

⫺.64* ⫺.47* ⫺.29*

⫺.46* ⫺.42* ⫺.17*

⫺.57* ⫺.47* ⫺.20*

Note. All coefficients control for ideal–perception consistency across the two other ideal dimensions. In Study 2a, coefficients also controlled for the associations across self- and partner regulation within the specific ideal dimension, and all paths are pooled across gender (see Footnote 12). * p ⬍ .05.

management strategies that promote intimacy within the relationship (e.g., Sanderson & Cantor, 2001; Sanderson & Karetsky, 2002). Thus, individuals with higher ideals may work harder to attain their high standards and/or requiring change of specific partner qualities may increase the importance attached to discrepancy-related attributes. To rule out this possible explanation, we regressed regulation attempts on both ideal–perception consistency and ideal standards simultaneously. For all three ideal dimensions, ideal–perception consistency remained a significant predictor of regulation attempts (␤s ⫽ ⫺.37 to ⫺.59, ps ⬍ .05) and in all cases, was a stronger predictor than the importance attached to ideal standards (␤s ⫽ .01 to .24). These results (controlling for perceptions and ideal standards) provide substantive evidence that regulation is linked with the extent to which perceptions match ideal standards and not simply with how positively or negatively self and partner are viewed or the importance attached to the qualities within each dimension. 8

7 To examine whether order produced any mean differences across the main variables, we ran a series of 3 (ratings across all three ideal dimensions) ⫻ 2 (receiving perceptions–ideal consistency vs. regulation scales first) analyses of variance with the first factor as repeated measures. No main or interaction effects for order were significant. 8 In our general treatment of regulation processes, we have proposed that regulation behavior is associated with perceiving the partner as falling short of ideal standards. Can peoples’ perceptions exceed their ideal standards and is this type of inconsistency related to regulation? To examine this, we created groups consisting of participants (a) whose perception ratings fell below the importance attached to associated ideal standards and (b) whose perceptions matched or exceeded the relevant ideal ratings. As expected, participants whose partner perceptions matched or exceeded the importance placed on each dimension had significantly higher ideal–perception consistency ratings (Ms ⫽ 5.80 to 6.30) than those individuals whose perception ratings fell below ideals (Ms ⫽ 4.64 to 5.29) ts(198) ⫽ 7.01 to 8.31, ps ⬍ .05. As established in the prior analyses, individuals who have high ideal–perception consistency are those who have exhibited very low regulation attempts. To explore this further, we regressed regulation attempts on ideal standards, partner perceptions, the grouping variable, and all possible two-way and three-way interactions. Examination of significant interactions revealed that when participants whose perceptions were below ideal ratings reported higher regulation attempts, the lower their partner perceptions and the higher their ideal

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS

669

Table 3 Regression Coefficients for All Dimensions Testing Whether Regulation Success Moderated the Links Between Regulation Attempts and Ideal–Perception Consistency Partner ideal–perception consistency Study 1 Partner regulation and regulation success Warmth/trustworthiness Partner regulation Regulation success Interaction Attractiveness/vitality Partner regulation Regulation success Interaction Status/resources Partner regulation Regulation success Interaction

Study 2a

N 174

172

155

Both

N

⫺.88* .47* .90* ⫺.75* .52* .35 ⫺.70* .57* .99*

46

44

40

Women

Men

⫺.62* .34* .28*

⫺.77* .39* .26*

⫺.87* .42* .33*

⫺.88* .41* .23

⫺.26† .06 .06

⫺.35† .10 .08

Note. Main effects have been calculated without the interaction. Analyses for Study 1 only include participants who reported regulation attempts for the specific ideal dimension. Analyses for Study 2a only include couples who both reported regulation attempts for the specific ideal dimension. The path coefficients for Study 2a are pooled across men and women (see Footnote 12), except paths that were significantly different across gender (shown in italics). † p ⬍ .10. * p ⬍ .05.

Does Perceived Regulation Success Moderate the Link Between Regulation and Ideal–Perception Consistency? As shown in Model 2 (see Figure 1), we hypothesized that the relation between regulation and ideal–perception consistency would be moderated by perceived regulation success. In other words, individuals who have tried harder to change their partner in the previous 6 months should generally have lower levels of ideal–perception consistency, but this pattern should be more pronounced for those who have been less successful in their regulation attempts. To test these predictions, we performed hierarchical regression analyses separately for each ideal dimension with ideal–perception consistency as the dependent variable. In the first step of each analysis, regulation and perceived success of regulation were entered as predictors, after which the interaction term was entered in Step 2. Only data from those individuals who reported actually attempting at least some regulation with respect to specific ideal dimensions were included in each analysis. Table 3 displays the standardized regression coefficients for each ideal dimension (see Study 1 of Table 3). The main effects for partner regulation attempts were significant for each ideal dimension, revealing that the more individuals tried to change their partners, the less they perceived their partners as meeting their ideal standards. The main effects for perceived

standards (as predicted). However, participants whose partners exceeded their ideal standards reported the same (relatively low) levels of regulation regardless of high versus low perceptions or high versus low ideal standards (see Murray et al., 2005, for an example of this kind of approach). Thus, most of the regulatory action is occurring in the context of partner perceptions falling short of ideal standards.

regulation success were also significant for all three dimensions, indicating that individuals who perceived their regulation attempts as more successful perceived relatively higher ideal–perception consistency. Finally, just as predicted, the interaction between partner regulation attempts and perceived regulation success was significant for the warmth/trustworthiness and the status/resources dimensions. These interactions were consistent across the two ideal dimensions and are illustrated in Figure 2. Although more strenuous efforts to change the partner were typically associated with lower ideal–perception consistency, this trend was more marked for individuals who were less successful in accomplishing change.

Does Ideal–Perception Consistency Mediate the Link Between Partner Regulation and Relationship Quality? We predicted that ideal–perception consistency would mediate the link between partner regulation and relationship quality (see Figure 1, Model 3). In order to demonstrate mediation, four conditions must be met (see Baron & Kenny, 1986). First, partner regulation must be significantly associated with perceived relationship quality. Second, partner regulation must be significantly associated with partner ideal–perception consistency. Third, partner ideal consistency must be significantly associated with relationship quality when controlling for regulation attempts. Finally, the size of the path from partner regulation to perceived relationship quality should be significantly reduced when partner ideal–perception consistency is controlled. We tested the mediation model with reports of partner regulation attempts across all three ideal dimensions. The results of the path analyses using multiple regression are shown in Figure 3. Solid support was marshaled for our mediation model across all three dimensions. In all cases, more effortful attempts to regulate

OVERALL, FLETCHER, AND SIMPSON

670 7

Low Succe ss High Succe ss

Ideal-Perception Consistency

6.5 6 5.5 5 4.5 4 3.5 3

Low W/T

High W/T

Low S/R

High S/R

Partner Regulation Figure 2. Interaction of partner regulation and regulation success on the warmth/trustworthiness and status/ resources dimensions, Study 1. W/T ⫽ warmth/trustworthiness and S/R ⫽ status/resources. Low scores are 1 standard deviation below the mean; high scores are 1 standard deviation above the mean.

the partner during the prior 6 months predicted lower partner ideal–perception consistency, which in turn fed into more negative perceptions of relationship quality. Moreover, the indirect effect (equivalent to the drop in the direct path between regulation attempts and relationship quality when the mediating variable is controlled) was significant in all cases (zs ⫽ ⫺3.64 to ⫺6.10, ps ⬍ .01).9 To rule out potential artifacts, we recalculated the mediation models sequentially controlling for relationship length, relationship seriousness, and gender. None of the direct or indirect paths changed in their levels of significance, and the size of the paths altered very little. In addition, as when examining the direct links between regulation and ideal–perception consistency, we recalculated all of the mediation models partialing out (a) partner perceptions and (b) ideal standards. Although the size of the paths were generally reduced, the direct effects between ideal–perception consistency and relationship quality remained positive and significant (␤s ⫽ .24 to .33, ps ⬍ .05), with one exception: The association between warmth/trustworthiness ideal–perception consistency and relationship quality fell below significance when controlling for partner perceptions (␤ ⫽ .11). These results provide further support for the impact of ideal–perception consistency on relationship perceptions beyond positive or negative evaluations of the partner or how important partner attributes are viewed.

Regulation, Regulation Success, Ideal–Perception Consistency, and Relationship Quality

the effect regulation success has on ideal–perception consistency. To test this prediction, we repeated the above mediation analyses but entered both partner regulation and perceived regulation success as predictor variables (including only those participants who reported partner regulation; see Table 3). As predicted, both partner regulation (␤s ⫽ ⫺.27 to ⫺.47, ps ⬍ .05) and regulation success (␤s ⫽ .21 to .33, ps ⬍ .10) predicted relationship quality across ideal dimensions. Also as predicted, when adding ideal– perception consistency into the model, across ideal dimensions (a) partner regulation (␤s ⫽ ⫺.70 to ⫺.88, ps ⬍ .05) and regulation success (␤s ⫽ .47 to .57, ps ⬍ .05) significantly predicted ideal– perception consistency, (b) ideal–perception consistency significantly predicted relationship quality (␤s ⫽ .33 to .42, ps ⬍ .05), and (c) the direct paths between relationship quality and partner regulation (␤s ⫽ .02 to ⫺.23, ps ⬎ .05) and regulation success (␤s ⫽ .01 to .16, ps ⬎ .05) were significantly reduced (zs ⫽ 3.43 to 5.65, ps ⬍ .01). We then added the interaction term between partner regulation and regulation success into the model. The interaction significantly predicted relationship quality in the same way as with ideal– perception consistency (see Figure 2) for the warmth/trustworthiness ( p ⬍ .10) and the status/resources ( p ⬍ .05) dimensions (see Table 3). However, as before, the direct associations were reduced below significance when controlling for ideal–perception consistency (zs ⫽ 2.93 and 3.26, ps ⬍ .05). Thus, when controlling for 9

Finally, as presented previously (see Table 3 and Figure 2), perceiving regulation attempts as unsuccessful amplified the negative impact of regulation on ideal–perception consistency. Thus, perceiving regulation as unsuccessful should also negatively influence relationship quality but, as with regulation attempts, do so via

Note that the initial path between status/resources partner regulation and relationship quality was not statistically significant, although it was in the predicted direction (r ⫽ ⫺.11). Even though this violates one of the conditions for mediation, we included this model in Figure 3 for the sake of completeness. In addition, when we controlled for ideal–perception consistency this path was significantly reduced (z ⫽ 3.64, p ⬍ .01).

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS

671

W/T Partner IdealPerception Consistency .44**

-.59**

Partner Regulation: Warmth/Trustworthiness

-.07 (-.32**)

Perceived Relationship Quality

A/V Partner IdealPerception Consistency .36**

-.43**

Partner Regulation: Attractiveness/Vitality

-.13 (-.29**)

Perceived Relationship Quality

S/R Partner IdealPerception Consistency .34**

-.34**

Partner Regulation: Status/Resources

.01 (-.11)

Perceived Relationship Quality

Figure 3. Models show ideal–perception consistency mediating the path between partner regulation and perceived relationship quality, Study 1. Values are standardized regression coefficients. Coefficients when partner ideal–perception consistency is not controlled are shown in parentheses. W/T ⫽ warmth/trustworthiness, AV ⫽ attractiveness/vitality, and S/R ⫽ status/resources. **p ⬍ .01.

level of partner regulation attempts, we found that perceiving regulation as unsuccessful was associated with more negative judgments of relationship quality (particularly when exerting strong efforts to change the partner), but this effect occurred because unsuccessful regulation attempts reduced ideal–perception consistency.

Discussion The results from Study 1 provide compelling evidence for the proposed regulation processes developed from the Ideal Standards Model. We found strong connections between partner regulation attempts and perceived consistency with ideal standards within all three ideal dimensions. It is important to note that these links were domain specific, such that more partner regulation on one ideal dimension (e.g., warmth/trustworthiness) was associated with lower partner ideal–perception consistency on the same ideal dimension and not the other dimensions (e.g., attractiveness/vitality or status/resources). In addition, these effects were not attributable to how positively or negatively the partner was perceived in a given domain or to the importance attached to the qualities within each dimension. We also tested and found provisional support for the proposed causal models regarding the consequences of regulation behavior (see Figure 1, Models 1 through 3). Across ideal dimensions, lower perceived regulation success during the prior 6 months was associated with lower partner ideal consistency. Beyond this main effect, however, the perceived success of warmth/trustworthiness and status/resources regulation attempts generally moderated the relation between regulation and ideal consistency (Model 2). More specifically, individuals who tried harder to change the partner yet

were unsuccessful reported the lowest perceptions of ideal consistency. The final model we tested proposed that ideal–perception consistency should mediate the relation between partner regulation and perceived relationship quality (Model 3). This model was strongly supported across all three ideal dimensions. More fervent partner regulation attempts during the previous 6 months (particularly if perceived as unsuccessful) were associated with lower partner ideal–perception consistency, which in turn predicted lower judgments of relationship quality. In general, the results of Study 1 confirmed our predictions. However, Study 1 examined individuals rather than romantic couples, meaning that relationship-level processes that might be critical to understanding regulation within this context could not be investigated. Study 2a sought to redress this limitation by investigating how regulation processes operate within relationship dyads.

Study 2a In Study 2a, partners involved in romantic relationships completed the same questionnaires used in Study 1. Assessing regulation and ideal–perception consistency with both members of a couple allowed us to test (and replicate) the within-partner associations found in Study 1 (e.g., the link between partner ideal– perception consistency and perceived relationship quality) but also enabled us to examine associations across partners (e.g., the link between men’s partner ideal consistency and women’s perceived relationship quality). Thus, Study 2a had two major objectives: (a) to replicate the effects found in Study 1 and (b) to test for partner effects across regulation, ideal–perception consistency, and perceived relationship quality.

OVERALL, FLETCHER, AND SIMPSON

672

Replicating Study 1 Findings The predictions for actor effects (i.e., the effect that an individual’s independent variable score has on his or her dependent variable score in an analysis, controlling for the partner’s independent variable score) were the same as Study 1. Guided by the Ideal Standards Model, we first predicted that stronger regulation attempts should be associated with lower ideal–perception consistency (see Figure 1, Model 1). However, we also expected that these effects would be channeled through each ideal dimension rather than be driven by global evaluative judgments. Second, we predicted that regulation success should moderate the link between regulation attempts and ideal–perception consistency (see Figure 1, Model 2). Third, we predicted that greater partner regulation should predict lower perceived relationship quality but that this association would be mediated by ideal–perception consistency (see Figure 1, Model 3).

Partner Effects We also predicted several partner effects. A partner effect is evident when the partner’s independent variable score predicts the actor’s dependent variable score, controlling for the actor’s independent variable score. Campbell et al. (2001), for example, have shown that lower partner ideal–perception consistency for one partner is associated with more negative relationship quality perceived by the other partner. We expected to find the same partner effect in this research, but we also wanted to test the extent to which regulation received from the partner (i.e., being the target of the partner’s regulation attempts) is related to self-judgments and regulation. As described previously, we had three main predictions. We expected that receiving strong regulation attempts from the partner would (a) reduce the extent to which individuals believed they possessed the targeted qualities (self-perceptions), (b) lower the degree to which individuals perceived they matched their partner’s ideal standards (a new variable we call inferred ideal–perception consistency), and (c) increase the efforts individuals applied to improving targeted self-attributes (self-regulation). We tested these hypotheses using structural equation modeling and predicted that we would find the same kind of domain specificity with partner effects as were demonstrated for actor effects in Study 1.

Perceptions of Partner’s Regulation In addition to examining the links between regulation reported by one partner and the judgments made by the targeted partner, we also tested the above three predictions examining participants’ perceptions of their partner’s regulation. Whether John’s perceptions and behavior are influenced by Mary’s regulatory efforts will depend upon whether John is aware of her partner regulation attempts. Previous research has found a multitude of ways in which partners try to influence each other, including direct tactics such as coercion and rational reasoning and indirect tactics such as manipulation and supplication (e.g., Bui, Raven, & Schwarzwald, 1994; Falbo & Peplau, 1980; Howard, Blumstein, & Schwartz, 1986; Orin˜a, Wood, & Simpson, 2002). Some types of regulation attempts, such as more direct strategies, may be more obvious to the partner, and other strategies may not be perceived by one or

either partner as actual attempts to change (e.g., an objective two-sided discussion). Moreover, recent research has demonstrated that perceptions of partner’s behavior (beyond the actual behavior reported by the partner) can have important effects on self. For example, Gable, Reis, and Downey (2003) explored the effects of partner’s behavior that individuals accurately detected (e.g., perceiving regulation behavior that the partner reports enacting), falsely detected (e.g., perceiving regulation behavior that is not reported by partner), or missed (e.g., not perceiving regulation behavior that the partner reports enacting). Examining daily records of both partners over a 28-day period, Gable et al. found that perceiving negative behavior from the partner (such as criticizing) was associated with reductions in mood and relationship evaluations regardless of whether the behavior was accurately or falsely detected. Other research also supports the importance of perceptions of partner’s behavior. For example, Bolger, Zuckerman, and Kessler (2000) found that perceiving support from the partner reduced adjustment toward a stressor (quite the opposite of the intentions behind support behavior). However, partner’s reports of support provision increased coping, suggesting that support is best when it is given but not perceived (i.e., missed). Thus, the effects of partner’s regulation attempts on self-perceptions and behavior may depend (to a large degree) on the extent or the way that those attempts are perceived. Accordingly, we also measured and tested the associations between perceptions of the regulation attempts received from one’s partner and individuals’ (a) self-perceptions, (b) inferences regarding self in comparison to partner’s ideal standards, and (c) self-regulatory efforts.

Method Participants Sixty-two couples involved in heterosexual romantic relationships for a minimum of 6 months were recruited via poster advertisements at the University of Canterbury. Women ranged from 18 to 43 years of age (M ⫽ 23.10, SD ⫽ 4.96) and men ranged from 18 to 49 years of age (M ⫽ 23.80, SD ⫽ 5.75). Of the sample, 28 couples were living together and 10 were married. Of the remaining couples, 16 reported their relationship as serious and 8 as steady. The mean length of the relationships was 33.90 months (SD ⫽ 33.65 months).

Scales and Psychometric Analyses Both partners of each couple completed the same scales as in Study 1 along with five additional measures that were developed to assess (a) self-perceptions, (b) inferred ideal–perception consistency, (c) self-ideal– perception consistency, (d) self-regulation, and (e) perceptions of partners’ regulation. These measures were included to test the associations across partners outlined above and to enable the elimination of alternative explanations of these effects. Self-perceptions. Participants were asked to rate each attribute from the Partner Ideal Scales (see Study 1) in terms of how accurately it described them (1 ⫽ not at all like myself, 7 ⫽ very much like myself). Higher scores reflect more positive self-perceptions. Inferred ideal–perception consistency. Participants rated each attribute from the Partner Ideal Scales in terms of the extent to which they believed they matched their partner’s ideal (1 ⫽ I do not match my partner’s ideal at all, 7 ⫽ I completely match my partner’s ideal). Higher scores reveal more positive perceptions that the self matches the ideal standards held by the partner.

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS Self-ideal–perception consistency. Participants rated each attribute according to the degree to which their current self matched their expectations regarding their ideal self (1 ⫽ do not match my ideal at all, 7 ⫽ completely match my ideal). Higher scores indicate greater consistency between an individual’s self-ideal standards and his or her self-perceptions. Self-regulation. For each attribute, participants also rated the extent to which they had tried in some way to change that aspect of themselves during the past 6 months (1 ⫽ not tried at all to change, 7 ⫽ tried hard to change), with higher scores reflecting stronger efforts to change self-attributes. Perceptions of partner’s regulation. Participants also rated each attribute according to the extent to which they had received regulation attempts from their partner. Specifically, participants were asked to “rate the extent to which your partner has actually tried in some way to change (or attempted to get you to change) this aspect of you over the last six months” (1 ⫽ not tried at all to change, 7 ⫽ tried hard to change). Higher scores represent stronger perceived regulation attempts received from partner. As with Study 1, confirmatory factor analysis confirmed the three-factor structure (representing the three ideal dimensions) of all scales,10 and the items within each dimension were internally consistent for each scale (Cronbach’s alphas ranged from .66 to .92; shown in Table 4). Consequently, for each measure, items within each dimension were summed and averaged to provide single indexes for each dimension. The PRQC Scale was used to measure perceived relationship quality (see Study 1). As before, the PRQC Scales had good internal reliability for both men and women (see Table 4). Finally, participants also completed the Rosenberg (1965) Self-Esteem Scale in order to control for overall self-evaluation across analyses. This scale assesses global feelings of self-worth. Participants were asked to rate the extent to which they agree with a series of 10 statements about themselves (e.g., “On the whole, I am satisfied with myself ”; 1 ⫽ strongly disagree, 7 ⫽ strongly agree). Items were keyed so that higher scores indicated higher self-esteem, and the items were then averaged to form an overall self-esteem score. The scale had good internal consistency (see Table 4).

Procedure The general procedures and order of questionnaires paralleled Study 1. Participants first provided background information and completed the PRQC and Self-Esteem Scales, followed by the regulation and ideal– perception consistency measures. Half of the sample completed the scales regarding the self prior to the scales regarding the partner, and half of the sample completed the scales assessing perceptions, ideal standards, and ideal–perception consistency before the scales concerning regulation and regulation success.11 Partners completed the questionnaires in separate rooms, after which they engaged in videotaped discussions (not reported here). At the conclusion of the study, each couple was debriefed, paid $40 for their participation, and entered into a $50 cash draw.

Results

673

ated with lower partner ideal–perception consistency. Because we used couples, rather than individuals as in Study 1, we used the EQS structural equation modeling (SEM) program (Bentler, 1995) to test all of our predictions. Using SEM allowed us to (a) test both partners simultaneously, (b) control for the associations in the variables across partners, and (c) control for a number of potential artifacts. An example of this analysis strategy is illustrated in Figure 4 displaying the links between partner regulation and partner ideal–perception consistency. We expected that the within-individual paths running from partner regulation to partner ideal–perception consistency would be negative and significant. In this set of analyses, however, we did not expect that the cross-partner paths (e.g., women’s partner regulation to men’s partner ideal–perception consistency) would produce significant effects. Nevertheless, the inclusion of all paths, as well as the double-headed arrow controlling the initial associations between women’s and men’s partner regulation, ensured that the within-individual paths were calculated with all associations across partners controlled. Equivalent analyses were run across all three ideal dimensions. Following the strategy described in Study 1, these models were run while partialing out ideal–perception consistency ratings for the two other ideal dimensions to control for overall partner evaluation and test whether regulation is channeled through specific mate value dimensions. Furthermore, in this study we collected ratings of individuals’ self-directed regulation attempts. Not surprisingly, participants’ levels of self-regulation and partner regulation were correlated within ideal dimensions (rs ⫽ .27 to .53, ps ⬍ .05). Thus, we calculated the models while also partialing out selfregulation within the specific ideal dimension to control for the overall tendency to regulate these qualities within the relationship. In addition, all paths were pooled across gender (e.g., constraining the path between women’s partner regulation and partner ideal– perception consistency to be equal to the equivalent path for men). Lagrange multiplier (LM) tests revealed there were no significant 2 differences in the paths across gender, ␹LM (1, N ⫽ 62) ⫽ 0.01 to 12 2.13, ps ⬎ .05. The SEM coefficients for the within-individual paths between partner regulation and partner ideal–perception consistency are presented in Table 2 (see right hand columns). As predicted, for both men and women, across ideal dimensions, individuals who reported greater attempts to change their partner perceived lower consistency between their perceptions of these qualities in their partner and their ideal standards for these attributes.

Descriptive Statistics Means, standard deviations, and internal reliabilities for all scales are shown in Table 4. The means of the scales were similar to both those in Study 1 and those reported in prior research (e.g., Campbell et al., 2001; Fletcher et al., 2000a). As in Study 1, the majority of participants (94%) reported attempting to change some aspect of their partner, and all participants reported attempting to change some aspect of the self over the prior 6 months.

Replicating the Links Between Regulation and Ideal–Perception Consistency We expected to replicate the findings from Study 1, which showed that stronger attempts to change the partner were associ-

10 Replicating Study 1, for all 10 scales, a three-factor model (representing the three ideal dimensions) produced a good fit, ␹2(25, N ⫽ 62) ⫽ 22.69 to 58.66, ps ⫽ .75 to ⬍.01, CFIs ⫽ .88 to 1.00, RMSEAs ⫽ .00 to .14, and a significantly better fit, ⌬␹2(2, N ⫽ 62) ⫽ 36.47 to 134.02, ps ⬍ .001, than a one-factor model, which consistently demonstrated poor fit, ␹2(27, N ⫽ 62) ⫽ 59.85 to 166.37, ps ⬍.01, CFIs ⫽ .44 to .84, RMSEAs ⫽ .14 to .29. 11 As in Study 1, order effects were tested. A small number of effects were significant (3 out of a total 20 possible main effects) but were not theoretically meaningful. 12 Note that although the paths were constrained to be equal across men and women, the standardized path coefficients reported can differ slightly due to gender differences in the variances of the measures.

OVERALL, FLETCHER, AND SIMPSON

674

Table 4 Means, Standard Deviations, and Reliability Coefficients of All Scales (Study 2a) Women Variable Partner ideal standards Warmth/trustworthiness Attractiveness/vitality Status/resources Partner perceptions Warmth/trustworthiness Attractiveness/vitality Status/resources Partner ideal–perception consistency Warmth/trustworthiness Attractiveness/vitality Status/resources Partner regulation Warmth/trustworthiness Attractiveness/vitality Status/resources Partner regulation successa Warmth/trustworthiness (N ⫽ 46) Attractiveness/vitality (N ⫽ 44) Status/resources (N ⫽ 40) Self-perceptions Warmth/trustworthiness Attractiveness/vitality Status/resources Self ideal–perception consistency Warmth/trustworthiness Attractiveness/vitality Status/resources Inferred ideal–perception consistency Warmth/trustworthiness Attractiveness/vitality Status/resources Self-regulation Warmth/trustworthiness Attractiveness/vitality Status/resources Perceptions of partner’s regulation Warmth/trustworthiness Attractiveness/vitality Status/resources Relationship quality Self-esteem

Men

M

SD

M

SD

Women IR

Men IR

6.22 4.77 4.97

0.57 0.94 1.24

5.97 5.12 4.23

0.62 0.74 1.28

.77 .77 .92

.76 .66 .88

5.70 5.61 5.36

0.81 0.79 0.87

5.75 5.42 5.58

0.78 0.86 0.98

.83 .67 .78

.84 .75 .81

5.79 5.81 5.77

0.96 0.86 1.03

5.84 5.41 5.88

0.78 0.83 0.78

.87 .73 .85

.81 .76 .80

2.49 2.03 2.56

1.18 1.14 1.38

2.33 2.58 2.15

1.19 1.17 1.23

.78 .80 .75

.83 .78 .85

2.86 2.14 2.56

1.27 1.12 1.18

2.66 2.75 2.69

1.20 1.11 1.17

.76 .74 .64

.82 .77 .72

5.60 4.38 5.36

0.80 0.99 0.93

5.31 4.77 5.15

0.71 0.79 0.89

.83 .77 .82

.72 .69 .78

5.35 4.18 4.89

0.94 1.12 1.35

5.13 4.66 4.78

0.84 0.90 0.96

.81 .80 .90

.78 .75 .80

5.29 4.92 5.34

1.04 1.05 1.08

5.01 5.11 4.90

0.93 0.75 1.08

.84 .77 .85

.84 .68 .84

3.08 3.32 3.48

1.35 1.09 1.48

3.42 3.18 4.00

1.20 1.10 1.30

.81 .70 .77

.80 .75 .74

2.44 2.46 2.32 6.17 5.33

1.25 1.24 1.35 0.62 1.02

3.00 2.42 3.16 6.02 5.44

1.26 1.13 1.43 0.66 0.90

.82 .76 .78 .82 .90

.80 .77 .80 .85 .88

Note. Internal reliability (IR) was measured with Cronbach’s alphas. a Descriptive data regarding regulation success include only those couples who both reported actually attempting to change their partner on specific ideal dimensions. Thus, the sample sizes vary across couples for these analyses.

As in Study 1, we wanted to eliminate the possibility that the associations across ideal–perception consistency and regulation were produced by partner perceptions or the importance attached to each ideal dimension (rather than the inconsistency between perceptions and ideal standards). Accordingly, we ran a series of SEM analyses that included (a) paths between regulation and both actual perceptions and ideal–perception consistency and (b) in a separate set of analyses, paths between regulation and both ideal standards and ideal–perception consistency. Across ideal dimensions, for both men and women, partner ideal–perception consistency was more strongly associated with partner regulation attempts (␤s ⫽ ⫺.14 to ⫺.55, average ␤ ⫽ ⫺.37) than partner perceptions (␤s ⫽ ⫺.25 to .10, average ␤ ⫽ ⫺.12) or ideal standards (␤s ⫽ .11 to .35, average ␤ ⫽ .23). These results

replicate Study 1 and provide further evidence that partner regulation is associated with the discrepancy between perceptions and ideal standards rather than the importance attached to particular partner qualities or perceptions of the partner per se.13

13

As in Study 1, the predicted pattern of effects emerged most strongly when participants perceived their partner as falling short of ideal standards compared with when partner perceptions matched or exceeded ideal standards. That is, those individuals whose partner ratings matched or exceeded their ratings of ideal importance had higher ideal–perception consistency ratings, ts(60) ⫽ 2.24 to 5.54, ps ⬍ .05, and exhibited low levels of regulation regardless of how positively or negatively they viewed their current or ideal partner (also see Footnote 8).

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS

Women’s Partner Regulation

Women’s Partner Ideal-Perception Consistency

Men’s Partner Regulation

Men’s Partner Ideal-Perception Consistency

675

Figure 4. Structural equation model testing the associations between partner regulation and partner ideal– perception consistency, Study 2a.

Does Perceived Regulation Success Moderate the Link Between Regulation and Ideal–Perception Consistency? To test our hypothesis that perceived regulation success should moderate the link between regulation and ideal–perception consistency (see Figure 1, Model 2), we repeated the hierarchical regression analyses carried out in Study 1 using SEM. We first ran a model with regulation attempts and perceived regulation success predicting ideal– perception consistency. Models were run separately for each ideal dimension and only included couples in which both members reported regulation attempts for the specific dimension (samples ranged from 40 to 46 couples; see Table 4). We again pooled the paths across gender, and there were no significant gender differences in the paths, 2 ␹LM (1, Ns ⫽ 40 – 46) ⫽ 0.07 to 2.34, ps ⬎ .05. The SEM coefficients for these analyses are presented on the right side of Table 3. As predicted, main effects revealed that the more individuals tried to change their partner, the lower their perceptions of consistency. Furthermore, the more successful warmth/trustworthiness and attractiveness/vitality regulation attempts were perceived, the higher the ratings of ideal–perception consistency. We then entered the interaction term to each model—again pooling 2 across gender with no differences in the paths, ␹LM (1, Ns ⫽ 40 – 46) ⫽ 0.12 to 0.82, ps ⬎ .05, except for one path noted below. The interactions between partner regulation and perceived regulation success were significant on the warmth/trustworthiness and—for women 2 only, ␹LM (1, N ⫽ 44) ⫽ 6.72, p ⬍ .05—attractiveness/vitality dimensions. The patterns of the interactions were the same across ideal dimensions and identical to those found in Study 1 (see Figure 2). Specifically, greater attempts to change partner attributes were associated with lower ideal–perception consistency. However, ideal– perception consistency was reduced further for participants whose regulation efforts were viewed as unsuccessful, particularly if the individual had tried very hard to bring about change.

Does Ideal–Perception Consistency Mediate the Link Between Partner Regulation and Relationship Quality? To test for mediation (see Model 3), we again used SEM to test the model for both relationship partners simultaneously (see Figure 5). Two cross-partner paths were entered into the equation; women’s ideal–perception consistency to men’s relationship quality, and men’s ideal–perception consistency to women’s relationship quality. (No other cross-partner paths were significant, and therefore no other paths were included in the model.) We also pooled all paths across gender (including the direct path from regulation to relationship quality), with no significant differences in the paths, 2 ␹LM (1, N ⫽ 62) ⫽ 0.01 to 1.22, ps ⱖ .05. The one exception was the cross-partner paths between women’s status/resources partner

ideal–perception consistency and men’s relationship quality, 2 ␹LM (1, N ⫽ 62) ⫽ 5.17, p ⬍ .05. Thus, the associated paths were left unconstrained in the status/resources model. The results for the mediation model are depicted in Figure 5. The model produced an excellent fit across all three ideal dimensions, ␹2(9, N ⫽ 62) ⫽ 4.32 to 9.39, ps ⬎ .05, CFIs ⫽ .99 to 1.00, RMSEAs ⫽ .00 to .03. For both men and women, more effortful attempts to regulate the partner in the past 6 months predicted lower levels of partner ideal–perception consistency, which in turn predicted more negative perceptions of relationship quality. Moreover, the indirect effect for all three models was significant; zs ⫽ ⫺2.55 to ⫺4.62, ps ⬍ .05. These latter results indicate that the direct paths were significantly reduced when the mediating variable was controlled. In addition, consistent with prior research (Campbell et al., 2001), higher levels of women’s ideal–perception consistency were positively related to men’s relationship quality and vice versa (with the exception of women’s status/resources ideal–perception consistency). Individuals’ perceptions of their relationship quality, in other words, were not solely a product of their own perceptions of partner ideal consistency but were also a product of the partner ideal consistency judgments of their partners. To rule out potential artifacts, we recalculated the models sequentially controlling for self-esteem, relationship length, and relationship status. None of the direct or indirect paths changed in their levels of significance, and the size of the paths altered very little. As in Study 1, to examine whether ideal–perception consistency predicts relationship quality beyond partner perceptions or ideal standards, we recalculated the mediation models partialing out these two variables. As with the links between ideal– perception consistency and regulation, the direct within-individual paths between ideal–perception consistency and relationship quality remained positive and significant (␤s ⫽ .16 to .52, ps ⬍ .05). These results provide further support that a key variable associated with partner regulation and, in turn, relationship quality is the consistency between perceptions and ideal standards.14 14

As in Study 1, we repeated the mediation analyses incorporating regulation success and the Regulation ⫻ Regulation Success interaction term as predictor variables to examine whether regulation success influenced relationship quality via perceptions of ideal consistency. Low sample sizes (40 – 46) combined with the number of variables included in these models meant that, where regulation success and the interaction term predicted ideal–perception consistency (see Table 3), three of four failed to significantly predict relationship quality, although these paths were in the predicted direction. Nevertheless, the indirect effect was significant in all cases (zs ⫽ 2.06 to 2.77, ps ⬍ .05).

OVERALL, FLETCHER, AND SIMPSON

676 Warmth/Trustworthiness

Women’s W/T Partner Ideal-Consistency .36*

-.43*

e Women’s Partner Regulation: Warmth/Trustworthiness

.34*

Women’s Perceived Relationship Quality

-.08 (-.26*)

.30*

e

.10

Men’s Partner Regulation: Warmth/Trustworthiness

.29*

Men’s Perceived Relationship Quality

-.07 (-.25*)

e

e

.30*

-.52*

.28*

Men’s W/T Partner Ideal-Consistency

Attractiveness/Vitality

Women’s A/V Partner Ideal-Consistency .42*

-.50*

e Women’s Partner Regulation: Attractiveness/Vitality

.20*

Women’s Perceived Relationship Quality

-.17* (-.35*)

-.16

e

.18

Men’s Partner Regulation: Attractiveness/Vitality

.40*

Men’s Perceived Relationship Quality

-.16* (-.34*)

e

e

.23*

-.51*

.37*

Men’s A/V Partner Ideal-Consistency

Status/Resources

Women’s S/R Ideal-Consistency .48*

-.25*

e Women’s Partner Regulation: Status/Resources

.19

Women’s Perceived Relationship Quality

-.08 (-.23*)

-.00

e

-.18

Men’s Partner Regulation: Status/Resources

.30*

Men’s Perceived Relationship Quality

-.07 (-. 19*)

e

-.29*

e

.47* .37*

Men’s S/R Ideal-Consistency

Figure 5. Structural equation models (SEM) showing ideal–perception consistency mediating the path between partner regulation and perceived relationship quality, Study 2a. Values are standardized SEM coefficients. Coefficients when partner ideal–perception consistency is not controlled are shown in parentheses. W/T ⫽ warmth/trustworthiness, A/V ⫽ attractiveness/vitality, and S/R ⫽ status/resources. e represents the error term for each variable. *p ⬍ .05.

Testing Partner Effects: Links Between Partner’s Regulation Attempts, Self-Perceptions, Inferred Ideal–Perception Consistency, and Self-Regulation We had three main predictions regarding the links between partner’s regulation attempts and judgments held by the targeted

partner. Specifically, we expected that receiving strong regulation attempts from one’s partner would be associated with (a) lower self-perceptions regarding targeted qualities, (b) lower inferred ideal–perception consistency (perceptions of the degree to which self matches partner’s ideal standards) on the targeted dimension, and (c) stronger efforts to change targeted self-attributes.

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS

677

garding how the self is perceived by the partner could be influenced by self-perceptions (e.g., Murray, Holmes, & Griffin, 2000; Murray, Holmes, Griffin, Bellavia, & Rose, 2001), we also controlled for self-perceptions on the corresponding ideal dimension. As before, the cross-paths were pooled across gender, and LM tests revealed no gender differences for any of the cross–paths, 2 ␹LM (1, N ⫽ 62) ⫽ 0.01 to 1.57, ps ⬎ .05. As shown in Table 5, for both women and men, greater regulation received from the partner on the attractiveness/vitality dimension was associated with reduced levels of inferred ideal–perception consistency. Finally, using the same analysis strategy, we tested whether participants responded to their partner’s regulation efforts by attempting to change the self. Although participants reported on their self-regulation and partner regulation attempts over the same time period, a positive association between the partner’s regulation attempts and attempts to change self may suggest that individuals increase their self-regulatory efforts when they receive regulation attempts from their partner. As before, to illustrate that the effects of partner’s regulation attempts were domain specific, we controlled self-regulation on the other two ideal dimensions. The cross-partner paths were significantly different across sex for the 2 attractiveness/vitality and status/resources dimensions, ␹LM (1, N ⫽ 62) ⫽ 3.93, p ⬍ .05, so we left the cross-paths unconstrained for these sets of analyses. As predicted, stronger regulation attempts from the partner were associated with increased self-directed regulation across ideal dimensions. That is, the more both women and men received warmth/trustworthiness regulation attempts from their partner, the more they tried to change these aspects of themselves. However, only women responded to their partner’s attractiveness/vitality regulation attempts, and only men responded to their partner’s regulatory efforts focusing on status/resources self-attributes. In summary, we found some evidence for partner effects (as predicted). In addition, when controlling for possible confounding

Using SEM, we first examined the associations between regulation attempts made by the partner (e.g., women’s regulation attempts of their male partners) and individuals’ self-perceptions (e.g., men’s self-perceptions). Our analysis strategy is equivalent to that shown in Figure 4 (with partner ideal– consistency variables replaced by self-perception ratings). If partners’ regulation attempts were related to self-perceptions as we predicted, the crosspaths (e.g., path running from women’s partner regulation to men’s self-perceptions) should be negative and significant. To control for overall self-evaluation, we partialed out selfperceptions on the other two ideal dimensions in all analyses. Thus, any effects found are domain specific (e.g., women’s warmth/trustworthiness partner regulation influences men’s perceptions of these particular qualities and not attractiveness/vitality or status/resources self-attributes). In addition, we pooled the paths across gender. For the warmth/trustworthiness and attractiveness/ vitality dimensions, there were no differences in the paths across 2 gender, ␹LM (1, N ⫽ 62) ⫽ 0.06 to 0.64, ps ⬎ .05 (see Footnote 12). However, the cross-paths in the status/resources model were 2 significantly different, ␹LM (1, N ⫽ 62) ⫽ 4.60, p ⬍ .05, and were therefore left unconstrained. The cross-path coefficients for all three ideal dimensions are presented in the first column of Table 5 (see top half of Table). As predicted, for both women and men, the more individuals attempted to change the warmth/trustworthiness of their partner, the more negative were their partners’ self-perceptions on this dimension. However, the cross-partner paths for attractiveness/vitality and status/resources were not significant. Next, we examined the association between partner’s regulation attempts and inferred ideal–perception consistency (i.e., the extent to which the individual perceives the self as matching the ideal standards held by the partner). Again, these analyses controlled for inferred ideal consistency across ideal dimensions to rule out the effects of global evaluation. In addition, because judgments re-

Table 5 Structural Equation Modeling Coefficients for Paths From Partner’s Regulation Attempts to Self-Perceptions, Inferred Ideal–Perception Consistency, and Self-Regulation Attempts (Study 2a) Partner’s regulation attempts and perceptions of partner’s regulation attempts Partner’s regulation attempts Warmth/trustworthiness Vitality/attractiveness Status/resources Perceptions of partner’s regulation attempts Warmth/trustworthiness Vitality/attractiveness Status/resources

Self-perceptions

Inferred ideal–perception consistency

Self-regulation

Women

Men

Women

Men

Women

Men

⫺.20* .03 .19

⫺.18* .03 ⫺.19

⫺.02 ⫺.21* ⫺.10

⫺.02 ⫺.15* ⫺.09

.21* .29* .04

.23* .01 .24*

⫺.39* ⫺.08 ⫺.14

⫺.46* ⫺.09 ⫺.16

⫺.24* ⫺.16* ⫺.20*

⫺.28* ⫺.20* ⫺.20*

.34* .35* .31*

.37* .30* .37*

Note. Coefficients in the top half of the table represent the association between partners’ regulation attempts (e.g., women’s partner regulation attempts) and self-perceptions (e.g., men’s self-perceptions), inferred ideal–perception consistency (e.g., men’s inferences regarding the extent to which they match their female partner’s ideal standards), and self-regulation (e.g., men’s self-directed regulatory efforts). Coefficients in the bottom half of the table represent the equivalent analyses using perceptions of partner’s regulation attempts (e.g., men’s perceptions of their female partner’s regulation attempts) as the predictor variable. The coefficients for self-perceptions control for self-perceptions across the two other ideal dimensions. The coefficients for inferred ideal–perception consistency control for inferred ideal– consistency across the two other ideal dimensions as well as self-perceptions on the corresponding dimension. The coefficients for self-regulation control for self-regulation attempts across the two other ideal dimensions. All paths are pooled across men and women (see Footnote 12), except those which were significantly different across gender (shown in italics). * p ⬍ .05.

678

OVERALL, FLETCHER, AND SIMPSON

variables—such as the extent to which individuals matched their own ideal standards (self-ideal–perception consistency), global levels of self-esteem, and partner perceptions and partner ideal– perception consistency—95% of the effects remained significant across analyses. The exact pattern of findings, however, was inconsistent across ideal dimensions and dependent measures. As noted previously, prior research has demonstrated that perceptions of partners’ behavior have important and perhaps more direct effects on relationship judgments beyond those predicted by the behavior reported by the partner (e.g., Bolger et al., 2000; Gable et al., 2003). Thus, we (conceptually) reexamined the above partner effects, but this time we tested the impact of perceptions of partner’s regulation on self-judgments and self-regulation at the within-individual level.

Perceptions of Partner’s Regulation The correlations between participants’ perceptions of their partner’s regulation attempts (i.e., ratings of the degree to which individuals had received regulation from their partners) and their partner’s reported regulation attempts were positive across all three ideal dimensions for both men and women (rs ⫽ .07 to .33, average r ⫽ .20), indicating that participants were tracking, to some extent, their partner’s regulation efforts. Nevertheless, the size of these correlations suggests that some discrepancy existed between the perceptions of regulation behavior across partners. We repeated the above set of analyses testing partner effects, but using perceptions of partner’s regulation attempts as the predictor variable. As before, all paths were calculated while controlling for global evaluation (e.g., partialing out self-perceptions on the other two ideal dimensions), and all paths were pooled across gender 2 (with no gender differences evident across analyses) ␹LM (1, N ⫽ 62) ⫽ 0.01 to 2.93, ps ⬎ .05. The SEM coefficients for perceptions of the partner’s regulation predicting (a) self-perceptions, (b) inferred ideal–perception consistency, and (c) self-regulation are shown in the bottom half of Table 5. First, as before, for both men and women, perceptions of partner’s regulation reduced how positively individuals perceived their own warmth/trustworthiness attributes, but there were null findings for the remaining two dimensions. Second, in contrast, across all three ideal dimensions, perceiving that the partner was attempting to change self was associated with more negative inferences regarding the extent to which self matched the partner’s ideal standards. And, third, for both men and women, across ideal dimensions, the more individuals perceived their partner as attempting to change them over the past 6 months, the more they reported trying to change these same characteristics in themselves. Moreover, all paths remained significant when partialing out individuals’ own self-ideal–perception consistency, self-esteem, partner perceptions, and partner ideal–perception consistency both within and across partners. Thus, perceiving that the partner had tried to change self over the past 6 months was associated with (a) more negative (warmth/trustworthiness) self-perceptions, (b) more negative evaluations regarding how closely self matches the ideal standards held by the partner, and (c) an increase in self-regulatory efforts, regardless of individuals’ own self- or partner evaluations or the evaluations made by their partner. Finally, we examined whether perceptions of partner’s regulation attempts and inferred ideal–perception consistency ratings

were associated with relationship quality. As expected, perceiving stronger regulation attempts from the partner was associated with lower perceptions of relationship quality across ideal dimensions (␤s ⫽ ⫺.21 to ⫺.41, ps ⬍ .05), as was lower inferred ideal– perception consistency (␤s ⫽ .18 to .51, ps ⬍ .05, with the exception of men’s warmth/trustworthiness inferred ideal consistency, ␤ ⫽ ⫺.04), and these effects remained significant when controlling for partner regulation attempts. Thus, the partner effects demonstrated above do have negative implications for relationship evaluations (also see cross-partner paths in Figure 5) consistent with previous research (e.g., Murray et al., 2000, 2001).

Discussion The results of Study 2a replicated the central findings of Study 1. For all three ideal dimensions, and for both men and women, more strenuous attempts to change the partner were associated with lower partner ideal–perception consistency. Moreover, these links were specific to particular ideal dimensions and were not a function of global evaluations, perceptions of the partner, or ideal standards. The results of Study 2a also provided further support for the models shown in Figure 1. Although the power to find interaction effects was fairly low in this study because of relatively small sample sizes, we still found evidence that less successful regulation attempts reduced the consistency between ideal standards and partner perceptions (and vice versa; see Model 2). Moreover, the mediation model (Model 3) received solid support across all three ideal dimensions, with attempts to change the partner during the previous 6 months predicting lower partner ideal–perception consistency, which in turn predicted lower perceived relationship quality. One novel objective of Study 2a was to determine how regulation, ideal–perception consistency, and relationship satisfaction were related between relationship partners. As predicted, we found that relationship quality was not only a function of how people viewed their partners but also influenced by how their partners viewed them. Moreover, regulation received from the partner (both as reported by the partner and as perceived by the target) had important consequences for self-perceptions and behavior. First, stronger regulation received from the partner was related to more negative self-perceptions, but only for warmth/trustworthiness attributes. Given the importance of warmth/trustworthiness characteristics in close relationships, and the deeply interpersonal quality of these attributes, it is perhaps not surprising that partner regulation exerted a direct effect on targets’ self-perceptions of these specific qualities. Attractiveness/vitality and status/resources attributes, on the other hand, tend to be more objective. Accordingly, self-perceptions on these dimensions may be less vulnerable to partners’ expectations and regulation attempts. Second, although partner’s regulation did not influence selfperceptions on the other two dimensions, the associations across partner’s regulation and inferred ideal–perception consistency suggested that individuals did not ignore the regulation attempts of their partners. Specifically, we found evidence that stronger regulation received from the partner was related to more negative perceptions of how closely individuals believed they matched their partner’s standards. Third, the associations between partner’s regulation and individuals’ self-regulation across dimensions suggested that

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS

participants often responded to their partner’s regulation attempts by increasing self-regulatory efforts to change the targeted characteristics. Taken together, these data provide good preliminary evidence that partner regulation generally has detrimental impacts on relationship well-being because such attempts produce more negative perceptions and relationship evaluations for both partners. A major limitation of both Study 1 and Study 2a is that they used cross-sectional designs. For example, although the results in the prior studies supported the hypothesized links between regulation, ideal–perception consistency, and relationship quality, we were unable to test the possibility of the reverse mediation chain (i.e., regulation mediating the association between ideal consistency and relationship quality). A longitudinal component of Study 2 (Study 2b) was designed to remedy these limitations by investigating how ideal–perception consistency and regulation attempts influence each other and judgments of relationship quality over time (see Figure 1, Model 4).

Study 2b Couples who had participated in Study 2a reported on their (a) partner regulation, (b) partner ideal–perception consistency, and (c) perceived relationship quality in a 6-month follow-up telephone interview. Assessing both ideal–perception consistency and regulation attempts on two separate occasions permitted a crosslagged design, which allowed us to test the extent to which ideal consistency and regulation might influence each other across time. In Study 1 and Study 2a, we provided evidence that regulation attempts during the past 6 months appear to influence current perceptions of ideal consistency (see Figure 1, Model 1). We expected to replicate the same finding examining these variables over time (see Figure 1, Model 4, Path a). Another crucial component of the Ideal Standards Model, however, is the proposition that ideal–perception consistency should motivate regulation attempts (see Figure 1, Model 4, Path b). Accordingly, we had two main predictions. First, lower partner ideal–perception consistency at Time 1 should predict greater partner regulation attempts at Time 2. Second, greater regulation attempts at Time 1 should predict lower partner ideal–perception consistency at Time 2. Finally, we also examined how both ideal–perception consistency and regulation influenced changes in relationship quality over time. We expected that reductions in ideal–perception consistency would predict more negative perceived relationship quality at Time 2 (controlling for relationship quality assessed at Time 1). However, consistent with the mediation model supported in the previous studies (see Figure 1, Model 3), we did not expect direct links between changes in regulation and relationship quality (when controlling for ideal–perception consistency). This finding would support our contention that the primary direct outcome of regulation attempts is modifying the consistency between perceptions and ideal standards, which in turn influences judgments of relationship quality.

Method Participants Of the 62 couples who participated in Study 2a, 51 reported on their partner ideal–perception consistency and partner regulation 6 months after

679

their initial testing session. Of those who did not participate, 9 couples had broken up and 2 chose not to participate.

Telephone Follow-Up at 6 Months To have a more efficient and practical questionnaire for the telephone interview, all participants completed a short version of the partner ideal– perception consistency scale and the partner regulation questionnaire. Two items from each of the mate ideal dimensions were included for each measure. The items were “understanding” and “supportive” for the warmth/trustworthiness dimension, “attractive appearance” and “good lover” for the attractiveness/vitality dimension, and “successful” and “financially secure” for the status/resources dimension. Participants rated each item on 7-point scales. For the partner ideal–perception consistency scales, participants rated each attribute in terms of the extent to which their partner matched their ideal (1 ⫽ does not match my ideal at all, 7 ⫽ completely matches my ideal). For the partner regulation questionnaires, participants rated the extent to which they had tried in some way to change that aspect of their partner during the past 6 months (1 ⫽ not tried at all to change, 7 ⫽ tried hard to change). We then computed equivalent (two item) measures for each construct for both Time 1 and Time 2. Table 6 presents descriptive statistics and reliability indexes. The means for each scale were similar to those reported in the previous studies, which used the full set of items (see Tables 1 and 4). For each measure assessed at both time periods, the two items tapping each dimension correlated positively and at adequate levels (average r ⫽ .55 at Time 1 and .52 at Time 2). Hence, the items from each dimension were summed and averaged to provide single indexes for each dimension. The final column of Table 6 shows the within-subject longitudinal correlations. For both men and women across all three ideal dimensions, there generally was good consistency for each measure across time (average rs ⫽ .55 for ideal–perception consistency and .49 for regulation). Thus, the short versions of our scales were reasonably reliable. Participants also completed the seven-item version of the PRQC Inventory (Fletcher et al., 2000b) to assess perceived relationship quality (see Study 1). The means for each scale were similar to those reported previously (M ⫽ 6.13, SD ⫽ 0.78, and M ⫽ 6.04, SD ⫽ 0.70, for women and men, respectively), and the scale had good internal reliability (Cronbach’s ␣s ⫽ .90 and .88, respectively).

Procedure Both members of each couple were phoned 6 months after their initial testing session, and each partner verbally responded to the follow-up questionnaires described above (as well as several additional questions not germane to the current study). Participants completed the scales in the same order they did in the laboratory. All participants initially completed the relationship quality measure, followed by the ideal consistency and regulation questionnaires (with half of the couples responding to the ideal consistency scales first and half responding to the regulation scales first). At the completion of the interview, couples were entered into a cash draw for $75.

Results Cross-Lagged Analyses To analyze the cross-lagged relations, we again used an SEM approach. The design strategy is illustrated in Figure 6. First, we examined the cross-time associations between partner ideal– perception consistency and partner regulation. We set the longitudinal within-subject paths (e.g., paths running from Time 1 ideal consistency to Time 2 ideal consistency) to be equal within gender. This was done to ensure that any differences in the cross-lagged paths were not a function of differential reliabilities across mea-

OVERALL, FLETCHER, AND SIMPSON

680

Table 6 Means, Standard Deviations, and Correlations of Partner Ideal–Perception Consistency and Partner Regulation (Short) Scales at Times 1 and 2 (Study 2b) Time 1 Women Partner ideal–perception consistency and partner regulation Partner ideal–perception consistency Warmth/trustworthiness Attractiveness/vitality Status/resources Partner regulation Warmth/trustworthiness Attractiveness/vitality Status/resources

Time 2

Men

Women Women Men r r

M

SD

M

SD

6.09 6.14 5.79

1.07 0.81 1.14

6.14 5.53 5.89

0.83 1.14 0.95

.64 .46 .71

2.38 2.25 2.78

1.31 1.53 1.71

2.33 2.67 2.45

1.38 1.49 1.56

.45 .44 .49

Across Times 1 and 2

Men Women Men r r

M

SD

M

SD

.62 .59 .55

5.83 6.06 4.90

0.80 0.79 1.01

5.86 5.94 6.07

0.86 0.89 0.92

.40 .38 .57

.54 .51 .64

2.70 1.93 2.22

1.60 1.26 1.44

2.69 2.30 2.42

1.41 1.16 1.33

.57 .37 .52

Women r

Men r

.58 .78 .65

.66 .58 .57

.51 .46 .50

.66 .37 .34

.20a .64 .49

.68 .33 .58

Note. Both Time 1 and Time 2 measures comprise the equivalent two-item scales described in the Method section. Correlations at Times 1 and 2 are the correlations between the two ratings for each dimension. Correlations across Times 1 and 2 are the within-subject longitudinal correlations (i.e., measure at Time 1 correlated with the equivalent measure at Time 2). a p ⬎ .05. All other correlations are significant at p ⬍ .05.

sures (which can be a problem with cross-lagged analyses). LM tests indicated that, for all analyses, there were no differences in 2 the within-subject longitudinal paths, ␹LM (1, N ⫽ 62) ⫽ 0.02 to 2.57, ps ⬎ .05. To test for gender differences, we pooled the diagonal paths across women and men (i.e., each cross-lagged path for women was set equal to the equivalent path for men). None of the cross-lagged paths were significantly different across gender, 2 ␹LM (1, N ⫽ 62) ⫽ 0.02 to 1.54, ps ⬎ .05. Thus, these paths were left pooled (see Footnote 12). Over time, we predicted that stronger attempts to regulate the partner would reduce the consistency between perceptions and ideal standards and that lower ideal–perception consistency would motivate

stronger regulation attempts (see Figure 1, Model 4, Paths a and b). As shown in Table 7, solid evidence was found for both predictions. Strikingly, all of the cross-lagged paths were negative, and 8 of the 12 paths were significant. To test whether the cross-lagged results were simply a function of general levels of positivity, we recalculated all the cross-lagged analyses, controlling for relationship length, selfesteem, and judgments of relationship quality assessed at Time 1. The results were unchanged, with all significant cross-lagged paths remaining significant. Finally, as we had done previously, we controlled for partner perceptions and ideal standards assessed at Time 1, and, as before, the size of the significant paths in Table 7 were typically not reduced (␤s ⫽ ⫺.11 to ⫺.38).

Time 1

Time 2

Women’s Partner Ideal-Consistency

Women’s Partner Ideal-Consistency e e

Women’s Partner Regulation

Women’s Partner Regulation

Men’s Partner Ideal-Consistency

Men’s Partner Ideal-Consistency e e

Men’s Partner Regulation

Women’s Perceived Relationship Quality

e

Men’s Perceived Relationship Quality

e

Men’s Partner Regulation

Figure 6. Structural equation model (SEM) testing the associations between partner regulation, partner ideal–perception consistency, and relationship quality over a 6-month period, Study 2b.

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS

Table 7 SEM Coefficients From Cross-Lagged Analyses Testing the Associations Between Partner Regulation and Ideal–Perception Consistency Across Time (Study 2b) Partner regulation as predictor variable Partner ideal– perception consistency Warmth/trustworthiness Attractiveness/vitality Status/resources

Partner regulation as dependent variable

Women

Men

Women

Men

⫺.21* ⫺.17 ⫺.19*

⫺.19* ⫺.14 ⫺.20*

⫺.13 ⫺.23* ⫺.21*

⫺.13 ⫺.35* ⫺.20*

Note. All paths are pooled across gender (see Footnote 12). * p ⬍ .05.

Partner Regulation, Ideal–Perception Consistency, and Relationship Quality The paths in Figure 6 running from Time 2 ideal–perception consistency and regulation attempts to relationship quality (assessed at Time 2) also allowed us to examine whether regulation attempts and partner ideal–perception consistency influenced later judgments of relationship quality. These paths were calculated while controlling for relationship quality measured at Time 1, and the resulting coefficients are displayed in Table 8. Across dimensions, strong positive associations were found between ideal– perception consistency and relationship quality. In contrast, partner regulation attempts were not associated with perceptions of relationship quality, with one exception: Attractiveness/vitality partner regulation was positively, although less strongly, associated with changes in relationship quality. In addition, the significant paths between ideal–perception consistency and relationship quality remained strong and significant when controlling for partner perception ratings (␤s ⫽ .30 to .63) and partner ideal standards (␤s ⫽ .32 to .60) gathered at Time 1. Thus, reductions in ideal consistency across time are associated with decreasing levels of perceived relationship quality.15 These findings confirm our contention that the pivotal proximal determinant of relationship evaluation is the consistency across actual perceptions and ideal standards (as opposed to regulation attempts), and they are consistent with the previously reported mediation models suggesting that regulation feeds back into ideal– perception consistency, which in turn influences judgments of relationship quality.

Discussion These results provide evidence for the bidirectional nature of the connections between ideal–perception consistency and regulation (see Model 4). As expected, we found that (a) greater attempts to change the partner predicted reduced judgments of partner ideal– perception consistency across time and (b) lower perceived consistency between partner perceptions and ideal standards motivated more strenuous regulation attempts over time. Moreover, when controlling for the current and longitudinal associations across ideal–perception consistency and regulation, reductions in ideal consistency predicted more negative relationship evaluations across time.

681

Studies 1 and 2a used cross-sectional samples to test models of how regulation attempts during the previous 6 months are related to current perceptions of ideal consistency and, in turn, relationship evaluation. This extension replicated these findings using a longitudinal design and supported the notion that the impact of partner regulation on relationship quality is mediated via the effects that regulation has on ideal–perception consistency. Moreover, the cross-lagged results support a key prediction of the Ideal Standards Model that partner regulation should be motivated by low ideal–perception consistency.

General Discussion Extending the Ideal Standards Model (Simpson et al., 2001), the current research tested several novel predictions regarding regulation processes in intimate relationships. As predicted, more strenuous partner regulation attempts were associated with lower consistency between partner perceptions and ideal standards (above and beyond either ideal standards or partner perceptions operating on their own). In addition, across all samples, these links occurred within specific mate-evaluation dimensions (warmth/trustworthiness, attractiveness/vitality, and status/resources) and were not driven by global evaluation biases. General support was also found for four causal models regarding the consequences partner regulation attempts have for partner and relationship evaluations (see Figure 1). First, we proposed and found good evidence that greater attempts to change the partner reduced how closely the partner was perceived to match ideal standards (see Model 1 and Path a, Model 4). That regulation attempts generally produce more negative perceptions of ideal consistency may seem counterintuitive, particularly given that the aim of regulation attempts is presumably the exact opposite. However, attempts to change the partner will be powerful signals that he or she is failing to meet expectations. Moreover, regulation attempts may increase ideal–perception consistency, but only if such efforts are successful in bringing about change, whereas unsuccessful regulation attempts seem more likely to increase the salience and the psychological significance of the discrepancy. The low levels of perceived regulation success reported within these studies showed that participants were generally only able to produce small perceived partner improvements (mean levels ranged from 2.14 to 2.86 where 1 represents lack of success and 7 equals success; see Tables 1 and 4). Nevertheless, in support of Model 2 (see Figure 1), more successful regulation attempts were associated with relatively higher ideal–perception consistency. Those participants who had engaged in more intensive regulation and failed, however, had the lowest perceptions of partner ideal consistency. Thus, high levels of partner regulation were likely to exacerbate low ideal consistency, particularly if regulatory efforts were unsuccessful. Third, we found strong evidence that the negative links between regulation and ideal–perception consistency filter through to judgments of relationship quality. Both cross-sectional and longitudinal 15 We also tested for partner effects by running paths from ideal– perception consistency and regulation at Time 2 to partners’ relationship quality. Consistent with the cross-sectional analyses reported in Figure 5, all paths from ideal–perception consistency to partners’ relationship quality ratings were positive (␤s ⫽ .11 to .34), but were generally not significant.

OVERALL, FLETCHER, AND SIMPSON

682

Table 8 SEM Coefficients for Paths From Ideal–Perception Consistency and Partner Regulation to Relationship Quality (Time 2; Study 2b) Perceived relationship quality Partner ideal–perception consistency and partner regulation Partner ideal–perception consistency Warmth/trustworthiness Attractiveness/vitality Status/resources Partner regulation Warmth/trustworthiness Attractiveness/vitality Status/resources

Women

Men

.38* .52* .32*

.57* .61* .03

.05 .20* .10

.05 .21* .10

Note. Coefficients were calculated controlling for relationship quality assessed at Time 1. All paths were pooled across gender (except for those shown in italics). There were generally no differences in the paths across 2 gender, ␹LM (1, N ⫽ 62) ⫽ 0.01 to 3.65, ps ⬎ .05 (see Footnote 12), with the exceptions marked in italics; only women’s status/resources partner ideal–perception consistency was significantly associated with relationship 2 quality, ␹LM (1, N ⫽ 62) ⫽ 5.05, p ⬍ .05. These paths were left unconstrained. * p ⬍ .05.

analyses demonstrated that more strenuous partner regulation attempts reduced perceptions of ideal consistency, which, in turn, fed into more negative relationship evaluations (see Figure 1, Model 3 and Model 4). These effects are consistent with the hypothesized feedback-loop characteristic of regulation processes; namely, attempts to change the partner impact on relationship quality to the extent that regulation influences the consistency between perceptions and ideal standards. Finally, consistent with prior accounts of self-regulation that propose regulation attempts should be produced by discrepancies between perceptions and ideals, longitudinal analyses illustrated that lower consistency between partner perceptions and ideal standards motivated an increase in partner regulation attempts (see Path b, Model 4). To summarize, we found that lower ideal consistency motivated attempts to improve the partner. However, in ironic contrast to the purpose of improvement attempts, partner regulation appears to be generally detrimental for the individual’s partner and relationship evaluations. Furthermore, we also identified a second route by which partner regulation attempts are likely to have deleterious consequences for the relationship. We discuss these partner effects next.

The Dyadic Nature of Relationship Regulation This research breaks new ground by examining how regulation and ideal consistency are related between partners within romantic relationships. Consistent with previous research (Campbell et al., 2001), for all three ideal dimensions we found that perceptions of relationship quality were associated not only with individuals’ own judgments of partner ideal consistency but also with how closely individuals matched the ideal standards of their partners. This important partner effect suggests that participants were sensitive to how they were evaluated by their romantic partners.

Moreover, extending prior research, Study 2a suggested that one central way in which individuals form judgments of how they are regarded by their partners is the degree to which they are targets of their partners’ regulatory efforts. With regard to warmth/trustworthiness characteristics, for example, the more individuals received regulation attempts from their partner during the previous 6 months, the less glowingly they evaluated themselves on this dimension. It is not surprising that regulation of this category of characteristics was negatively associated with individuals’ selfperceptions. Warmth/trustworthiness is consistently rated as more important than the other dimensions in long-term relationships by both men and women (see Buss, 1999, and Fletcher, 2002, for reviews), and these interpersonal characteristics are routinely expressed and entwined within day-to-day interactions between partners. Thus, people are likely to be particularly sensitive to both self and partner’s possession of warmth/trustworthiness qualities and the extent to which they pass muster on these attributes in the eyes of their partners. However, individuals did not disregard their partner’s regulation behavior on the other two dimensions. Instead, we found evidence that participants adjusted perceptions of their partner’s evaluations according to the amount of regulation received—the more regulation the less they believed they matched their partner’s standards. These findings suggest that partner regulation or improvement attempts convey crucial information regarding how individuals feel about their partners. Thus, regulation processes do not merely occur within individuals’ heads, but are tied to the objective reality of the relationship. This point is also illustrated by the relatively accurate judgments that men and women produced when evaluating their partners’ ideal standards. Correlations between inferred ideal–perception consistency (i.e., ratings of the degree to which individuals believed they matched the ideal standards of their partner) and partners’ actual partner ideal-consistency ratings across all three ideal dimensions were all positive and typically significant (rs ⫽ .15 to .51, average r ⫽ .30). Prior research has shown that perceiving the partner as holding relatively negative judgments about self (such as low ideal– perception consistency) is strongly associated with more negative partner perceptions and lower relationship satisfaction (e.g., Murray et al., 2000, 2001). Thus, partner improvement attempts translate into negative relationship outcomes partly because receiving regulation attempts from one’s partner communicates lack of acceptance and negative views of the self held by the partner. Such judgments exert powerful corrosive effects on perceptions of relationship quality. Other dyadic effects were consistent with prior suggestions that self-regulation may be motivated by perceived discrepancies between self-perceptions and either the goals and wishes of significant others (Moretti & Higgins, 1999) or individuals’ perceived relational value (Leary, 2004). When individuals received regulation attempts from their partner they were more likely to engage in regulatory efforts to improve their partner’s evaluation. These findings suggest that partner regulation attempts do lead, at least in part, to their intended effect by eliciting attempts from partner to change targeted self-attributes. However, we found some intriguing sex differences. Men’s self-regulatory efforts were associated with their partners’ reported regulation attempts for status/resources but not for attractiveness/vitality, whereas for women the pattern was the reverse (see Table 5). Previous research has shown

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS

that men typically attach higher importance to attractiveness/vitality partner attributes, and women attach greater value to status/ resource partner characteristics (Fletcher et al., 2004; a finding replicated in this research; see Tables 1 and 4). Thus, our results again suggest that individuals’ motivation to self-regulate is sensitive to the importance placed on these dimensions by their partners. Because both self- and partner regulation were reported for the same time period, we were unable to establish whether the receipt of regulation from partner motivates self-regulation efforts or attempts to change self simply involves recruiting the partner to help (or both). Future research could profitably examine these connections over time to throw light on the causal links between partners’ regulation attempts, self- and relationship perceptions, and self-regulatory behavior. Finally, the impact of partners’ regulation attempts was examined using both partner-reported regulation attempts and targets’ perceptions of that regulation behavior. There was reasonable consistency in the findings across both of these measures; however, perceptions of partner’s regulation attained the strongest links with inferred ideal–perception consistency and reports of self-regulation attempts. This finding is consistent with one of the most well-researched and robust findings within the intimate relationship domain; that is, beyond the actual behavior exhibited, how partners perceive and attribute cause to relationship events and interactions powerfully determines how individuals think, feel, and behave within their relationships (see Fincham, 2001).

Implications and Novel Contributions This research extends prior regulation theories and research in several critical ways. First, consistent with what is already known about the motivating influence of goals, these findings support our hypothesis that important goals or standards are prime determinants of motivating and regulating other individuals in interpersonal settings. Thus, the basic processes traditionally identified as underlying self-regulation also appear to explain the regulation of intimate relationship partners. This extension is important given that intimate relationships have a substantial impact on personal well-being and should be a central domain in which regulation processes are played out. Second, the specific content of the standards that drive regulation has been relatively neglected within the regulation literature. An important contribution of the Ideal Standards Model is the identification of three major dimensions that individuals use to evaluate and regulate their intimate partner: warmth/trustworthiness, attractiveness/vitality, and status/resources. Past research has focused on how these dimensions influence partner and relationship evaluation. The current studies show how these ideal dimensions are also implicated in regulation processes. Moreover, to recap, one important result is the indication that regulation processes work through each of the three ideal constructs rather than being driven by global partner or relationship judgments. Third, many regulation theories (explicitly or implicitly) conceptualize regulation processes in terms of circular feedback loops, in which current states (i.e., perceptions) are compared with a reference value of some kind (i.e., ideal standards or goals), the effects of resulting regulation are monitored, and these judgments, in turn, influence levels of ideal–perception consistency (e.g.,

683

Carver & Scheier, 1998). Inherent in these accounts is the fundamental role that regulation success should have in moderating the feedback loop. The present research clearly demonstrated the bidirectional associations across ideal consistency and regulation in intimate relationships, and it (at least in part) confirmed the moderating effect of perceived regulation success. Fourth, we were able to rule out the possibility that evaluative perceptions of the partner (vs. ideal–perception consistency) were driving these regulation processes by statistically controlling for partner perceptions when calculating the associations between regulation attempts and ideal consistency. Indeed, when pitting straight perceptions of the partner directly against ideal–perception consistency ratings, ideal consistency continued to predict regulation, whereas straightforward perceptions typically did not. In addition, although (as expected) ideal importance was positively associated with partner regulation, ideal–perception consistency was a stronger and more robust predictor of regulation in all cases. These results support a fundamental principle of the Ideal Standards Model (and prior regulation theories) that ideal–perception consistency is a key proximal variable in terms of both relationship evaluation and regulation processes. Finally, as already noted, our results suggest that regulation efforts tend to backfire, and people become even unhappier with their relationship. This raises a question about the functions of the relationship monitoring and regulation system. From a distal evolutionary approach, the functions of an adaptation or behavior are defined in terms of the costs and benefits vis-a`-vis reproductive fitness and do not necessarily equate to increased happiness. For example, one reason why humans may have evolved the relationship monitoring and regulation system is to loosen the powerful bonds of love and attachment when standards are not being met, thus enabling individuals to look elsewhere for a new partner and relationship. Alternatively, perhaps a principal function of our ancestral relationship monitoring and regulation system is indeed to improve relationships, but it fails because the contemporary social and cultural environment has changed so that it no longer fits the ancestral environment within which the relevant adaptations developed. For example, maybe contemporary Western cultures, with thousands of accessible partners apparently a mouseclick away, barrages of self-help books and TV shows about relationships and how to make them better, constant images of attractive alternatives, and people apparently having great sex everywhere have ramped up people’s expectations and standards to the extent that the relationship monitoring and regulation system has been put into overdrive and, thus, has become relatively dysfunctional in the modern environment. These speculations remain to be tested.

Limitations and Future Directions First, across studies, the results demonstrated that partner improvement attempts generally have negative implications for partner and relationship evaluations. However, these studies provided little information regarding how individuals actually go about regulating their partners or the differential effects that particular regulation tactics may have on relationship outcomes (e.g., direct confrontational harangues vs. kind and supportive suggestions). The way in which regulation is enacted, and the way in which particular strategies are likely to be perceived, should have pow-

OVERALL, FLETCHER, AND SIMPSON

684

erful moderating effects on the extent to which partner regulation reduces ideal–perception consistency and relationship quality. This might be one reason why individual’s perceptions of their partners’ regulation are so important in determining the consequences of improvement attempts. We might expect, for example, that more indirect regulation attempts that are not noticed or perceived by the partner or that are not interpreted as attempts to change may be the most innocuous and perhaps effective regulation attempts (also see Bolger et al., 2000). Similarly, regulation attempts that focus on the relationship rather than the individual might also be more effective. Second, although several studies have examined influence tactics within intimate relationships (e.g., Bui et al., 1994; Falbo & Peplau, 1980; Howard et al., 1986; Orin˜a et al., 2002), prior studies have not examined (a) whether regulation strategies produce change over time, (b) which strategies are most likely to be successful in bringing about change, or (c) the long-term consequences of successful versus unsuccessful regulation attempts. An examination of these processes should help in determining when partner regulation attempts successfully accomplish the goal of relationship improvement versus jeopardize relationship satisfaction and stability. Third, the tendency to engage in regulation tactics should depend on the history of regulation successes and failures. Individuals who have been successful in their regulation attempts might be more likely to engage in future efforts to change their partners. People may also develop a general sense of efficacy that either promotes or hinders future partner regulation attempts (Bandura, 1992) and form attributions about the changeability and controllability of specific partner characteristics (e.g., Ruvolo & Rotondo, 1998; see also Fincham, 2001). Similarly, general implicit relationship theories, such as believing that relationships grow and develop through efforts to maintain and improve them, may influence the salience of ideal discrepancies and the likelihood of engaging in regulation tactics (see Knee et al., 2001). Other intrapersonal factors are also likely to moderate how regulation attempts are received. For example, individuals with low selfesteem or who are anxious regarding their partner’s love and acceptance may be especially vulnerable to a drop in selfevaluation and relationship satisfaction when subject to their partner’s regulation attempts (Murray et al., 2000, 2001). Finally, when large discrepancies between perceptions and ideal standards on important dimensions simply cannot be diminished, individuals may either resign themselves to enduring the problem behavior or decide to dissolve the relationship. We were unable to examine the connection between regulation processes and relationship dissolution in the current research because only a small percentage of couples ended their relationship across 6 months (only 9 out of 62 couples disbanded; Study 2b). Nevertheless, we suspect that the negative relationship implications of low ideal consistency, paired with high, but unsuccessful, regulation attempts, should increase the probability of relationship dissolution.

Conclusions These studies investigated relationship regulation attempts directed toward specific characteristics of the partner. However, attempts to improve the relationship should also include attempts to change self-attributes and/or aspects of the relationship that

involve both self and partner (e.g., intimacy). Across both studies we did measure the relations between self-ideal–perception consistency (ratings of how closely individuals matched their own ideal standards) and self-regulation. Although we have not reported these results here, consistent with prior theory and research we found that lower consistency between self-perceptions and ideal standards was typically associated with greater attempts to change the self within the three ideal dimensions (e.g., Carver & Scheier, 1998; Higgins, 1987). However, own judgments of selfideal–perception consistency and self-regulation were not related to perceptions of relationship quality. Thus, it is not the experience of general ideal inconsistencies within the relationship that produces negative relationship evaluations, but the extent to which the partner is not living up to ideals (and associated partner regulation attempts). These findings are also consistent with previous research illustrating that partner judgments play a more powerful role in predicting relationship satisfaction than self-judgments (e.g., Fletcher & Fincham, 1991; Fletcher & Thomas, 2000; Friesen, Fletcher, & Overall, 2005; Su¨mer & Cozzarelli, 2004). There remain many unanswered questions about relationship regulation, including the factors that govern when and how regulation occurs, the effectiveness of different regulation strategies and tactics, and their long-term consequences. The current research relied exclusively on self-reports and partner reports, which raises the question of whether similar results might emerge when other methods (including more behavioral ones) are used. This limitation notwithstanding, the current research has several strengths. They include the systematic replication of results across different studies and different measurement strategies, statistically controlling for several potential artifacts, and utilizing both crosssectional and longitudinal designs. Understanding when, how, and why individuals try to change their partners or relationships and the consequences of such regulation attempts are important topics for the science of relationships. The current studies provide some initial steps toward answering these questions and affirm the important role that ideal standards play in regulation processes as they unfold in romantic relationships.

References Bandura, A. (1992). Exercise of personal agency through the self-efficacy mechanism. In R. Schwarzer (Ed.), Self-efficacy: Thought control of action (pp. 3–38). Washington, DC: Hemisphere. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117, 497–529. Bem, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 6, pp 1– 62). New York: Academic Press. Bentler, P. M. (1995). EQS structural equations program manual. Encino, CA: Multivariate Software. Bolger, N., Zuckerman, A., & Kessler, R. C. (2000). Invisible support and adjustment to stress. Journal of Personality and Social Psychology, 79, 933–961. Bui, K. T., Raven, B. H., & Schwarzwald, J. (1994). Influence strategies in

REGULATION PROCESSES IN INTIMATE RELATIONSHIPS dating relationships: The effects of relationship satisfaction, gender, and perspective. Journal of Social Behavior and Personality, 9, 429 – 442. Buss, D. M. (1999). Evolutionary psychology: The new science of the mind. Boston: Allyn and Bacon. Campbell, L., Simpson, J. A., Kashy, D. A., & Fletcher, G. J. O. (2001). Ideal standards, the self, and flexibility of ideals in close relationships. Personality and Social Psychology Bulletin, 27, 447– 462. Carver, C. S., & Scheier, M. F. (1998). On the self-regulation of behavior. New York: Cambridge University Press. Falbo, T., & Peplau, L. A. (1980). Power strategies in intimate relationships. Journal of Personality and Social Psychology, 38, 618 – 628. Fincham, F. D. (2001). Attributions in close relationships: From balkanisation to integration. In G. J. O. Fletcher & M. S. Clark (Eds.), Blackwell handbook of social psychology: Interpersonal processes (pp. 3–31). Oxford, England: Blackwell. Fletcher, G. J. O. (2002). The new science of intimate relationships. Cambridge, England: Blackwell. Fletcher, G. J. O., & Fincham, F. D. (1991). Attribution processes in close relationships. In G. J. O. Fletcher & F. D. Fincham (Eds.), Cognition in close relationships (pp. 7–35). Hillsdale, NJ: Erlbaum. Fletcher, G. J. O., Simpson, J. A., & Thomas, G. (2000a). Ideals, perceptions, and evaluations in early relationship development. Journal of Personality and Social Psychology, 79, 933–940. Fletcher, G. J. O., Simpson, J. A., & Thomas, G. (2000b). The measurement of perceived relationship quality components: A confirmatory factor analytic approach. Personality and Social Psychology Bulletin, 26, 340 –354. Fletcher, G. J. O., Simpson, J. A., Thomas, G., & Giles, T. (1999). Ideals in intimate relationships. Journal of Personality and Social Psychology, 76, 72– 89. Fletcher, G. J. O., & Thomas, G. (2000). Behavior and on-line cognition in marital interaction. Personal Relationships, 7, 111–130. Fletcher, G. J. O., Tither, J. M., O’Loughlin, C., Friesen, M. D., & Overall, N. C. (2004). Warm and homely or cold and beautiful? Sex differences in trading off traits in mate selection. Personality and Social Psychology Bulletin, 30, 659 – 672. Friesen, M. D., Fletcher, G. J. O., & Overall, N. C. (2005). A dyadic assessment of forgiveness in intimate relationships. Personal Relationships, 12, 61–77. Gable, S. L., Reis, H. T., & Downey, G. (2003). A quasi-signal detection analysis of daily interactions between close relationship partners. Psychological Science, 14, 100 –105. Gangestad, S. W., & Simpson, J. A. (2000). The evolution of human mating: Trade-offs and strategic pluralism. Behavioral and Brain Sciences, 23, 573– 644. Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94, 319 –340. Howard, J. A., Blumstein, P., & Schwartz, P. (1986). Sex, power and influence tactics in intimate relationships. Journal of Personality and Social Psychology, 51, 102–109. Knee, R. C., Nanayakkara, A., Vietor, N. A., Neighbors, C., & Patrick, H. (2001). Implicit theories of relationships: Who cares if romantic partners

685

are less than ideal? Personality and Social Psychology Bulletin, 27, 808 – 819. Leary, M. R. (2004). The sociometer, self-esteem, and the regulation of interpersonal behaviour. In R. F. Baumeister & K. D. Vohs (Eds.), Handbook of self-regulation: Research, theory, and applications (pp 373–391). New York: Guilford Press. Moretti, M. M., & Higgins, E. T. (1999). Own versus other standpoints in self-regulation: Developmental antecedents and functional consequences. Review of General Psychology, 3, 188 –223. Murray, S. L., & Holmes, J. G. (1999). The mental ties that bind: Cognitive structures that predict relationship resilience. Journal of Personality and Social Psychology, 77, 1228 –1244. Murray, S. L., Holmes, J. G., & Griffin, D. W. (1996). The self-fulfilling nature of positive illusions in romantic relationships: Love is not blind but prescient. Journal of Personality and Social Psychology, 71, 1155– 1180. Murray, S. L., Holmes, J. G., & Griffin, D. W. (2000). Self-esteem and the quest for felt security. How perceived regard regulates attachment processes. Journal of Personality and Social Psychology, 78, 478 – 498. Murray, S. L., Holmes, J. G., Griffin, D. W., Bellavia, G., & Rose, P. (2001). The mismeasure of love: How self-doubt contaminates relationship beliefs. Personality and Social Psychology Bulletin, 27, 423– 436. Murray, S. L., Rose, P., Holmes, J. G., Derrick, J., Podchaski, E. J., Bellavia, G., & Griffin, D. W. (2005). Putting the partner within reach: A dyadic perspective on felt security in close relationships. Journal of Personality and Social Psychology, 88, 327–347. Orin˜a, M. M., Wood, W., & Simpson, J. A. (2002). Strategies of influence in close relationships. Journal of Experimental Social Psychology, 38, 459 – 472. Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press. Ruvolo, A. P., & Rotondo, J. L. (1998). Diamonds in the rough: Implicit personality theories and views of partner and self. Personality and Social Psychology Bulletin, 24, 750 –758. Sanderson, C. A., & Cantor, N. (2001). The association of intimacy goals and marital satisfaction: A test of four mediational hypotheses. Personality and Social Psychology Bulletin, 27, 1567–1577. Sanderson, C. A., & Karetsky, K. H. (2002). Intimacy goals and strategies of conflict resolution in dating relationships: A mediational analysis. Journal of Social and Personal Relationships, 19, 317–337. Simpson, J. A., Fletcher, G. J. O., & Campbell, L. (2001). The structure and function of ideal standards in close relationships. In G. J. O. Fletcher & M. Clark (Eds.), Blackwell handbook of social psychology: Interpersonal processes (pp. 86 –106). Oxford, England: Blackwell. Su¨mer, N., & Cozzarelli, C. (2004). The impact of adult attachment on partner and self-attributions and relationship quality. Personal Relationships, 11, 355–371.

Received January 23, 2006 Revision received February 8, 2006 Accepted February 12, 2006 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 686 – 697

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.686

Procedural Justice and the Hedonic Principle: How Approach Versus Avoidance Motivation Influences the Psychology of Voice Jan-Willem van Prooijen

Johan C. Karremans

Free University Amsterdam

Radboud University Nijmegen

Ilja van Beest Leiden University The authors investigate the relation between the hedonic principle (people’s motivations to approach pleasure and to avoid pain) and procedural justice. They explore whether approach or avoidance motivation increases the effect that people feel they were treated more fairly following procedures that do versus do not allow them an opportunity to voice their opinion. Experiments 1 and 2 reveal that these procedures influence procedural justice judgments more strongly when people conduct approach motor action (arm flexion) than when they conduct avoidance motor action (arm extension). Experiment 3 indicates that individual-difference measures of participants’ approach motivations predicted procedural justice judgments following voice versus no-voice procedures. The authors conclude that people’s motivational orientations stimulate their fairness-based reactions to voice procedures. Keywords: procedural justice, voice, approach–avoidance, hedonic motivation

An illustration of a typical procedural justice phenomenon is the finding that people evaluate decision-making procedures that allow them an opportunity to voice their opinions to be more fair than procedures that do not allow them such an opportunity. This finding is referred to as the voice effect (Folger, 1977). Besides having a strong influence on procedural justice judgments, voice procedures have been found to exert positive effects on many of people’s other perceptions and behaviors. For example, voice procedures (when compared with no-voice procedures) have been found to increase positive affect, decrease negative affect, increase people’s willingness to accept decisions, improve relations with authorities, and improve task performance (Folger, Rosenfield, Grove, & Corkran, 1979; see also Brockner et al., 1998; Greenberg & Folger, 1983; Lind, Kanfer, & Earley, 1990; Tyler & Lind, 1992; Van den Bos, 2001, 2003; Van den Bos, Wilke, Lind, & Vermunt, 1998; Van Prooijen, Van den Bos, & Wilke, 2002, 2004a, 2005). People’s more positive responses to the granting as opposed to the denial of voice have been described as one of the most robust findings in procedural justice literature, and evidence for these voice effects has been found in both applied and experimental settings (e.g., Brockner et al., 1998; Lind et al., 1990; Tyler, 1987; Van den Bos & Van Prooijen, 2001). In many social situations, voice versus no-voice procedures precede decisions about positive outcomes (i.e., gains). In correspondence with this, a substantial number of procedural justice studies have investigated the voice effect in a gain-framed context. For example, recipients in numerous empirical procedural justice studies received voice or no-voice procedures about decisions regarding pay distributions (Folger, 1977; Van den Bos, 1999; Van den Bos & Van Prooijen, 2001; Van Prooijen et al., 2004a), lottery tickets (Van den Bos, 2001, 2003; Van den Bos et al., 1997, 1998; Van Prooijen et al., 2005), student grades (Tyler, Rasinski, & Spodick, 1985; Study 2), or goal-setting opportunities (Lind et al.,

Social justice is a key issue in understanding human behavior: People are influenced profoundly by the extent to which they perceive social situations as fair or unfair. For example, people display signs of appreciation when they believe that justice has been done, but acts of injustice lead people to show aversive reactions, such as feelings of anger, fear, and disgust (Folger & Cropanzano, 1998; Lind & Tyler, 1988; Tyler & Lind, 1992). It has even been suggested that social justice may be one of the most important norms and values in human society (Folger, 1984). A conceptualization of social justice that has been extensively studied by social psychologists is the extent to which people perceive decision-making procedures as fair, a conceptualization commonly referred to as procedural justice (Thibaut & Walker, 1975). Procedural justice has been found to influence a wide range of people’s perceptions and behaviors in various social situations (for overviews, see Brockner & Wiesenfeld, 1996; Cropanzano, Byrne, Bobocel, & Rupp, 2001; Folger & Cropanzano, 1998; Lind & Tyler, 1988; Tyler & Blader, 2000; Tyler & Lind, 1992; Van den Bos & Lind, 2002; Van Prooijen, Van den Bos, & Wilke, 2004b).

Jan-Willem Van Prooijen, Department of Social Psychology, Free University Amsterdam, Amsterdam, the Netherlands; Johan C. Karremans, Department of Social Psychology, Radboud University Nijmegen, Nijmegen, the Netherlands; Ilja van Beest, Department of Social and Organizational Psychology, Leiden University, Leiden, the Netherlands. We thank Jooke Veenstra for her assistance in collecting the data of Experiment 2 and Anouk Hofman for her assistance in collecting the data of Experiment 3. Furthermore, we thank Eric van Dijk for his helpful comments on a draft of this article. Correspondence concerning this article should be addressed to JanWillem Van Prooijen, Department of Social Psychology, Free University Amsterdam, Van der Boechorststraat 1, 1081 BT Amsterdam, the Netherlands. E-mail: [email protected] 686

PROCEDURAL JUSTICE AND THE HEDONIC PRINCIPLE

1990; Van Prooijen et al., 2004a). As a consequence, influential procedural justice theories are to a substantial extent based on gain-framed social situations (Thibaut & Walker, 1975; Tyler & Lind, 1992). Although we wish to emphasize that procedural justice phenomena are important in loss-framed situations as well (e.g., Brockner et al., 1998; Thibaut & Walker, 1975), these considerations suggest that gain-framed situations provide a good point of departure to study voice effects. Our focus on gains was partially inspired by early instrumental explanations of why people attach importance to voice procedures. These instrumental explanations were based on research findings that people prefer procedures that allow them a substantial amount of process control, that is, control over the manner in which decisions are made (Thibaut & Walker, 1975). According to instrumental explanations, one of the main reasons people desire process control is “because it enables them to obtain predictable and satisfactory outcomes” (Houlden, LaTour, Walker, & Thibaut, 1978, p. 16). As such, instrumental explanations suggest that people value voice procedures because they associate such procedures with obtaining positive outcomes. Besides instrumental explanations, in the 1980s procedural justice research indicated that people also value voice procedures for noninstrumental reasons (e.g., Lind et al., 1990; Tyler, 1987; Tyler et al., 1985). These noninstrumental concerns are highlighted in the relational model of authority (Tyler & Lind, 1992), which asserts that people value voice procedures because such procedures have positive implications for their sense of self-worth. More specifically, voice procedures are usually provided by group authorities, and these group authorities are generally perceived as representative for the entire group. An authority that uses fair procedures therefore communicates that recipients are respected members of their community and that they are included in social groups. Unfair procedures, however, communicate that recipients are disrespected by their community and that they are excluded from social groups (Lind, 2001; Lind & Tyler, 1988; Tyler, 1987, 1989; Tyler & Lind, 1992; Van Prooijen et al., 2004a, 2004b). The relational model thus proposes that people attach importance to voice procedures because they associate such procedures with obtaining important relational benefits, such as being respected and included in valuable social groups. Both instrumental and noninstrumental explanations of voice effects have contributed in important ways to scientists’ understanding of when people feel treated fairly or unfairly by decisionmaking authorities. Yet in the current article we suggest that both perspectives offer an incomplete account of the motivational nature of the voice effect: Do people value voice procedures because these procedures increase their chances of obtaining important instrumental or relational benefits, such as positive outcomes or a sense of inclusion (i.e., approach motivated)? Or do people value voice procedures because these procedures imply the avoidance of detrimental instrumental or relational issues, such as being denied valuable outcomes or being excluded from social groups (i.e., avoidance motivated)? Although the described theoretical perspectives seem to assume the first possibility, empirical research has ignored the question of whether voice effects are explained by approach or avoidance motivation. As a consequence, it is as yet unclear why people are motivated to have voice in a decisionmaking process, which constitutes a void in scientific knowledge on procedural justice. In the current research, we focus on the

687

question of whether the voice effect is driven by people’s approach or avoidance motivations. As such, the current research is designed to make a novel contribution to the procedural justice field by investigating the underlying motivational principles that explain people’s fairness-based reactions to voice and no-voice procedures, thus specifying and extending existing procedural justice theories. To explore the motivational nature of the voice effect, we empirically tested the theoretical assumption that in a gain-framed context, the voice effect is driven by people’s approach motivations, and we contrasted this proposition with the alternative possibility that the voice effect is driven by people’s avoidance motivations.

The Current Research People’s motivations to approach pleasure and to avoid pain are commonly referred to as the hedonic principle (Fo¨rster, Higgins, & Idson, 1998; Higgins, 1997). Although the hedonic principle has been ignored by empirical procedural justice research, it is central in other domains in social psychology. It has even been proposed that “the distinction between approach and avoidance motivation is fundamental and integral to the study of affect, cognition, and behavior” (Elliot & Thrash, 2002, p. 804). To investigate how the hedonic principle influences the voice effect, we base our line of reasoning on the idea that people’s motivational state stimulates their sensitivity to congruent social information (Higgins, 1997, 2000). To illuminate this point, it is likely that people who are in an avoidance motivational state are particularly sensitive to cues in their social environment that they associate with the avoidance of undesirable issues. As a consequence, people respond positively when they are successful in avoiding negative stimuli and respond negatively when they are unable to avoid negative stimuli. A similar process is likely to occur among people who are in an approach motivational state: People who are in such a state are particularly sensitive to cues in their social environment that they associate with the acquisition of desirable issues. As a consequence, people respond positively when they obtain desired benefits and respond negatively when they are denied these benefits. Thus, approach and avoidance motivation may direct people’s attention to different types of social information. This causal influence of approach and avoidance motivation on the type of information that people focus on has implications for how people respond to decision-making procedures. In correspondence with previous theorizing, we propose that people associate voice procedures with the acquisition of instrumental and relational benefits (Lind & Tyler, 1988; Thibaut & Walker, 1975; Tyler & Blader, 2003; Tyler & Lind, 1992). It therefore stands to reason that especially those people who are in an approach motivational state will pay attention to the extent to which they regard these decision-making procedures as fair or unfair. As a consequence, approach motivation is likely to stimulate people’s fairness evaluations following voice and no-voice procedures. According to this line of reasoning, it would be expected that people’s procedural justice judgments are particularly sensitive to the granting versus denial of voice when they are in an approach motivational state, as compared with when they are in an avoidance motivational state. The alternative possibility, however, is that people associate voice procedures with the avoidance of undesirable issues (e.g., being denied valuable outcomes, being excluded

VAN PROOIJEN, KARREMANS, AND VAN BEEST

688

from social groups). If this were the case, then it would be predicted that avoidance motivation stimulates people’s fairness evaluations following voice and no-voice procedures. According to this alternative line of reasoning, it would be expected that people’s procedural justice judgments are particularly sensitive to the granting versus denial of voice when they are in an avoidance motivational state as opposed to an approach motivational state. In the current research, we tested these opposing ideas in a series of three experiments. In Experiments 1 and 2, we tested our hypotheses by means of orthogonal manipulations of approach versus avoidance motor action. Furthermore, in Experiment 1 we manipulated voice versus no-voice procedures in a scenario, whereas in Experiment 2 participants directly experienced variations in voice versus no-voice procedures. In Experiment 3, we extended our research by measuring people’s approach and avoidance motivations as chronic individual-difference measures, followed by a direct manipulation of voice versus no-voice procedures. The main dependent variables in all three experiments were judgments that are typically assessed in procedural justice research: participants’ procedural justice judgments—that is, items referring to the extent to which participants believed that they were fairly treated by decision-making authorities (Lind & Tyler, 1988; Tyler & Lind, 1992).

Experiment 1 How can people’s approach and avoidance motivations be manipulated in an experimental setting? Previous research has indicated that people’s approach versus avoidance motivations are reflected in their automatic social–physiological behaviors, such as their motor actions (e.g., Chen & Bargh, 1999; Fo¨rster, 2003; Fo¨rster & Strack, 1996). For example, Chen and Bargh (1999) found that it is easier for people to pull positive items toward the body, whereas it is easier for people to push negative items away from the body. The arm movement of pulling items toward the body (“approach”) is referred to as arm flexion; the arm movement of pushing items away from the body (“avoidance”) is referred to as arm extension. When induced as an independent variable, arm flexion and extension produce bodily feedback that activates people’s approach and avoidance motivational orientations outside of their conscious awareness (e.g., Cacioppo, Priester, & Berntson, 1993; Fo¨rster, 2003; Friedman & Fo¨rster, 2000, 2002; Priester, Cacioppo, & Petty, 1996). According to Cacioppo et al. (1993), the reason for this can be found in classical conditioning principles: During one’s lifetime, muscle stimulations produced by arm flexion are most closely associated with the consumption or acquisition of desired stimuli (approach), whereas muscle stimulations produced by arm extension are most closely coupled with withdrawal or rejection of undesired stimuli (avoidance). On the basis of this line of reasoning, Cacioppo and his colleagues had participants either place their hand below a table and press upward (arm flexion) or place their hand on top of a table and press downward (arm extension). The results revealed that participants evaluated Chinese ideographs more positively in the arm flexion condition than in the arm extension condition. Furthermore, the motor action manipulation was cognitively associated with participants’ motivational orientations (approach vs. avoidance), not with their emotional orientations (pleasant vs. unpleasant). Moreover, empirical

data of six experiments ruled out alternative explanations, such as self-perception and demand characteristics. The research by Cacioppo and his colleagues is important because it indicated that nonaffective bodily feedback, as produced by arm flexion versus extension, can produce approach versus avoidance motivations that shape attitude development (see also Priester et al., 1996). The line of reasoning laid out in this article leads us to expect that contractions of the flexion versus extension muscles influence people’s justice-based reactions to voice versus no-voice procedures. After all, if people value opportunities to voice their opinions predominantly because they associate such procedures with the acquisition of instrumental or relational benefits, then they should be particularly sensitive to the granting versus denial of voice when they are in an approach motivational state (as produced by arm flexion) as compared with when they are in an avoidance motivational state (as produced by arm extension). On the basis of this line of reasoning, it can be predicted that people’s procedural justice judgments are influenced more strongly by voice as opposed to no-voice procedures when they flex their arms than when they extend their arms. Alternatively, if people value opportunities to voice their opinions predominantly because they associate such procedures with the avoidance of undesirable instrumental or relational issues, then they should be particularly sensitive to the granting versus denial of voice when they are in an avoidance motivational state (as produced by arm extension) as compared with when they are in an approach motivational state (as produced by arm flexion). The alternative prediction would then be that people’s procedural justice judgments are influenced more strongly by voice as opposed to no-voice procedures when they extend their arms than when they flex their arms.

Method Participants and design. Participants were assigned randomly to conditions of a 2 (motor action: arm flexion vs. arm extension) ⫻ 2 (procedure: voice vs. no-voice) factorial design. A total of 82 participants (31 men and 51 women, varying in age from 17 to 48 years) were recruited in the restaurant of the Free University Amsterdam. The experiment was preceded by other, unrelated studies. The studies lasted a total of 45 min. Participants engaged voluntarily in all of the experiments and were paid 5 euros (U.S.$6.25). Procedure. Upon entry in the laboratory, participants were led to separate individual cubicles. In the cubicles, participants found computer equipment, which was used to present the stimulus information and to register the data. The experiment was introduced as a study on how people respond to social situations if they simultaneously exert a slight physical effort. We then manipulated motor action: Participants in the arm flexion condition were asked to put the palm of one of their hands under the table and press upward, whereas participants in the arm extension condition were asked to put the palm of one of their hands on the table and press downward. Participants in both conditions were then asked to maintain a slight pressure against the table during the entire experiment and to work through the stimulus information with their one free hand using the computer mouse and keyboard. Participants were then presented with the following situation: Imagine that you have had a summer job in a company. The company has made good business in recent times. Because of a number of financial windfalls, the management has decided to give every employee a once-only financial bonus. These bonuses will be divided by every employee’s direct supervisor. Thus, your personal supervisor

PROCEDURAL JUSTICE AND THE HEDONIC PRINCIPLE during your summer job will decide what financial bonus you will receive. We then manipulated procedure:

Table 1 Means and Standard Deviations of Participants’ Procedural Justice Judgments as a Function of Motor Action and Procedure (Experiment 1)

Your direct supervisor gives you [voice/no voice] about what bonus you think that you should receive. Eventually, your supervisor decides to give you a bonus of 50 euros. After this, participants were asked to answer questions that pertained to the dependent variables. To measure procedural justice, we asked the following three questions: “How fair was the way you were treated by your supervisor?” (1 ⫽ very unfair, 7 ⫽ very fair), “How just was the way you were treated by your supervisor?” (1 ⫽ very unjust, 7 ⫽ very just), and “How appropriate was the way you were treated by your supervisor?” (1 ⫽ very inappropriate, 7 ⫽ very appropriate). These three items were averaged into a reliable procedural justice scale (␣ ⫽ .94). We then asked participants how annoying and how physically strenuous it was to press their hand against the table, and how much physical effort it took to press their hand against the table (1 ⫽ not at all, 7 ⫽ very much). Finally, we asked participants to indicate how they were pressing their hand against the table (1 ⫽ pressing upward, 2 ⫽ pressing downward, 3 ⫽ not pressing) and whether they used their left or right hand to press the table. After this, participants were told that they could stop exerting pressure against the table. Participants were fully debriefed, thanked, and paid for their participation.

Results A total of 5 participants indicated that they were not pressing their hand against the table in the correct way during the experiment. These participants were deleted from further analyses. Of the remaining 77 participants, 76 indicated that they had used their left hand to press against the table. One male participant indicated that he had used his right hand to press the table, and this participant was included in the analyses (results were similar when this participant was excluded). Unless noted explicitly, participants’ gender did not show significant effects on the variables analyzed below and was excluded as a factor in the reported analyses. Physical discomfort. A 2 ⫻ 2 multivariate analysis of variance (MANOVA) did not show any significant effects on either the multivariate or univariate level on the questions concerning how annoying and how physically strenuous it was for participants to press their hand against the table, and how much physical effort it took (for annoyance, overall M ⫽ 4.48, SD ⫽ 1.83; for strenuousness, overall M ⫽ 3.71, SD ⫽ 1.61; for physical effort, overall M ⫽ 3.26, SD ⫽ 2.70). When included in the analyses, gender showed a significant univariate main effect on strenuousness, F(1, 69) ⫽ 6.46, p ⬍ .02. Women found it more strenuous to press their hand against the table (M ⫽ 4.06, SD ⫽ 1.58) than men (M ⫽ 3.14, SD ⫽ 1.55). However, this main effect was independent from the experimental conditions and was nonsignificant on the multivariate level. More important, these results confirmed that participants in the various conditions did not differ significantly in their reported physical discomfort as a result of their motor action. Physical discomfort produced by motor action therefore cannot explain the results presented here. Procedural justice judgments. The means and standard deviations of participants’ procedural justice judgments are displayed in Table 1. A 2 ⫻ 2 analysis of variance (ANOVA) on procedural justice judgments revealed a significant main effect of procedure, F(1, 73) ⫽ 22.00, p ⬍ .001. More important, the results yielded the

689

Motor action Arm flexion

Arm extension

Procedure

M

SD

M

SD

Voice No-voice

5.91 3.68

0.86 1.40

4.75 4.09

1.54 1.47

Note. Higher means indicate more positive procedural justice judgments.

predicted interaction, F(1, 73) ⫽ 6.39, p ⬍ .02. Simple main effect analyses indicated that the procedure manipulation exerted a significant effect on procedural justice judgments in the arm flexion condition, F(1, 73) ⫽ 25.73, p ⬍ .001, but not in the arm extension condition, F(1, 73) ⫽ 2.32, p ⫽ .13. These results corroborate the hypothesis that participants are more strongly influenced by voice procedures if they conduct approach motor action (arm flexion) than if they conduct avoidance motor action (arm extension). Further, the motor action simple main effect was significant in the voice condition, F(1, 73) ⫽ 6.27, p ⬍ .02, but nonsignificant in the no-voice condition, F(1, 73) ⫽ 1.23, ns. Although approach motivation can elicit both positive and negative responses (i.e., by obtaining vs. not obtaining positive stimuli; Higgins, 1997), in the current experiment approach motor action stimulated participants’ positive responses to voice procedures, not their negative responses to no-voice procedures. We revisit this finding in the General Discussion.

Discussion The results of Experiment 1 indicate that participants’ procedural justice judgments were influenced more strongly by voice as opposed to no-voice procedures when they flexed their arms than when they extended their arms. This suggests that participants who conducted approach motor action were more sensitive to the procedure manipulation than participants who conducted avoidance motor action. In correspondence with our analysis of previous procedural justice research, the findings obtained in Experiment 1 support the idea that people’s fairness-based responses to voice as opposed to no-voice procedures are primarily driven by their approach motivations. Before drawing strong conclusions, it is important to replicate and extend these results. After all, in Experiment 1 participants responded to voice as opposed to no-voice procedures in a hypothetical situation. From the results of Experiment 1 alone, we do not know how arm flexion and extension influence people’s procedural justice judgments and satisfaction ratings if they directly experience variations in voice procedures. We therefore tested our hypotheses again in an experiment in which participants directly experienced a voice or a no-voice procedure.

Experiment 2 In Experiment 2, we tried to replicate and extend the findings of Experiment 1. In this experiment we made a number of modifica-

690

VAN PROOIJEN, KARREMANS, AND VAN BEEST

tions to our stimulus materials. As a first modification, in Experiment 2 we added a third condition to the motor action manipulation: a control condition in which participants were not exerting any flexion or extension pressure. It was important to do so in order to get an indication of whether arm flexion increases the strength of participants’ reactions to procedures or, alternatively, arm extension decreases the strength of participants’ reactions to procedures. On the basis of our line of reasoning in the introduction, we expected that approach motor action (arm flexion) would enhance the strength of the voice effect. After all, our line of reasoning implies that approach motivation increases people’s sensitivity to decision-making procedures. As a second modification, we adjusted the motor action manipulation: Participants either flexed or extended their arm muscles for 5 min while simultaneously conducting tasks, ostensibly to create an additional difficulty in participants’ task performance. After these 5 min, participants stopped exerting flexion or extension pressure, before they encountered the procedure manipulation or the dependent variables. We reasoned that if bodily feedback produced by contraction of the flexion or extension muscles activates people’s approach versus avoidance tendencies, then these tendencies should also be more accessible during the immediate relaxation period. This delayed induction of the procedure manipulation following the motor action manipulation should shed light on the time span within which nonaffective bodily feedback produces effects on people’s reactions to decision-making procedures. As a third modification to our stimulus materials, participants directly experienced a voice or a no-voice procedure: Participants were told that a number of lottery tickets would be divided among all participants. They then received a message from the experimenter stating that they were either granted or denied voice about the division of the lottery tickets. This manipulation of voice as opposed to no-voice procedures has been frequently used in procedural justice research (e.g., Van den Bos, 2003; Van den Bos et al., 1997, 1998; Van den Bos & Van Prooijen, 2001; Van Prooijen et al., 2002). The main dependent variables were again procedural justice judgments.

Method Participants and design. Participants were assigned randomly to conditions of a 3 (motor action: arm flexion vs. arm extension vs. control) ⫻ 2 (procedure: voice vs. no-voice) factorial design. A total of 118 participants (52 men and 66 women, age varying from 16 to 34 years) were recruited in the restaurant of the Free University Amsterdam. The experiment was followed by other, unrelated experiments. The experiments lasted a total of 45 min, and participants, who took part voluntarily, were paid 5 euros (U.S.$6.25). Procedure. Participants were seated in the same cubicles and behind the same computers as in Experiment 1. We then explained the experimental procedure to the participants. The experiment was introduced as a study on how people perform tasks. Participants were informed that they would perform two rounds of tasks: a practice round of 2 min and a work round of 5 min. Additionally, participants were led to believe that all computers in the lab were interconnected and that the experimenter, who was supposed to be in one of the cubicles, could send messages to all participants during the experiment. (In reality, all stimulus information was preprogrammed, a procedure none of the participants objected to upon debriefing.) Finally, participants were informed that a lottery with a prize of 50 euros would take place among all participants, and that following the

tasks the experimenter would allocate a number of lottery tickets to the participant. Next, the tasks were explained to the participants. Figures would be presented on the upper right side of the computer screen. Each figure consisted of 36 squares, and each square showed one of eight distinct patterns. One of these patterns was presented at the upper left side of the computer screen, and participants had to count the number of squares with this pattern in the figure on the right side of the screen. After participants had indicated the correct number, a new figure was presented. This procedure was repeated for 2 min in the practice round and for 5 min in the work round. In each round, the time remaining was presented on the lower left side of the computer screen, and the number of tasks completed (i.e., the number of figures the participant had counted during the round) was presented in the lower right side of the computer screen. After completing the practice round, we induced the motor action manipulation. Participants in the arm flexion and extension conditions were informed that they would face an additional difficulty in the work round: Ostensibly to investigate how they would perform the tasks if they simultaneously exerted a slight physical effort, participants were asked to press one of their hands against the table during the work round. In correspondence with Experiment 1, in the arm flexion condition participants were asked to put their palm under the table and press upward. In the arm extension condition participants were asked to put their palm on the table and press downward. In both conditions, participants were asked to maintain a slight pressure against the table during the entire work round and to work through the tasks using their one free hand. In the control condition, participants were not given the assignment to exert pressure against the table. The control condition thus is similar to that used in previous procedural justice experiments (e.g., Van den Bos, 2001, 2003; Van den Bos & Miedema, 2000; Van den Bos et al., 1998; Van Prooijen et al., 2002). After the work round, all participants were informed that their score on the tasks (in comparison with other participants) was about average. Furthermore, participants in the arm flexion and extension conditions were informed that they could stop exerting pressure against the table. We then asked participants in the arm flexion and extension conditions how annoying and how physically strenuous it was to press their hand against the table (1 ⫽ not at all, 7 ⫽ very much). Also, we asked participants to indicate how they were pressing their arm against the table (1 ⫽ pressing upward, 2 ⫽ pressing downward, 3 ⫽ not pressing) and whether they used their left or right arm to press the table. Additionally, we asked participants in all conditions how much physical effort it took to conduct the tasks and how pressured they felt during the tasks (1 ⫽ not at all, 7 ⫽ very much). The manipulation of procedure was then administered. Participants in the voice condition were informed that they were allowed an opportunity to voice their opinion about the number of lottery tickets that should be allocated to them. These participants were asked to type in the number of lottery tickets they thought they should receive. Participants in the no-voice condition were informed that they were not allowed an opportunity to voice their opinion about the number of lottery tickets that should be allocated to them. These participants were not asked to type in the number of lottery tickets they thought they should receive. Participants were then informed that they would be asked a number of questions before being informed about the number of lottery tickets they would receive. These questions constituted the dependent measures and the manipulation checks. To assess participants’ procedural justice judgments, we asked the following three questions: “How correctly were you treated by the experimenter?” (1 ⫽ very incorrectly, 7 ⫽ very correctly), “How dignified were you treated by the experimenter?” (1 ⫽ not very dignified, 7 ⫽ very dignified), and “How politely were you treated by the experimenter?” (1 ⫽ very impolitely, 7 ⫽ very politely). These three items were averaged into a reliable procedural justice scale (␣ ⫽ .93). To check the procedure manipulation, we asked the following two questions (1 ⫽ not at all, 7 ⫽ very much): “To what extent did the experimenter allow you an opportunity to

PROCEDURAL JUSTICE AND THE HEDONIC PRINCIPLE voice your opinion about the number of lottery tickets that should be allocated to you?” and “How much attention did the experimenter give to your opinion about the number of lottery tickets that should be allocated to you?” These two items were averaged into a reliable procedure check scale (␣ ⫽ .83). After this, participants were fully debriefed, thanked, and paid for their participation.

Table 2 Means and Standard Deviations of Participants’ Procedural Justice Judgments as a Function of Motor Action and Procedure (Experiment 2) Motor action Arm flexion

Results One male participant in the arm flexion condition indicated that he was not pressing his arm against the table during the tasks. This participant was excluded from further analyses, leaving 77 participants in the motor action conditions and 40 in the control condition. In the motor action conditions, 70 participants indicated that they used their left arm to press the table and 7 participants indicated that they used their right arm to press the table. When arm preference (left vs. right) was included as a dichotomous covariate in the analyses comparing the procedure effect in the arm flexion versus extension conditions, the results were similar to the results without this covariate, described below. Unless noted explicitly, gender of the participants did not influence the variables analyzed below and was excluded as a factor in the reported analyses. Manipulation check. A 3 ⫻ 2 ANOVA on the procedure check scale showed a significant procedure main effect only, F(1, 111) ⫽ 87.55, p ⬍ .001. Participants in the voice condition indicated having received more opportunities to voice their opinions (M ⫽ 5.16, SD ⫽ 1.45) than participants in the no-voice condition (M ⫽ 2.42, SD ⫽ 1.66). These results showed that participants had perceived the procedure manipulation as intended. Physical discomfort of motor action. Given that the questions concerning how annoying and how physically strenuous it was for participants to press their hands against the table could be posed only in the motor action conditions (i.e., arm flexion and extension), we analyzed these items with a 2 ⫻ 2 MANOVA. This analysis did not show significant effects on either the multivariate or the univariate level (for annoyance, overall M ⫽ 4.66, SD ⫽ 1.59; for strenuousness, overall M ⫽ 3.57, SD ⫽ 1.51). When included in the analyses, gender of the participant again influenced physical strenuousness on the univariate level, F(1, 69) ⫽ 5.54, p ⬍ .03. In correspondence with Experiment 1, women found it more physically strenuous to press their hand against the table (M ⫽ 3.91, SD ⫽ 1.48) than men (M ⫽ 3.12, SD ⫽ 1.45). However, this main effect was again nonsignificant on the multivariate level, and it was independent from the experimental manipulations. More important was that these results indicated that all participants experienced an equal amount of physical discomfort as a function of the motor action manipulation. Physical discomfort of motor action thus cannot explain the results reported here. Pleasantness of the tasks. A 3 ⫻ 2 MANOVA on the questions concerning how much physical effort it took to conduct the tasks and how pressured participants felt when conducting the tasks revealed no significant effects on either the multivariate or the univariate level (for physical effort, overall M ⫽ 2.77, SD ⫽ 1.48; for pressure, overall M ⫽ 3.42, SD ⫽ 1.60). These results indicated that participants in all conditions rated the tasks as equally pleasant. Pleasantness of the tasks therefore cannot explain the results described below. Procedural justice judgments. The means and standard deviations are displayed in Table 2. A 3 ⫻ 2 ANOVA on procedural

691

Arm extension

Control

Procedure

M

SD

M

SD

M

SD

Voice No-voice

5.47 2.15

1.09 0.84

5.22 2.90

1.00 1.14

5.20 3.12

0.98 1.42

Note.

Higher means indicate more positive procedural justice judgments.

justice judgments revealed a significant procedure main effect, F(1, 111) ⫽ 162.42, p ⬍ .001. Participants who received a voice procedure reported more positive procedural justice judgments (M ⫽ 5.29, SD ⫽ 1.01) than participants who received a no-voice procedure (M ⫽ 2.72, SD ⫽ 1.21). More important for the current purposes was that the results also yielded the predicted interaction, F(2, 111) ⫽ 3.41, p ⬍ .04. To more directly test our hypotheses, we conducted three interaction contrast analyses. First, we compared the procedure effect in the arm flexion condition with the procedure effect in the arm extension condition. This analysis showed a significant interaction contrast, F(1, 111) ⫽ 3.97, p ⬍ .05. Simple main effect analyses revealed that the procedure manipulation exerted stronger effects on procedural justice judgments in the arm flexion condition, F(1, 111) ⫽ 87.39, p ⬍ .001, ␩2 ⫽ .44, than in the arm extension condition, F(1, 111) ⫽ 45.81, p ⬍ .001, ␩2 ⫽ .29. This result corroborates our main hypothesis and replicates the findings of Experiment 1. Second, we contrasted the procedure effect in the arm flexion condition with the procedure effect in the control condition. This interaction contrast also turned out to be significant, F(1, 111) ⫽ 6.10, p ⬍ .02. Simple main effect analyses showed that the procedure simple main effect in the arm flexion condition was also stronger than in the control condition, F(1, 111) ⫽ 36.46, p ⬍ .001, ␩2 ⫽ .25. These results extend the findings obtained in Experiment 1 by revealing that approach motor action (i.e., arm flexion) enhances people’s reactions to voice versus no-voice procedures relative to a control condition. Third, we contrasted the procedure effect in the arm extension condition with the procedure effect in the control condition. This interaction contrast was nonsignificant (F ⬍ 1). As an aside, we note that the motor action simple main effect was nonsignificant in the voice condition (F ⬍ 1) but significant in the no-voice condition, F(2, 111) ⫽ 4.50, p ⬍ .02. This finding reflects one of the possible consequences of approach motivation. After all, approach motivation can produce either positive or negative reactions as a result of obtaining versus not obtaining positive stimuli (Higgins, 1997). However, this finding is inconsistent with Experiment 1, in which the motor action manipulation influenced reactions to voice instead of no-voice; we address this issue in the General Discussion.

Discussion In correspondence with Experiment 1, Experiment 2 revealed that people respond more strongly to voice as opposed to no-voice

VAN PROOIJEN, KARREMANS, AND VAN BEEST

692

procedures when they flex their arms than when they extend their arms. Furthermore, the results of Experiment 2 indicate that this difference between motor action conditions is attributable to an increased impact of the procedure manipulation in the arm flexion condition, not to a decreased impact of the procedure manipulation in the arm extension condition. After all, whereas the relative strength of the voice effect did not differ between the arm extension and control conditions, the arm flexion condition produced a significantly stronger voice effect than the control condition. These findings are supportive for the idea that approach motor action amplifies people’s fairness-based reactions to decision-making procedures. Experiments 1 and 2 focused on a specific and situational operationalization of approach and avoidance motivation through a manipulation of motor action. Although previous research has established clear indications that this motor action manipulation indeed reflects the hedonic principle (Cacioppo et al., 1993; Fo¨rster, 2003; Friedman & Fo¨rster, 2000, 2002), we believe that it is important to replicate the current results with a different operationalization of approach versus avoidance motivation. To get an indication of the generality of our findings and to increase our confidence that the hedonic principle provides an appropriate theoretical framework for the present findings, in Experiment 3 we measured participants’ chronic approach and avoidance motivations as individual-difference variables. We investigated the relation of these measures with a subsequent manipulation of voice versus no-voice procedures.

Experiment 3 In Experiment 3, we first measured Carver and White’s (1994) Behavioral Activation Scale (BAS) and Behavioral Inhibition Scale (BIS). These scales are designed to measure people’s approach versus avoidance motivational orientations as individualdifference variables. In the case of behavioral activation (BAS), we specifically focused on Carver and White’s Reward Responsiveness scale, which is designed to measure people’s approach responses to the occurrence or anticipation of rewarding events. This BAS scale most closely fits our theoretical line of reasoning, which has focused on the idea that people expect procedures to be rewarding because the procedures may imply the acquisition of instrumental gains or positive self-relevant social information (Tyler & Lind, 1992).1 After the measurement of participants’ approach and avoidance motivations, participants were informed that they would conduct another, ostensibly unrelated study. In this second study we manipulated voice versus no-voice procedures in the same way as we did in Experiment 2. The main dependent variables consisted of participants’ procedural justice judgments (Tyler & Lind, 1992). On the basis of our line of reasoning in the introduction, and the findings of Experiments 1 and 2, we predicted that people’s approach motivations (and not their avoidance motivations) would moderate procedural justice judgments as a function of voice as opposed to no-voice procedures. Specifically, we expected that the voice effect would be stronger among those high in behavioral activation than among those low in behavioral activation.

Method Participants and design. We tested our hypothesis in a design in which we measured participants’ approach and avoidance tendencies as continuous independent variables and manipulated procedure by randomly assigning participants to voice and no-voice conditions. Participants were 113 Leiden University students (42 men and 71 women, varying in age from 17 to 31 years). The experiment was preceded by another, unrelated experiment. The experiments lasted a total of 1 hr. Participants voluntarily engaged in the experiments and were paid 7 euros (U.S.$8.75). Procedure. On arrival at the laboratory, participants were led to separate cubicles. The cubicles contained computer equipment that was used to present the stimulus information and to register the data. The experiment was presented as two separate studies. Participants started with “Study 1,” which was presented as a study on “life experiences.” The study consisted of a series of questionnaires with 7-point scales. Among these questionnaires was Carver and White’s (1994) BIS, a seven-item scale designed to measure participants’ avoidance responses to the occurrence or anticipation of undesirable events (example item: “I worry about making mistakes”; ␣ ⫽ .81). Also included was Carver and White’s (1994) five-item Reward Responsiveness scale of the BAS. This scale is designed to measure people’s approach responses to the occurrence or anticipation of rewarding events (example item: “When I see an opportunity for something I like, I get excited right away”; ␣ ⫽ .78). In correspondence with Carver and White’s validation study, the BIS and BAS scales were positively correlated (r ⫽ .19, p ⬍ .05). The first study then ended, and participants continued with “Study 2.” In correspondence with Experiment 2, this study was presented as a study on how people conduct tasks. Furthermore, participants were informed that the experimenter could send messages to the participants using the computer network and additionally that a lottery would take place among all participants. The winner of the lottery would receive a prize of 50 euros. A total of 200 lottery tickets would be divided among all participants, and some of these lottery tickets would be allocated to the participant. Participants then started with the tasks. The tasks consisted of the counting of squares in the same way as in Experiment 2 (Van den Bos et al., 1997, 1998; Van den Bos & Miedema, 2000; Van den Bos & Van Prooijen, 2001; Van Prooijen et al., 2002, 2004a). However, in this experiment, participants would conduct a total of 25 tasks, without the time constraint that we used in Experiment 2. Participants were asked to complete all 25 tasks. Following the tasks, we manipulated procedure. This manipulation was the same as in Experiment 2. After the procedure manipulation, participants were again informed that they would first be asked some questions before being informed about the number of lottery tickets they would receive. These questions constituted the dependent measures and the manipulation checks. To measure procedural justice judgments, we asked the following questions: “How correctly were you treated by the experimenter?” (1 ⫽ very incorrectly, 7 ⫽ very correctly), “How respectfully were you treated by the experimenter?” (1 ⫽ very disrespectfully, 7 ⫽ very respectfully), “How politely were you treated by the experimenter?” (1 ⫽ very impolitely, 7 ⫽ very politely), and “To what extent do you respect the experimenter?” (1 ⫽ not at all, 7 ⫽ very much). Confirmatory factor analysis revealed an excellent fit of a one-factor model comprising these four items, ␹2(2, N ⫽ 113) ⫽ 2.27, ns; normed fit index ⫽ .99; nonnormed

1

Carver and White (1994) identified one BIS scale and three BAS scales. The two other BAS scales are the Drive scale, which measures people’s persistence when pursuing desirable goals, and the Fun Seeking scale, which measures people’s tendencies to approach new and exciting situations on the spur of the moment. Given that these two scales have a much poorer fit to our theoretical argument than the Reward Responsiveness scale, the Drive and Fun Seeking scales were ignored in the current research.

PROCEDURAL JUSTICE AND THE HEDONIC PRINCIPLE fit index ⫽ 1.00; comparative fit index ⫽ 1.00. We therefore averaged these items into a reliable procedural justice scale (␣ ⫽ .85). To check the procedure manipulation, we asked the same two questions as in Experiment 2, which were again averaged into a reliable procedure check scale (␣ ⫽ .78). The experiment then ended, and participants were debriefed, paid, and thanked for their participation.

Results The results were analyzed with linear regression analyses that specified as independent variables the main effects (the BIS scale, the BAS scale, and the procedure manipulation), the two-way interactions (BIS ⫻ Procedure, BAS ⫻ Procedure, and BIS ⫻ BAS), and the three-way interaction (BIS ⫻ BAS ⫻ Procedure). Following Cohen, Cohen, West, and Aiken’s (2003) recommendations, we centered participants’ answers on the BIS and BAS scales and effect-coded the procedure manipulation (1 and ⫺1). The interaction terms were based on the products of the centered BIS and BAS scales and the effect-coded procedure manipulation.2 When included as an independent variable in the regression analyses, gender did not show any main effects or interactions with the other independent variables on the manipulation check or the dependent variable. We therefore dropped gender in the statistical analyses reported here. Manipulation check. A regression analysis on the procedure check scale indicated that the regression equation accounted for a significant amount of variance (R2 ⫽ .48), F(7, 105) ⫽ 13.82, p ⬍ .001. Participants’ answers on the procedure check scale were predicted by the procedure main effect only (␤ ⫽ .70, p ⬍ .001). Participants in the voice condition perceived more opportunities to voice their opinions (M ⫽ 4.67, SD ⫽ 1.37) than participants in the no-voice condition (M ⫽ 2.11, SD ⫽ 1.47). From these analyses we conclude that participants perceived the procedure manipulation as intended. Procedural justice judgments. The results of the regression analysis on procedural justice judgments are displayed in Table 3. The regression equation accounted for a significant amount of variance on procedural justice judgments (R2 ⫽ .16), F(7, 105) ⫽ 2.79, p ⬍ .02. As displayed in Table 3, participants’ procedural justice judgments were predicted by the procedure manipulation (␤ ⫽ .32, p ⬍ .01). More important, the results also yielded a significant BAS ⫻ Procedure interaction term (␤ ⫽ .19, p ⬍ .05). To further explore this interaction term, we conducted simple

Table 3 Results From Regression Analyses: Procedural Justice Judgments as a Function of the Behavioral Inhibition Scale (BIS), the Behavioral Activation Scale (BAS), and Procedure (Experiment 3) Predictor

␤

t (105)

Procedure BIS BAS BIS ⫻ Procedure BAS ⫻ Procedure BIS ⫻ BAS BIS ⫻ BAS ⫻ Procedure

.32 ⫺.09 ⫺.11 .01 .19 ⫺.02 ⫺.03

3.49** ⫺0.94 ⫺1.15 0.07 2.06* ⫺0.19 ⫺0.33

* p ⬍ .05.

** p ⬍ .01.

693

slopes analyses (Cohen et al., 2003). Among those high in approach motivation, voice as opposed to no-voice procedures positively influenced procedural justice judgments (␤ ⫽ .41, p ⬍ .001), whereas among those low in approach motivation, voice as opposed to no-voice procedures did not influence procedural justice judgments (␤ ⫽ .13, p ⬎ .42). The BAS ⫻ Procedure interaction is illustrated in Figure 1. These results indicate that, as predicted, the procedure manipulation exerted stronger effects on participants’ procedural justice judgments among those high in approach motivation than among those low in approach motivation. Further, the BAS did not predict procedural justice judgments in the voice condition (␤ ⫽ .10, p ⬎ .45) but did predict procedural justice judgments in the no-voice condition (␤ ⫽ ⫺.28, p ⬍ .04). These latter results, which are consistent with the findings obtained in Experiment 2, are revisited in the General Discussion.

Discussion The results of Experiment 3 extend the results obtained in Experiments 1 and 2 in two ways. First, the finding that people’s approach motivations (and not their avoidance motivations) moderated the effects of voice as opposed to no-voice procedures further corroborates the idea that procedural justice judgments are shaped by people’s approach motivations and less so by their avoidance motivations. These findings are in correspondence with the findings obtained in Experiment 2, which indicated that approach motor action enhanced the voice effect relative to a control condition. Second, the findings in Experiment 3 increase the plausibility of the assumption that the motor action manipulation of Experiments 1 and 2 activated participants’ motivational orientations (cf. Cacioppo et al., 1993) and that it was these motivational orientations that influenced people’s reactions to decisionmaking procedures. After all, the findings of Experiment 3, in which we directly measured participants’ motivational orientations, are consistent with the effects of motor action in Experiments 1 and 2. Together with the previous experiments, Experiment 3 has further revealed evidence that people’s procedural justice judgments are influenced profoundly by their motivational orientations.

General Discussion The three experiments presented here provide empirical evidence for the proposition that people’s motivational orientations (approach vs. avoidance) influences the voice effect. The results of Experiments 1 and 2 showed that arm flexion, a motor reflex that is associated with approach motivation (Cacioppo et al., 1993; Fo¨rster, 2003), leads to stronger voice effects than arm extension, a motor reflex that is associated with avoidance motivation. These 2

Given that our main prediction was a lower order interaction (i.e., a two-way interaction in the presence of a three-way interaction), it was in this case particularly important to center the BIS and BAS scales. As emphasized by Cohen et al. (2003, p. 261), the interpretation of lower order coefficients in linear regression is meaningful only if all main effects and interactions are based on centered predictors. After centering our predictors, the reported linear regression analyses produced the same beta weights as hierarchical regression analyses.

VAN PROOIJEN, KARREMANS, AND VAN BEEST

694 6 5

Voice No-voice

4 3 2 low

high BAS

Figure 1. Procedural justice judgments as a function of the Behavioral Activation Scale (BAS) and procedure.

findings were extended in a third experiment in which we measured people’s approach and avoidance motivational orientations as individual-difference variables (Carver & White, 1994). The findings of Experiment 3 were in correspondence with Experiments 1 and 2 by revealing that people high in approach motivation displayed stronger reactions to voice as opposed to no-voice procedures than people low in approach motivation. Furthermore, people’s avoidance motivations turned out to be unrelated to the effects of voice versus no-voice procedures on procedural justice judgments. Taken together, the results of these experiments clearly suggest that voice effects are shaped by people’s approach motivations, and less so by their avoidance motivations. Although these effects were found in three experiments, we note that approach motivation stimulated reactions to voice procedures in Experiment 1 and reactions to no-voice procedures in Experiments 2 and 3. These differential effects may reflect two possible approach-motivated reactions (Higgins, 1997): People may respond positively to obtaining pleasure (as implied by voice procedures) or negatively to being denied pleasure (as implied by no-voice procedures). (Likewise, Higgins noted that people’s avoidance motivations can also be reflected both by positive reactions to avoiding pain and by negative reactions to not avoiding pain.) The findings in Experiments 2 and 3 were in correspondence with previous notions that people’s negative responses to injustice usually are stronger than their positive responses to justice (Folger, 1984; Van Prooijen, Van den Bos, Lind, & Wilke, 2006). It may therefore be the case that reactions to no-voice procedures in particular are sensitive to people’s motivational orientations. However, whereas in Experiments 2 and 3 participants were not fully informed about their outcomes, in the scenario of Experiment 1 participants received full disclosure of the positive outcomes of voice. This explicit outcome information might have caused approach motivation to stimulate people’s responses to voice instead of no-voice procedures. These explanations are speculative, and future research might explore under what conditions motivational orientations influence reactions to voice or no-voice procedures. For now, it seems safe to conclude that people’s motivational orientations have the potential to influence reactions to both voice and no-voice procedures. More important for the current purposes is the finding that people’s approach motivations yielded stronger reactions to voice versus no-voice procedures than people’s avoidance motivations, a finding that was replicated in three experiments.

The main theoretical contribution of the current research is that it has increased insights into the motivational nature of the voice effect. We have made explicit that theoretical perspectives on procedural justice have assumed that people’s approach motivations stimulate their fairness-based responses to voice and novoice procedures (e.g., Lind & Tyler, 1988; Thibaut & Walker, 1975; Tyler & Blader, 2003; Tyler & Lind, 1992). In correspondence with this assumption, our results suggest that participants’ procedural justice judgments were predominantly associated with the chronic or situational accessibility of their approach motivations. This conclusion has implications for both fundamental and applied social justice research. After all, people are subjected to voice or no-voice procedures in numerous social settings, such as organizations, legal settings, schools, and other settings that involve interactions with decision-making authorities (Lind & Tyler, 1988). In such social situations, knowing whether people react out of approach or avoidance motivation might influence how to assuage their negative reactions when they are denied voice or when providing voice is impossible. Furthermore, understanding the motivational nature of voice effects may increase scientists’ ability to predict people’s reactions to voice or no-voice procedures in these social situations. These practical implications need to be tested further, of course, and it would be premature to draw firm conclusions regarding the real-life implications of the current findings. Although the findings presented here may not be instantly applicable to all possible types of social situations—as usually is the case in experimental research—the conclusions that we draw here extend existing procedural justice theories (Thibaut & Walker, 1975; Tyler & Lind, 1992) and may therefore provide a more solid theoretical base for both fundamental and applied research. It is important to note that the current findings do not imply that people’s reactions to procedures necessarily are always approach motivated (cf. Camacho, Higgins, & Luger, 2003). Notably, a constraint of our experiments was that participants received voice versus no-voice procedures about decisions that involved gains instead of losses (i.e., a financial bonus in Experiment 1 and lottery tickets in Experiments 2 and 3). As noted in the introduction, this focus on gains is in correspondence both with many real-life decisions involving positive outcomes and with a substantial number of previous procedural justice studies. We therefore decided that a focus on gain-framed decisions would be a good starting point to explore the underlying motivational principles of voice effects. Having said this, we note that in everyday life, loss decisions do also happen of course, and it is noteworthy that some procedural justice research has explicitly focused on loss decisions. For example, Brockner and his colleagues have studied procedural justice effects in the context of job layoff decisions (e.g., Brockner et al., 1994, 1998). It would be interesting to investigate how approach versus avoidance motivation influences the voice effect in these loss-oriented situations. More specifically, it might be expected that a gain–loss distinction moderates the influence of motivational orientations on fairness-based reactions to voice procedures. Such a prediction would be consistent with regulatory focus theory, which has posited that the hedonic principle operates differently when serving different human needs (Higgins, 1997, 2000; see also Camacho et al., 2003; Fo¨rster et al., 1998; Lee & Aaker, 2004). According to this theory, people’s desired end state in a gain-framed context

PROCEDURAL JUSTICE AND THE HEDONIC PRINCIPLE

entails accomplishments (i.e., a promotion focus), and to achieve this end state, people strategically approach and avoid gains versus nongains. Given the focus on gains, however, approach motivation is the dominant tendency because it provides a better fit to people’s regulatory orientation than avoidance motivation. In a loss-framed context, people’s desired end state entails security (i.e., a prevention focus), and to achieve this end state, people strategically approach and avoid losses versus nonlosses. Given the focus on losses, however, avoidance motivation provides the better fit to people’s regulatory orientation. Extrapolating this theoretical framework to the procedural justice field, it may be expected that when people are confronted with voice or no-voice procedures in loss-framed situations (e.g., when material losses are at stake, or social losses such as when people are threatened to be excluded from valuable social groups), their fairness-based reactions may be stimulated by their avoidance motivations instead of their approach motivations. These ideas are beyond the scope of the current article but do constitute a challenging opportunity for future research. It is important to keep in mind that the current research was not focused on differential effects of gains versus losses but on the possibility that the hedonic principle constitutes an important motivational dimension to understand people’s procedural justice judgments. Another issue that we would like to raise is that we focused exclusively on voice versus no-voice procedures and not on other elements of procedural justice, such as procedural accuracy or consistency between persons (e.g., Leventhal, 1980). Voice versus no-voice procedures strongly influence procedural justice judgments and constitute the basis for influential procedural justice theories (Brockner et al., 1998; Folger, 1977; Lind & Tyler, 1988; Tyler & Lind, 1992; Van den Bos, 2001; Van den Bos et al., 1998). However, voice is also distinct from other elements of procedural justice because it has more of an agentic quality to it: In the case of voice procedures, procedural justice judgments partially depend on actions of the recipient, that is, their willingness and ability to provide input in the decision-making process. In contrast, most other elements of procedural justice almost exclusively refer to actions on the part of the decision maker rather than actions by the recipient. This active role of the recipient, which differentiates voice from other elements of procedural justice, may be psychologically associated with people’s approach motivations. This line of reasoning is speculative and leads to the empirical question of whether the current findings would generalize to other elements of procedural justice, such as accurate versus inaccurate procedures. Be that as it may, for now one can conclude that motivational orientations have a profound influence on fairness-based reactions to voice and no-voice procedures, and these procedures are a key factor in the psychology of procedural justice. Given that hedonic motivation precedes the hedonic states of pleasure and pain, it might be reasoned that the current findings are related to mood. We suspect that the current findings cannot be explained by variations in mood, for at least two reasons. First, previous research has indicated that the motor action manipulation does not influence mood (Cacioppo et al., 1993). This finding is consistent with the assumption that the motor action manipulation induces approach and avoidance motivation: After all, both approach and avoidance motivation can elicit positive and negative moods, depending on how successful people are in approaching pleasure or avoiding pain. Second, previous research has explicitly

695

investigated the influence of mood on reactions to voice and no-voice procedures (Van den Bos, 2003), and the results were different from the current findings. Van den Bos (2003) found that mood influenced procedural fairness judgments only if people did not have information concerning what decision-making procedures were adopted by authorities. In the current experiments, however, all participants received explicit information about the procedures used (voice or no-voice), yet motor action moderated the voice effect in predictable ways. Taken together, the current findings are more likely explained by hedonic motivation (approach vs. avoidance) rather than by the valence of specific hedonic states (positive vs. negative). The main ideas of the current article were inspired by both instrumental and noninstrumental perspectives of procedural justice effects. Nevertheless, it should be noted that the present studies were not intended as direct tests of either of these perspectives, nor did we intend to show that both outcomes and relational concerns explain the relation between the hedonic principle and procedural justice judgments. After all, our experimental designs did not incorporate manipulations of outcomes or of interpersonal relatedness. Rather, we have assumed that the hedonic principle constitutes a basic phenomenon that describes the motivational process of how people seek to fulfill their instrumental and relational needs. These instrumental and relational needs are typical procedural-justice-related concerns (Lind & Tyler, 1988), leading us to reason that the hedonic principle is a predictor of procedural justice judgments. Given that the results were supportive for this idea, the current research may provide a starting point to more directly explore the processes that form the motivational underpinnings of people’s strivings for fair outcome distributions and respectful interpersonal treatment. To conclude, we have tried to reveal here that the hedonic principle is related to people’s fairness-based reactions to decisionmaking procedures. Evidence for this idea was found by investigating the effects of physiological feedback produced by approach versus avoidance muscle stimulations (Experiments 1 and 2) and individual-difference measures of behavioral activation and behavioral inhibition (Experiment 3) on reactions to procedures. These three studies have led to the conclusion that people are particularly sensitive to common conceptualizations of voice versus no-voice procedures when they are approach motivated. As such, one can conclude that the hedonic principle is an important principle in the psychology of procedural justice.

References Brockner, J., Heuer, L., Siegel, P. A., Wiesenfeld, B., Martin, C., Grover, S., Reed, T., & Bjorgvinsson, S. (1998). The moderating effect of self-esteem in reaction to voice: Converging evidence from five studies. Journal of Personality and Social Psychology, 75, 394 – 407. Brockner, J., Konovsky, M., Cooper-Schneider, R., Folger, R., Martin, C., & Bies, R. J. (1994). Interactive effects of procedural justice and outcome negativity on victims and survivors of job loss. Academy of Management Journal, 37, 397– 409. Brockner, J., & Wiesenfeld, B. M. (1996). An integrative framework for explaining reactions to decisions: Interactive effects of outcomes and procedures. Psychological Bulletin, 120, 189 –208. Cacioppo, J. T., Priester, J. R., & Berntson, G. G. (1993). Rudimentary determinants of attitudes: II. Arm flexion and extension have differential effects on attitudes. Journal of Personality and Social Psychology, 65, 5–17.

696

VAN PROOIJEN, KARREMANS, AND VAN BEEST

Camacho, C. J., Higgins, E. T., & Luger, L. (2003). Moral value transfer from regulatory fit: What feels right is right and what feels wrong is wrong. Journal of Personality and Social Psychology, 84, 498 –510. Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: The BIS/BAS scales. Journal of Personality and Social Psychology, 67, 319 –333. Chen, M., & Bargh, J. A. (1999). Consequences of automatic evaluation: Immediate behavioral predispositions to approach or avoid the stimulus. Personality and Social Psychology Bulletin, 25, 215–224. Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Erlbaum. Cropanzano, R., Byrne, Z. S., Bobocel, D. R., & Rupp, D. E. (2001). Moral virtues, fairness heuristics, social entities, and other denizens of organizational justice. Journal of Vocational Behavior, 58, 164 –209. Elliot, A. J., & Thrash, T. M. (2002). Approach–avoidance motivation in personality: Approach and avoidance temperaments and goals. Journal of Personality and Social Psychology, 82, 804 – 818. Folger, R. (1977). Distributive and procedural justice: Combined impact of “voice” and improvement on experienced inequity. Journal of Personality and Social Psychology, 35, 108 –119. Folger, R. (1984). Emerging issues in the social psychology of justice. In R. Folger (Ed.), The sense of injustice: Social psychological perspectives (pp. 3–24). New York: Plenum Press. Folger, R., & Cropanzano, R. (1998). Organizational justice and human resource management. Thousand Oaks, CA: Sage. Folger, R., Rosenfield, D., Grove, J., & Corkran, L. (1979). Effects of “voice” and peer opinions on responses to inequity. Journal of Personality and Social Psychology, 37, 2253–2261. Fo¨rster, J. (2003). The influence of approach and avoidance motor actions on food intake. European Journal of Social Psychology, 33, 339 –350. Fo¨rster, J., Higgins, E. T., & Idson, L. C. (1998). Approach and avoidance strength during goal attainment: Regulatory focus and the “goal looms larger” effect. Journal of Personality and Social Psychology, 75, 1115– 1131. Fo¨rster, J., & Strack, F. (1996). Influence of overt head movements on memory for valenced words: A case of conceptual-motor compatibility. Journal of Personality and Social Psychology, 71, 421– 430. Friedman, R. S., & Fo¨rster, J. (2000). The effects of approach and avoidance motor actions on the elements of creative insight. Journal of Personality and Social Psychology, 79, 477– 492. Friedman, R. S., & Fo¨rster, J. (2002). The influence of approach and avoidance motor actions on creative cognition. Journal of Experimental Social Psychology, 38, 41–55. Greenberg, J., & Folger, R. (1983). Procedural justice, participation, and the fair process effect in groups and organizations. In P. B. Paulus (Ed.), Basic group processes (pp. 235–256). New York: Springer-Verlag. Higgins, E. T. (1997). Beyond pleasure and pain. American Psychologist, 52, 1280 –1300. Higgins, E. T. (2000). Making a good decision: Value from fit. American Psychologist, 55, 1217–1230. Houlden, P., LaTour, S., Walker, L., & Thibaut, J. (1978). Preference for modes of dispute resolution as a function of process and decision control. Journal of Experimental Social Psychology, 14, 13–30. Lee, A. Y., & Aaker, J. L. (2004). Bringing the frame into focus: The influence of regulatory fit on processing fluency and persuasion. Journal of Personality and Social Psychology, 86, 205–218. Leventhal, G. S. (1980). What should be done with equity theory? New approaches to the study of fairness in social relationships. In K. J. Gergen, M. S. Greenberg, & R. H. Willis (Eds.), Social exchange: Advances on theory and research (pp. 27–54). New York: Plenum. Lind, E. A. (2001). Thinking critically about justice judgments. Journal of Vocational Behavior, 58, 220 –226.

Lind, E. A., Kanfer, R., & Earley, P. C. (1990). Voice, control, and procedural justice: Instrumental and noninstrumental concerns in fairness judgments. Journal of Personality and Social Psychology, 59, 952–959. Lind, E. A., & Tyler, T. R. (1988). The social psychology of procedural justice. New York: Plenum Press. Priester, J. R., Cacioppo, J. T., & Petty, R. E. (1996). The influence of motor processes on attitudes toward novel versus familiar semantic stimuli. Personality and Social Psychology Bulletin, 22, 442– 447. Thibaut, J., & Walker, L. (1975). Procedural justice: A psychological analysis. Hillsdale, NJ: Erlbaum. Tyler, T. R. (1987). Conditions leading to value expressive effects in judgments of procedural justice: A test of four models. Journal of Personality and Social Psychology, 52, 333–344. Tyler, T. R. (1989). The psychology of procedural justice: A test of the group-value model. Journal of Personality and Social Psychology, 57, 830 – 838. Tyler, T. R., & Blader, S. L. (2000). Cooperation in groups: Procedural justice, social identity, and behavioral engagement. Philadelphia: Taylor & Francis. Tyler, T. R., & Blader, S. L. (2003). The group engagement model: Procedural justice, social identity, and cooperative behavior. Personality and Social Psychology Review, 7, 349 –361. Tyler, T. R., & Lind, E. A. (1992). A relational model of authority in groups. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 25, pp. 115–292). San Diego, CA: Academic Press. Tyler, T. R., Rasinski, K., & Spodick, N. (1985). The influence of voice on satisfaction with leaders: Exploring the meaning of process control. Journal of Personality and Social Psychology, 48, 72– 81. Van den Bos, K. (1999). What are we talking about when we talk about no-voice procedures? On the psychology of the fair outcome effect. Journal of Experimental Social Psychology, 35, 560 –577. Van den Bos, K. (2001). Uncertainty management: The influence of human uncertainty on reactions to perceived fairness. Journal of Personality and Social Psychology, 80, 931–941. Van den Bos, K. (2003). On the subjective quality of social justice: The role of affect as information in the psychology of justice judgments. Journal of Personality and Social Psychology, 85, 482– 498. Van den Bos, K., & Lind, E. A. (2002). Uncertainty management by means of fairness judgments. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 34, pp. 1– 60). San Diego, CA: Academic Press. Van den Bos, K., Lind, E. A., Vermunt, R., & Wilke, H. A. M. (1997). How do I judge my outcome when I do not know the outcome of others? The psychology of the fair process effect. Journal of Personality and Social Psychology, 72, 1034 –1046. Van den Bos, K., & Miedema, J. (2000). Towards understanding why fairness matters: The influence of mortality salience on reactions to procedural fairness. Journal of Personality and Social Psychology, 79, 355–366. Van den Bos, K., & Van Prooijen, J.-W. (2001). Referent cognitions theory: The psychology of voice depends on closeness of reference points. Journal of Personality and Social Psychology, 81, 616 – 626. Van den Bos, K., Wilke, H. A. M., Lind, E. A., & Vermunt, R. (1998). Evaluating outcomes by means of the fair process effect: Evidence for different processes in fairness and satisfaction judgments. Journal of Personality and Social Psychology, 74, 1493–1503. Van Prooijen, J.-W., Van den Bos, K., Lind, E. A., & Wilke, H. A. M. (2006). How do people react to negative procedures? On the moderating role of authority’s biased attitudes. Journal of Experimental Social Psychology, 42, 632– 645. Van Prooijen, J.-W., Van den Bos, K., & Wilke, H. A. M. (2002). Procedural justice and status: Status salience as antecedent of procedural fairness effects. Journal of Personality and Social Psychology, 83, 1353–1361.

PROCEDURAL JUSTICE AND THE HEDONIC PRINCIPLE Van Prooijen, J.-W., Van den Bos, K., & Wilke, H. A. M. (2004a). Group belongingness and procedural justice: Social inclusion and exclusion by peers affects the psychology of voice. Journal of Personality and Social Psychology, 86, 66 –79. Van Prooijen, J.-W., Van den Bos, K., & Wilke, H. A. M. (2004b). The role of standing in the psychology of procedural justice: Towards theoretical integration. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 15, pp. 33–58). East Sussex, England: Psychology Press.

697

Van Prooijen, J.-W., Van den Bos, K., & Wilke, H. A. M. (2005). Procedural justice and intragroup status: Knowing where we stand in a group enhances reactions to procedures. Journal of Experimental Social Psychology, 41, 664 – 676.

Received June 30, 2005 Revision received March 22, 2006 Accepted March 23, 2006 䡲

PROCEDURAL JUSTICE AND THE HEDONIC PRINCIPLE Van Prooijen, J.-W., Van den Bos, K., & Wilke, H. A. M. (2004a). Group belongingness and procedural justice: Social inclusion and exclusion by peers affects the psychology of voice. Journal of Personality and Social Psychology, 86, 66 –79. Van Prooijen, J.-W., Van den Bos, K., & Wilke, H. A. M. (2004b). The role of standing in the psychology of procedural justice: Towards theoretical integration. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 15, pp. 33–58). East Sussex, England: Psychology Press.

697

Van Prooijen, J.-W., Van den Bos, K., & Wilke, H. A. M. (2005). Procedural justice and intragroup status: Knowing where we stand in a group enhances reactions to procedures. Journal of Experimental Social Psychology, 41, 664 – 676.

Received June 30, 2005 Revision received March 22, 2006 Accepted March 23, 2006 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 698 –711

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.698

The Paradox of Group-Based Guilt: Modes of National Identification, Conflict Vehemence, and Reactions to the In-Group’s Moral Violations Sonia Roccas

Yechiel Klar and Ido Liviatan

The Open University of Israel

Tel Aviv University

The authors examined the relationships between 2 modes of national identification (attachment to the in-group and the in-group’s glorification) and reactions to the in-group’s moral violations among Israeli students. Data were collected during a period of relative calm in the Israeli–Palestinian conflict as well as during a period of great intensification of this conflict. As expected, in Study 1, the 2 modes of identification had contrasting relationships with group-based guilt: Attachment was positively related whereas glorification was negatively related to group-based guilt for in-group’s past infractions. Glorification suppressed the attachment effect but not vice versa. Both relationships were mediated by the use of exonerating cognitions. In Study 2, group-based guilt for the in-group’s current wrongdoings was increased by priming critical rather than conventional attachment to the in-group, suggesting a causal effect of mode of identification on the experience of negative group-based emotions. Keywords: group-based guilt, national identification, moral violations

mobilize internal and external support to continue the fight. However, it is precisely during times of open conflict that there is heightened danger that the conflicting parties may engage in brutal and inhumane acts, making willingness to recognize the potential immorality of in-group actions imperative. One of the goals of the current research is to examine factors affecting the ways people react to information questioning the morality of their in-group’s actions during different stages of an intense ongoing intergroup conflict. An additional goal of this study is to disentangle a basic theoretical puzzle in group-based guilt literature: the conflicting relationships between group identification and group-based guilt. In previous research on group-based guilt, identification with the group has sometimes been characterized as the major buffer to experiencing feelings of guilt. For example, Doosje et al. (1998) argued that high identifiers experience less group-based guilt because they quickly reject the notion that their group has committed immoral deeds and can thus absolve themselves from moral distress. Subsequent results, however, have provided mixed support for this position (see Branscombe, 2004). In this article, we present a multifaceted view of national identification and test the hypothesis that different modes of identification have opposing relations to feelings of group-based guilt. We propose that identification can simultaneously increase and decrease feelings of group-based guilt, depending on the mode of identification. More specifically, we argue that attachment to the group without glorifying it is the “active ingredient” in the likelihood of experiencing group-based guilt. In the studies reported here, we focused on the Israeli– Palestinian conflict, which is regarded by many as one of the most intractable contemporary intergroup conflicts (Bar-Tal, 1998; Rouhana & Bar-Tal, 1998). In the first study, we examined judgments of Jewish Israelis about moral violations perpetrated by Israelis in the context of this conflict. We examined two samples. The first sample responded in the spring of 2000, during a period of relative calm in the conflict. Unfortunately, in September 2000 there was

In recent years there has been increased interest in how groups conceptualize their troubled history (e.g., Branscombe & Doosje, 2004; Branscombe, Doosje, & McGarty, 2002; Doosje, Branscombe, Spears, & Manstead, 1998). Typical examples are post-Holocaust Germany (e.g., Michman, 2002; Niven, 2002; Rensmann, 2004); nations with a colonial past such as Belgium, Portugal, and the Netherlands; (e.g., Cohen, 2001; Doosje et al., 1998; Licata & Klein, 2004); or nations with a history of ethnic cleansing of native populations such as the United States, Canada, and Australia (e.g., Augoustinos, & LeCouteur, 2004; Barkan & Snowden, 2001). These studies have provided important insights into the ways in which group members react to information that challenges their group’s morality. However, the willingness of group members to question the morality of their in-group in the context of violent ongoing and unsolved conflicts is far from being fully understood. During an ongoing and violent conflict, people are likely to be especially reluctant to criticize the morality of their in-group. Belief in the moral superiority of the in-group and the justifiability of its actions is critical in times of conflict because it is needed to Sonia Roccas, Department of Psychology and Education, The Open University of Israel, Tel Aviv, Israel; Yechiel Klar and Ido Liviatan, Department of Psychology, Tel Aviv University, Tel Aviv, Israel. Ido Liviatan is now at the Department of Psychology, New York University. The work of Sonia Roccas was supported by a grant from the Solomon Asch Center for the Study of Ethnopolitical Conflict and by Israel Science Foundation Grant 785/03. The work of Yechiel Klar at the later stages of research was supported by Israel Science Foundation Grant 1211/05. We are thankful to Ilan Roziner for his help in analyzing the data and to Talia Fried and Michal Bitton for their help in designing the instruments and collecting the data. Correspondence concerning this article should be addressed to Sonia Roccas, Department of Psychology and Education, The Open University of Israel, 16 Klausner Street, P.O.B. 39328, Ramat-Aviv, Tel Aviv 61392, Israel. E-mail: [email protected] 698

IDENTIFICATION AND GROUP-BASED GUILT

699

a resurgence of open hostilities between the two parties. We decided to repeat the study and explore the effects of the resurgence of open conflict on the retrospective perception of the in-group’s past moral infractions (i.e., group actions that took place long before the recent escalation). The second sample in Study 1 responded in the fall of 2001 following 1 year of intifada (Palestinian armed struggle). In Study 2 we experimentally manipulated critical attachment (i.e., attachment without glorification) and conventional attachment to assess their effects on groupbased guilt for current wrongdoings.

who are strongly identified with their group are more likely to experience group-based emotions than are people whose identification is weak, but they qualified this claim as follows:

Identification and Group-Based Guilt: A Basic Paradox

The authors concluded that high identifiers are more likely than low identifiers to employ a variety of defensive reactions and thus are less likely to experience group-based guilt. However, there has been little support for their hypothesis that highly attached members experience less group-based guilt than their low attachment counterparts. A recent review (Branscombe, 2004) indicates that the relationship between group identification and guilt varies across studies: Identification was sometimes positively related to group-based guilt (Doosje, Branscombe, Spears, & Manstead, 2004), sometimes negatively related to it (Doosje et al., 1998), and sometimes there was no overall relationship between the two variables (Branscombe, Slugoski, & Kappen, 2004). Identification with the national group is defined in this literature as feelings of attachment to one’s group (Doosje et al., 2004) and is measured with items such as “I identify with other (name of group members),” “(name of group) is an important group to me,” and “Being an (name of group) is an important part of how I see myself at this moment” (Doosje et al., 1998). We propose that the inconsistencies found in past research on the relationship of identification and group-based guilt reflect the complex effects of identification, which cannot be detected with a unidimensional approach. More specifically, we propose, in light of the identification-guilt paradox, that identification with a group simultaneously increases and decreases the propensity toward group-based guilt. To capture these complex effects, a more finely tuned approach to identification is needed. For this purpose, we base much of our analysis of the different modes of identification on the extensive theoretical and empirical studies that have focused on identification with the nation.

The relationship between identification with one’s national group and the experience of group-based guilt presents an inherent contradiction: On the one hand, group-based guilt is by definition “guilt by association” (Doosje et al., 1998). That is, it refers to feeling guilty for deeds that a person did not commit—feelings that arise because people are associated with the perpetrators by virtue of their common group membership. To feel morally implicated in immoral acts performed by in-group members, individuals need at the very least to self-categorize as members of the perpetrating group (Branscombe et al., 2002; Doosje et al., 1998). Among people who do not evade the relevant self-categorization, those who are attached to the group should be most susceptible to group-based emotions. Thus, one may expect that individuals who are most highly identified with a group to be most likely to feel group-based guilt. Stated differently, it can be argued that individuals who do not include a group as a major component of their social identity are less likely to feel guilt over moral infractions committed by group members.1 Thus, it can be argued that because group-based guilt is guilt by association, identification should be positively related to feeling moral guilt over the in-group’s wrongdoings. On the other hand, those who are highly identified should also be the most motivated to defend their group identity (Branscombe et al., 2002). Extensive research indicates that individuals who identify with their group tend to feel that the group is good and moral (e.g., Janis, 1982; Staub, 1997). When group members are confronted with negative information regarding their group, they often reinterpret this information in ways that protect their ability to derive a positive social identity from their group membership (Mummendey, Klink, Mielke, Wenzel, & Blanz, 1999; Tajfel & Turner, 1979). According to this line of reasoning, identification with a group should lead to legitimization of the group’s actions and consequently should be negatively related to people’s feelings of group-based guilt. Thus the paradox may be delineated as follows: Being identified with one’s group should be associated with experiencing stronger group-based emotions and thus should be associated with feeling stronger group-based guilt. But being identified with the group should also be associated with legitimization of the group’s wrongdoings and hence feeling little or no guilt.

The Unidimensional Approach to the Identification-Guilt Paradox This identification-guilt paradox did not elude Doosje et al.’s (1998) attention. The authors suggested that in general, people

Negative group-image threatening emotions, such as guilt or shame, however are likely to be experienced only by people who are willing to admit or accept that their group has done something wrong in the first place. We argue that high identifiers are typically unlikely to accept a negative interpretation of their group history and that they may have other defensive means of dealing with such a group threatening situation. (Doosje et al., 1998, p. 879)

Dual Conceptualizations of Identification With the National Group Theoreticians and researchers from diverse areas have distinguished two main aspects of national identification. Adorno, Frankel-Brunswik, Levinson, and Sanford (1950) differentiated pseudo-patriotism, defined as “blind attachment to certain national cultural values, uncritical conformity with the prevailing group ways, and rejection of other nations as outgroups” (p. 107), from genuine patriotism, defined as “love of the country and attachment to national values based on critical understanding” (p. 107). Kosterman and Feshbach (1989) suggested a similar distinction be1

Obviously, individuals who do not identify with the group may still feel enraged over its acts, in the same way that nonmembers may be. However they are not expected to feel much responsibility for the deeds or feel resultant guilt.

700

ROCCAS, KLAR, AND LIVIATAN

tween nationalism, defined as a view that one’s nation is superior and should be dominant, and patriotism, defined as a feeling of attachment to one’s nation. More recently, Staub (1997) proposed distinctions among blind patriotism, its conceptual opposite constructive patriotism, and conventional patriotism. Blind patriotism was defined as a rigid and inflexible attachment to one’s country, characterized by unquestioning positive evaluation, staunch allegiance, and intolerance of criticism; conventional patriotism was defined simply as positive identification with and feelings of affective attachment to one’s nation (Schatz & Staub, 1997); and constructive patriotism was defined as attachment to one’s country characterized by critical loyalty that includes “questioning and criticism of current group practices that are driven by a desire for positive change” (Schatz, Staub, & Lavine, 1999, p. 153). A similar distinction can also be found in literature dealing with individual-level collectivism. Triandis (1995) described collectivism as a social pattern consisting of closely linked individuals who see themselves as parts of one or more collectives (family, co-workers, tribe, nation); are primarily motivated by the norms of, and duties imposed by, those collectives; are willing to give priority to the goals of these collectives over their own personal goals; and emphasize their connectedness to the members of these collectives. (p. 2)

Triandis and Gelfand (1998) suggested distinguishing two types of collectivism. Horizontal collectivists see themselves as similar to others and emphasize common goals, interdependence, and sociability but do not emphasize submitting to authority. Vertical collectivists “emphasize the integrity of the ingroup, are willing to sacrifice their personal goals for the sake of ingroup goals, and support competitions of their ingroup with outgroups” (Triandis & Gelfand, 1998, p. 119).

Glorification of the National Group and Attachment to the National Group We suggest that Adorno et al.’s (1950) pseudo patriotism, Kosterman and Feshbach’s (1989) nationalism, Staub’s (1997) blind patriotism, and Triandis and Gelfand’s (1998) vertical collectivism all express one mode of group identification (see also Roccas, Sagiv, Schwartz, Halevy, & Eidelson, 2006). We refer to this as glorification of the national group. Glorification is thus defined as viewing the national in-group as superior to other groups and having a feeling of respect for the central symbols of the group such as its flag, rules, and leadership. An individual who is highly identified in this sense believes that the in-group is better and more worthy than other groups and that group members should adhere to all the group’s rules and regulations and feels insulted if others do not show the utmost respect for the group’s symbols (Roccas et al., 2006). Similarly, we reason that Adorno et al.’s (1950) genuine patriotism, Kosterman and Feshbach’s (1989) patriotism, Staub’s (1997) conventional patriotism, and Triandis and Gelfand’s (1998) horizontal collectivism all express another mode of group identification. We refer to this as attachment to the national group. People who are highly identified in this sense define themselves in terms of their group membership and extend their self-concept to include the group. They feel emotionally attached to the group and want to contribute to it (Roccas et al., 2006). As already men-

tioned, attachment to the national group was captured in the identification items that were used in the studies examining groupbased guilt reviewed above (Branscombe, 2004; Doosje et al., 1998, 2004).

Glorification and Attachment Partly Overlap How do glorification and attachment relate to each other? The two modes of identification should be positively related because they both tap the common concept of identification with a group. Thus people who are strongly attached to their in-group should also tend to glorify it. The positive link between attachment and glorification is supported by studies that have found moderate positive relations between patriotism and nationalism (e.g., Karasawa, 2002; Li & Brewer, 2004) and between horizontal and vertical collectivism (Chiou, 2001). Note, however, that the two modes of identification are distinct from each other. Thus, it is possible to glorify the group without being particularly attached to it and it is possible to be attached to the group without necessarily glorifying every aspect related to the group. The latter mode of identification is of particular interest to us: It denotes a positive attachment to the group coupled with a critical approach to its possible shortcomings. In the terminology proposed by Staub (1989), people who complement their high affective attachment to their nation with low glorification are critical loyalists.

Glorification, Attachment, and Reactions to the InGroups’ Moral Infractions We suggest that distinguishing between the two modes of national identification may be the first step in resolving the paradox of identification and group-based guilt. Glorification of the national group involves the motivation to view the group in the best possible light. Inherent to this mode of identification is justification of the group’s acts and denial of any criticism toward the group (e.g. Staub, 1997). Thus, high glorifiers are likely to reject information that implies that the in-group has been involved in moral violations. In summary, high glorifiers are expected to react in the way hypothesized by Doosje et al. (1998) as characterizing all high identifiers. Attachment to the group expresses commitment to the group and inclusion of the group in one’s self-concept. This makes people who are strongly attached to their in-group particularly vulnerable to feeling morally responsible and distressed when exposed to possibly incriminating information on the group’s infractions. The predicted positive relations between attachment and groupbased guilt, however, might be eliminated by positive relations between attachment and glorification. Glorification may suppress these relations, with a net result of inconsistent relations between attachment to the national group and group-based guilt, as indeed has been found in the collective guilt literature. To reveal the predicted positive relations between attachment and group-based guilt, we need to control for the glorification effect. To reiterate, we predict that it is those who are attached to the nation without glorifying it who are the most susceptible to feelings of groupbased guilt.

IDENTIFICATION AND GROUP-BASED GUILT

Modes of Identification, Exonerating Cognitions, and Group-Based Guilt It is likely that relationships between attachment, glorification, and group-based guilt are affected by the way in which information regarding the in-group is attended to and interpreted. People are distressed when they perceive that their in-group deviates from grouplevel moral standards (e.g., Bizman, Yinon, & Krotman, 2001), and they employ creative cognitive mechanisms to protect their positive group identity. For example, people attempt to lower the severity of their group’s wrongdoings (Branscombe & Miron, 2004), they change the content of in-group and out-group stereotypes to justify discriminating an out-group (Rutland & Brown, 2001), and use external circumstances to explain negative historical actions carried out by their in-group to a greater extent than for similar actions committed by other nations (Doosje & Branscombe, 2003). Bandura (1999) systematically analyzed the multiple ways in which such mechanisms can lead to justification of wrongdoings. People can dispute the accuracy of the reports of harm done to the out-group, displace the responsibility for the harm done and attribute the offending behaviors to extenuating external circumstances rather than volition, or even interpret the event in a manner that assigns ultimate blame to the victims themselves (see Herbert & Dunkel-Schetter, 1992, for an example of the latter mechanism). All these mechanisms enable moral disengagement and have been related to support for military intervention against an enemy (Grussendorf, McAlister, Sandstrom, Udd, & Morrison, 2002; McAlister, 2001). In our study we focus on exonerating cognitions that minimize the perceived severity of moral violations committed by the ingroup. Past research indicates that using this legitimization strategy can decrease group-based guilt (Branscombe & Miron, 2004). How should glorification and attachment to the group be related to the use of exonerating cognitions? The relations between glorification and the use of such defensive cognitions seem rather straightforward. High glorifiers are likely to reject any negative interpretation of their in-group’s past deeds. Thus, glorification of the national group should be positively related to the use of exonerating cognitions. Admittedly, the relations between attachment to the group and the use of exonerating cognitions are less easily theoretically discerned. We, nevertheless, propose that attachment to the group (when glorification is controlled for) is negatively rather than positively related to the use of exonerating cognitions. Our reasoning relies on two lines of theory and research. First, we reason that being attached to the group and categorizing the self in terms of group membership implies high involvement in processing information about the group’s actions. Extensive research indicates that when people are involved in an issue, they process information more in depth, and the quality of information becomes a more important determinant of their attitudes (see Petty & Cacioppo, 1990). When the information is credible (as is the case in the current studies, in which true historical events are depicted) highly attached members may find themselves glued to the story more than their less attached counterparts. Furthermore, because of the perceived credibility of the story, it would be difficult to easily dismiss the events as false. Second, following Staub (1989), we reason that high attachment coupled with low glorification, which we have termed critical attachment, makes dismissal of negative

701

information about the group unlikely. In conclusion, we suggest that attachment to the group (when glorification is controlled for) is negatively rather than positively associated with the use of exonerating cognitions.

Study 1 We tested our model in the context of the Israeli–Palestinian conflict, which is rife with fierce violence by both parties. The study was conducted in two stages. The first sample was collected in a relatively calm period in the history of the Israeli–Palestinian conflict (January–April 2000), when many Israelis believed that the conflict was progressing toward a peaceful solution. Unfortunately, several months later, following the failed Camp David summit in the summer of 2000 and the outbreak of the al-Aqsa Intifada in October 2000, the Israeli–Palestinian conflict entered a new era of escalating and retaliatory violence that has claimed many lives on both sides. The belief in the resolvability of the conflict suffered a major blow.2 We thought that it would be worthwhile to examine the effects of the “back to conflict” atmosphere that took place following the continuation of the 2000 intifada on the retrospective assessments of the same past harmful in-group actions investigated in the first sample. Specifically, we sought to test whether the willingness to assume moral responsibility for a group’s past wrongdoing and experience group-based guilt was limited to circumstances in which the conflict was heading toward resolution. We explored these issues by replicating our study in the winter of 2001 (between October and December), 12⫹ months into the al-Aqsa Intifada. A major theoretical tenet in most psychological theories of intergroup conflict is that a perceived threat to the in-group (imaginary or real) leads to hardening of its positions toward the outgroup (for a recent review, see Stephan & Renfro, 2002). For example, some studies have revealed that outbreaks of hostilities coincide with the adoption of more negative stereotypes toward the out-group (e.g., Haslam, Turner, Oakes, McGarty, & Hayes, 1992) and that in-group conflict heightens current negative stereotypes and dehumanization of the enemy (Opotow, 1990; Staub, 1989). Thus, in Study 1 we examined how the intensification of the conflict affected identification, use of exonerating cognitions, and group-based guilt. We then tested our theoretical model depicting the relationships between glorification of the national group, attachment to it, use of exonerating cognitions, and feelings of group-based guilt.

2

The change in the Israeli public in viewing the chances for a peaceful solution to the Israeli–Palestinian conflict between the period in which we collected the first and the second sample is reflected in the Oslo Index, a monthly public opinion survey conducted by the Tami Steinmetz Center for Peace Research in Tel Aviv University. This index, which measures support for the Oslo process and belief in its positive conclusion, ranging from 100 (full support) to 0 (complete rejection), was 55 between January and April of 2000 when Study 1 was conducted and dropped to 32 between June and September of 2001 when Study 2 was conducted. The Oslo Index can be found at http://spirit.tau.ac.il/socant/peace/.

ROCCAS, KLAR, AND LIVIATAN

702

home from their jobs, they were stopped, lined up, and shot by the border patrol soldiers. 43 Arab residents were killed, among them 7 children and 9 women.

Method Participants and Procedure Participants were two samples of Jewish Israeli students. Data in Sample 1 (N ⫽ 216; 64% women, 36% men; age range ⫽ 19 –39, M ⫽ 23) were collected during a period of relative calm in the history of the Israeli– Palestinian conflict (January–April 2000). Data in Sample 2 (N ⫽ 165; 49% women, 51% men; age range ⫽ 18 –34, M ⫽ 23) were collected in late 2001, about a year after a renewal of hostilities between Israelis and Palestinians. Participants in both samples were approached at the end of class time and asked whether they would like to participate in the study for course credit or for a small payment (approximately $5). They completed the measures in groups of 20 – 40. The students were informed that participation in the study was voluntary and completely anonymous. Questionnaire completion took about 20 min. In Sample 2, the order of the identification, guilt, and exonerating cognition questionnaires was counterbalanced. Questionnaire order did not significantly affect means or correlations between the measures.

Instruments Participants received a booklet containing measures of identification with the national group (Israel), group-based guilt, and background questions.3 Identification with the national group. Eight items measured attachment to the national group and eight items measured glorification of the national group (see the Appendix). Respondents indicated their agreement with each of the items on a 7-point scale ranging from strongly disagree (1) to strongly agree (7). Sample items for attachment are “Being Israeli is an important part of my identity” and “I am strongly committed to my nation” (Cronbach’s ␣ ⫽ .89). Sample items for glorification are “Israel is better than other nations in all respects” and “One of the important things that we have to teach children is to respect our national leaders” (Cronbach’s ␣ ⫽ .78). The distinction between items measuring the two modes of identification was verified with a confirmatory factor analysis using maximum likelihood estimation. The eight attachment and eight glorification items were postulated to load on different latent factors. The analysis was conducted on this sample along with additional samples from Israel and from the United States. In both cultural groups the two-factor model of identification with the nation yielded adequate fit indices: ␹2(103, N ⫽ 484) ⫽ 374.323, comparative fit index (CFI) ⫽ .93, in the American sample; ␹2(206, N ⫽ 208) ⫽ 726.37, CFI ⫽ .87, in the Israeli sample (Roccas et al., 2006).4 Group-based guilt and exonerating cognitions. Participants read accounts of three historical events depicting harm done by Israelis to Palestinians. The events were presented in a factual paragraph indicating clearly that the harm resulted from an intentional act, without any explicit condemnation. Following is one of these events: On 29 October, 1956, with the opening of Operation Kadesh [the 1956 war between Israel and Egypt], a curfew was imposed in the Arab villages in the “triangle,” among these Kafir Kasem. The border patrol forces were responsible for enforcing the curfew in Kafir Kasem. The operational command that was given on that day stated that anyone disobeying the curfew should be shot with the aim to kill. The leader of the border patrol unit asked the commander of the decree what to do in the case of field workers who were unaware of the curfew and had not heard about it in the meantime. The commander replied in Arabic, “Allah yerachmo” [may God have mercy on them]. In a series of orders, the leader of the border patrol unit clarified that Arabs sitting in their homes were not to be injured, but that any person found outside of their home was to be shot dead. He said, “If there are a few casualties, it will only make enforcing the curfew easier in the coming days.” Towards evening, when the residents of the village returned

The two other historical events were the forced evacuation of Palestinians from the cities of Ramleh and Lidya (Lod) during the 1948 war and the 1994 killing of 29 Arab worshippers in the Tomb of the Patriarchs in Hebron by Baruch Goldstein, a Jewish settler from the West Bank. None of our participants had any personal involvement in any of the three historical events depicted in the studies—the earliest having taken place 52 years prior to the study and the most recent 6 years prior to it. Thus, these three events were part of the group histories of the participants rather than being part of their own personal history. Participants read the accounts of all three events. Then each account was presented again followed by a series of items on a 7-point scale ranging from do not agree at all (1) to completely agree (7). These items were used to construct the scales for exonerating cognitions and group-based guilt; each scale was the average of scores on its constituent items over the three events. By presenting participants with all three accounts before assessing their responses, we sought to minimize the effects of order and provide all participants with the same contextual anchoring. Exonerating cognitions. Seven items measured the use of exonerating cognitions in response to each of the three events. Sample items are “In my opinion, the description of the event is too harsh concerning the Israelis”; “Even if the event really happened the way it is described here, it pales in comparison to what the Arabs would have done to the Israeli side”; and “In my opinion, the Arabs brought the event upon themselves” (Cronbach’s ␣ ⫽ .92). Group-based guilt. Seven items measured willingness to assume moral responsibility for the harm done in response to each of the three events. Following previous research on group-based guilt (Doosje et al. 1998), we included items in our measure that tapped feelings of guilt directly and items measuring a more behavioral manifestation of guilt—willingness to compensate the victims. Sample items are “I feel guilty about what happened in the event depicted” and “I feel that as a result of the event depicted, I should help to improve the situation of Arabs” (Cronbach’s ␣ ⫽ .94).

Results Intensification of the Conflict, Modes of Identification, Exonerating Cognitions, and Group-Based Guilt Before testing the hypotheses regarding the relationships of different modes of identification with exonerating cognitions and group-based guilt, we examined whether the intensification of the conflict that occurred in the period between the times we collected responses from the two samples affected our core variables. Means of glorification, attachment, exonerating cognitions, and guilt in the low conflict phase (Sample 1) and high conflict phase (Sample 2) are presented in Table 1. The effects of the intensity of the conflict were tested with a multivariate analysis of variance, with the time of measurement (low conflict in early 2000 and high conflict in winter 2001) as the between-participants’ variable. This analysis revealed a multivariate significant main effect, F(1, 376) ⫽ 5.58, p ⬍ .001. At the univariate level, the effect was significant for group-based guilt, F(1, 379) ⫽ 9.53, p ⬍ .005, and 3

A total of 107 participants also reported their personal value priorities. In both cultural groups, the fit indices of the two-factor model were better than those resulting from a single-factor model and somewhat worse than those resulting from a four-factor model that included finer distinctions within glorification and attachment not relevant to the present study. 4

IDENTIFICATION AND GROUP-BASED GUILT

Table 1 Means and Standard Deviations for Attachment, Glorification, Exonerating Cognitions, and Collective Guilt in the Low and High Conflict Samples Low conflict sample

High conflict sample

Variable

M

SD

M

SD

Attachment Glorification Exonerating cognitions Collective guilt

5.42 3.45 2.59 4.29

1.07 0.91 0.90 1.21

5.25 3.39 2.94 3.90

1.07 1.00 0.95 1.24

for the use of exonerating cognitions F(1, 379) ⫽ 13.02, p ⬍ .001. As expected, participants in the high conflict phase reported lower levels of guilt compared with participants in the low conflict phase (high conflict: M ⫽ 3.90; low conflict: M ⫽ 4.29), and they made greater use of exonerating cognitions (high conflict: M ⫽ 2.94; low conflict: M ⫽ 2.59). There were no differences in attachment and glorification, F(1, 379) ⫽ 2.37, and F(1, 379) ⫽ .37, respectively, ps ⬎ .10.

Relations Between Glorification, Attachment, Exonerating Cognitions, and Group-Based Guilt The first step in studying the relations between glorification, attachment, and group-based guilt was exploring the zero-order correlations between these variables. Consistent with the notion that glorification of the group and attachment to the group partly overlap, these two concepts were positively related in both samples (rs ⫽ .57 and .52, ps ⬍ .001, in Samples 1 and 2, respectively). Also, as predicted, glorification was negatively correlated to group-based guilt (rs ⫽ ⫺.45 and ⫺.54, ps ⬍ .01, in Samples 1 and 2, respectively), whereas attachment was not correlated to guilt (rs ⫽ .09 and .03, ns, in Samples 1 and 2, respectively). Consistently, glorification was positively related to exonerating cognitions (rs ⫽ .46 and .52, p ⬍ .001, in Samples 1 and 2, respectively), whereas the correlations of attachment with exonerating cognitions were near zero (r ⫽ .14, p ⬍ .05, in Sample 1; r ⫽ .13, ns, in Sample 2). To test our model in full, we conducted a structural equation modeling (SEM) analysis using EQS Version 6 (Bentler, 2002), combining results of the two samples (see Figure 1). Each of the latent variables (attachment, glorification, exonerating cognitions, and group-based guilt) had two indicators. The two indicators were computed by splitting the items related to each latent variable into two sets: even versus odd numbered items. The structural model fitted the data well, ␹2(14, N ⫽ 381) ⫽ 43.35, p ⬍ .001, CFI ⫽ .98, normed fit index (NFI) ⫽ .98, nonnormed fit index (NNFI) ⫽ .97, standardized root-mean-square residual (SRMR) ⫽ .033. The independent variables explained 25% of the variance in groupbased guilt. Consistent with hypotheses, the path from glorification to exonerating cognitions was positive (␤ ⫽ .74, p ⬍ .05), the path from attachment to exonerating cognitions was negative (␤ ⫽ ⫺.39, p ⬍ .05), and the path from exonerating cognitions to guilt was negative (␤ ⫽ ⫺.41, p ⬍ .05). Indirect (mediated by the exonerating

703

cognitions) effects of both glorification and attachment on guilt were significant ( p ⬍ .05). The direct path (unmediated by the exonerating cognitions) from attachment to guilt was positive (␤ ⫽ .25, p ⬍ .05). The direct path from glorification to guilt was nonsignificant (␤ ⫽ ⫺.19), indicating complete mediation by the exonerating cognitions. All in all, the SEM results corroborated the theoretical model.5 Next, we conducted a series of analyses to examine finer points. We took a deeper look at our measure of group-based guilt and examined its emotional and behavioral components separately. As already indicated, following Doosje et al. (1998), we used a compound measure of group-based guilt that included items that directly tapped emotion and items that tapped the behavioral manifestation of guilt: willingness to compensate the victims. Past research indicates that feelings of guilt are closely related to willingness to compensate the victims. For example, Iyer, Leach, and Crosby (2003) found that among White people, feelings of guilt over racial discrimination and White privilege were related to support of policies designed to compensate African Americans. Similarly, in our sample, the emotional manifestation and the behavioral manifestation of guilt were highly correlated (r ⫽ .77). However, the emotional and the behavioral manifestations of guilt are theoretically distinguishable. We therefore sought to examine whether they were similarly related to exonerating cognitions, attachment, and glorification. We repeated the SEM analyses for each manifestation of guilt separately. The results were very similar across the two measures. Emotional guilt. The model fitted well to the data, ␹2(14, N ⫽ 381) ⫽ 34.25, p ⬍ .002, CFI ⫽ .99, NFI ⫽ .98, NNFI ⫽ .97, SRMR ⫽ .031. The independent variables in the model explained 33% of the variance in group-based guilt. The path from glorification to exonerating cognitions was positive (␤ ⫽ .89, p ⬍ .05), the path from attachment to exonerating cognitions was negative (␤ ⫽ ⫺.47, p ⬍ .05), and the path from exonerating cognitions to guilt was negative (␤ ⫽ ⫺.56, p ⬍ .05). Indirect (mediated by the exonerating cognitions) effects of both glorification and attachments on guilt were significant ( ps ⬍ .05). The direct path (unmediated by the exonerating cognitions) from attachment to guilt was positive (␤ ⫽ .21, p ⬍ .05). The direct path from glorification to guilt was near zero (␤ ⫽ ⫺.02), indicating complete mediation by the exonerating cognitions. Behavioral guilt. The model fitted well to the data, ␹2(14, N ⫽ 381) ⫽ 57.47, p ⬍ .001, resulting in the following fit indices, CFI ⫽ .97, NFI ⫽ .97, NNFI ⫽ .95, SRMR ⫽ .048. The model explained 31% of the variance in group-based guilt. The path from glorification to exonerating cognitions was positive (␤ ⫽ .87, p ⬍ .05), the path from attachment to exonerating cognitions was negative (␤ ⫽ ⫺.47, p ⬍ .05), and the path from exonerating cognitions to guilt was negative (␤ ⫽ ⫺.56, p ⬍ .05). Indirect (mediated by the exonerating cognitions) effects of both glorification and attachment on guilt were significant ( ps ⬍ .05). The direct 5

We examined the relationships between attachment, glorification, exonerating cognitions, and guilt separately in the two samples with hierarchical regressions. Results were very consistent across the two samples. Thus, the two modes of identification were similarly related to exonerating cognitions and to group-based guilt despite the changes in sociopolitical context.

ROCCAS, KLAR, AND LIVIATAN

704 Attachment

.25 -.39

.68 .74 Glorification

Exonerating Cognitions R2 = .31

-.41

Guilt R2 = .25

-.19

Figure 1. Structural equation model of the relationships of attachment, glorification, exonerating cognitions, and group-based guilt (standardized parameter estimates). Solid lines represent statistically significant paths ( p ⬍ .05). The dotted line represents a path that is not statistically significant. Measured indicators and their errors are not shown. Both R2 coefficients are statistically significant at p ⬍ .05.

paths (unmediated by the exonerating cognitions) from attachment to guilt and from glorification to guilt were near zero (␤s ⫽ .03 and ⫺.01, respectively), indicating complete mediation by the exonerating cognitions. In summary, the relationships between the two modes of identification, exonerating cognitions, and group-based guilt were consistent across the emotional and behavioral manifestations of guilt.

Discussion Effects of Conflict Intensification on Perception of Historical Events The resurgence of conflict brought with it some defensive revisions in the assessment of the in-group’s past wrongdoings. In our study, participants reacted to events that had occurred long ago: the most remote 52 years and the most recent 6 years before the first study. Nonetheless, the resurgence of violence apparently affected perceptions of these past events. These results suggest that in periods of heightened conflict, not only are group members likely to justify in-group acts that would be deemed unjustifiable in periods of relative calm, but they are also more likely to withhold moral condemnation of harmful acts committed by the in-group in the past. The high match between the two cohorts used in the studies, both taken from the same participant pool, makes any attribution of the results in terms of possible differences between the two samples rather unlikely. More probable explanations are that the decrease in group-based guilt and the increase in exonerating cognitions reflect some real temporal changes due to the intensification of the conflict. There are several possible mediating mechanisms that could explain how intensification of the conflict causes a decrease in group-based guilt. For example, conflict may lead people to focus on the in-group and its needs rather than on the out-group’s anguish, or it may lead to a reassessment of the history of the conflict in the light of current events. Further research is needed to test these possible mediating mechanisms.

The Two Modes of Identification With the Group, Exonerating Cognitions, and Group-Based Guilt Study 1 was set up to explore the paradoxical relations between identification with the national group and the extent of group-

based guilt. We used a two-mode conceptualization of identification in which identification with the national group is composed of glorification of the national group and attachment to it, two partially overlapping tendencies. We found that these two modes of identification were indeed positively related with each other. We also found that glorification of the national group was positively associated with the use of exonerating cognitions and negatively associated with experiencing group-based guilt. Attachment to the national group, however, showed a different pattern of results. Before controlling for glorification (i.e., in the zero-order correlations), attachment was positively related to the use of exonerating cognitions and unrelated to group-based guilt. But when glorification was controlled for (in the SEM analyses), attachment switched to become negatively related to exonerating cognitions and positively related to group-based guilt. Thus being attached to the national group (when glorification is controlled for) increases rather than decreases group-based guilt. Similarly, high attachment, when separated from glorification, is negatively associated with the use of exonerating cognitions, and the positive link between attachment and guilt is at least partly mediated by the failure to use exonerating cognitions. In summary, separating attachment from glorification served to reveal the effects of critical attachment. Drawing on Staub (1997) and on Petty and Cacioppo (1990), we reason that people who are critically attached to the group do not endorse exonerating cognitions and consequently are likely to feel guilty for the group’s wrongdoings. In Study 2 we examined more closely the distinction between critical and conventional attachment to the group. What about the direction of causality? Does the reaction to the in-group’s past moral violations determine the level of identification with the group rather than the other way around as argued in this article? Obviously, this direction of causality, in its broader sense, cannot be denied. There are historical examples of individuals who completely cut off their ties with their group in response to their group’s wrongdoings. In this case, group-based guilt reduces identification with the group. There are, however, a number of reasons to conclude that the current results cannot properly be explained by the “group-based guilt drives nonidentification” direction. First, reversing the causal direction between groupbased guilt and the two measures of identification would require explaining how increased group-based guilt simultaneously decreases and increases identification with the group. More specifi-

IDENTIFICATION AND GROUP-BASED GUILT

cally, it would have to both decrease glorification of the group (which makes theoretical sense) but also, at the same time, increase attachment to the group (which makes less theoretical sense). Similarly, reversing the causal direction would require explaining how the use of exonerating cognitions both increases glorification of the group (which makes theoretical sense) but also decreases attachment to the group (which makes less theoretical sense). Second, reversing the causal direction makes it difficult to explain the temporal pattern of results found in the two samples. As already discussed, we found that group-based guilt regarding past wrongdoings declined from the first (low conflict) to the second (high conflict) sample; however, levels of glorification and attachment were unchanged. This lack of change is also inconsistent with a direction of causality in which a decrease in groupbased guilt would result in increased identification with the group. Nonetheless, an experimental study might be appropriate to further test the causal path from identification to guilt.

Study 2 Of particular interest in our studies is the notion of critical attachment—attachment with relatively low-level glorification. In Study 1, we captured this conjunction by statistically controlling for glorification when examining the relationship between attachment and group-based guilt. In these studies, the zero-order correlation of attachment to guilt was nearly zero. The correlation of attachment to group-based guilt became positive only after glorification was controlled for. This is consistent with our reasoning that critical attachment is positively related to group-based guilt. In Study 2 we adopted an experimental approach to further examine this issue. Teasing apart the effects of attachment and glorification poses a methodological challenge. The two modes of identification are conceptually similar (both refer to identification with a group) and empirically positively correlated. Thus, a manipulation that elicits one mode of identification is likely to elicit the other mode as well. Because glorification suppresses the effects of attachment on identification, a manipulation that elicits attachment and glorification simultaneously would be powerless to test our core hypothesis that attachment without glorification causes group-based guilt. We thus sought to induce critical attachment directly and to compare its effects on group-based guilt with those of induction of conventional attachment (attachment intermixed with glorification). As already noted, Schatz et al. (1999) described constructive patriotism (i.e., critical attachment) as the individual’s readiness to “criticize and even actively oppose the nation’s action because he or she believes they violate fundamental national percepts that are contrary to long-term national interests” (p. 153). Conventional patriotism (i.e., conventional attachment), on the other hand, has been described by Schatz et al. simply as a feeling of affective attachment to the nation. Thus, a shift from a state of mind of conventional attachment to that of critical attachment should be followed by an increased readiness to criticize the actions of one’s nation that are perceived as a violation the nation’s long-term interests and fundamental values, and consequently by increased susceptibility to feelings of group-based guilt. The question is, however, how these two distinct modes of attachment can be induced in the laboratory.

705

To approach this goal, we were conceptually assisted by the distinction in self-concept literature between how people currently conceive themselves and how they wish to conceive themselves, that is, between the actual and the ideal self (Higgins, 1989). In a similar vein, when group members think about their attachment to their nation, they can mentally focus on attributes of their nation “as it is” or on the attributes of the nation as they “ideally would like it to be.” Research in the personal domain demonstrates that focusing on ideal self-guides may foster awareness of shortcomings of the individual’s current self and this may lead to personal distress and a desire to self-improve (Higgins, 1989, 1999). Discrepancies between current and ideal representations of one’s in-group have not been examined extensively. A notable exception is Bizman et al.’s (2001) research that indicates that people are more distressed the greater the extent to which they perceive their in-group to differ from their ideal image. We suggest that when the target of attachment is the nation as it ideally should be rather than the nation as it currently is, awareness of the nation’s current moral inadequacies and resultant group-based guilt should be higher. The first goal of Study 2 was to test this prediction. The second goal was to test these predictions in the context of the group’s current actions and policies. In Study 1, participants were asked to react to past events from the chronicles of the Israeli–Palestinian conflict, after being explicitly reminded about these events. In Study 2, we were interested in reactions to a more contemporary issue: the treatment of Israeli Arab citizens of Israel. Studying moral reactions to the in-group’s current rather than past wrongdoings is important because these current issues may lead to relevant social action. About 20% of Israeli citizens are Arabs (Palestinians). The treatment of this minority by the dominant Jewish majority raises many thorny issues of individual and collective rights. It is often argued that Israeli Arabs are discriminated against in many areas, such as land dispossession and allocations, education, employment, and political participation. Poverty and unemployment are substantially higher among members of this group than among the Jewish majority. Moreover, Israeli Arabs are often treated with suspicion and hostility by many majority members and viewed as being disloyal and antagonistic to the State of Israel (for discussions of these issues see, e.g., Ganim & Ghanem, 2001; Landau, 1993; Rouhana, 1998). We tested the hypothesis that thinking of the attachment to the in-group as it ideally should be leads participants to increase group-based guilt about in-group transgressions as compared with simply thinking about attachment to the in-group as it is.

Method Participants were Jewish Israeli students (N ⫽ 89; 80% women, 20% men; age range ⫽ 20 –35, M ⫽ 24) taking part in the study for course credit. They were randomly assigned to one of two groups: conventional and critical attachment.

Instruments Participants received a booklet containing the experimental manipulation followed by a measure of group-based guilt and background questions.

Manipulation In both conditions participants were asked to list attributes of the Israeli in-group. In the conventional attachment (actual in-group) condition the

ROCCAS, KLAR, AND LIVIATAN

706

instructions were designed to heighten the salience of those characteristics of Israel that participants view as determining their attachment to it. We used the following instructions: “Please briefly describe the attributes of Israel that prompt you to agree with the following sentence: ‘I love Israel and viewing myself as Israeli is important to me.’” In the critical attachment (ideal in-group) condition, participants were again instructed to list characteristics of Israel. But this time the phrasing of the instructions was designed to heighten the saliency of their conception of the ideal Israel: “Please briefly describe the attributes you would like to find in Israel in order to agree with the following sentence: ‘I love Israel and viewing myself as Israeli is important to me.’” Note that the tasks in the two experimental conditions were very similar: Participants listed those characteristics of Israel they associated with their attachment. The two statements used in both conditions (“I love Israel” and “viewing myself as an Israeli is important to me”) were taken from the attachment scale used in Studies 1 and 2. However, the two experimental conditions differed in their focus: In the conventional attachment condition, participants focused on the actual attributes of Israel; in the critical attachment condition, participants focused on desired or ideal attributes of Israel. In both cases we expected that participants would list positive attributes of the Israeli group, but we expected these attributes to be different in content and to refer to the actual group in the former condition and to the ideal group in the latter condition. To ascertain that thinking about one’s attachment to the ideal image of Israel lowers glorification, we examined the responses of 30 additional students who completed either the ideal (critical attachment) or the actual (conventional attachment) versions of the task and then completed the measure of glorification. Participants in the ideal condition (M ⫽ 3.27, SD ⫽ 0.58) expressed less glorification than those in the conventional attachment condition (M ⫽ 4.15, SD ⫽ 0.65), t(28) ⫽ 3.94, p ⬍ .005, confirming the effectiveness of the manipulation.

Group-Based Guilt Ten items measured feelings of guilt for harm done to the Arab minority in Israel. Sample items are “I believe that Israel treats the Arab minority unjustly,” “I believe that Israel should compensate Arabs citizens,” and “I feel guilty when I hear on the news about the bad conditions of Arab citizens” (Cronbach’s ␣ ⫽ .89). Given that issues related to the treatment of the Arab minority in Israel are often on the public agenda, we thought that unlike in our previous studies that dealt with the past, it would be unnecessary to provide our participants with any factual information about this matter.

Results and Discussion The Group Attributes Engendered in the Two Experimental Conditions We first analyzed the contents of the group attributes associated with attachment to Israel in the two experimental conditions. This was done to test whether the ideal versus actual manipulation produced two distinct sets of group attributes. Overall, participants wrote 191 different attributes. Two coders, blind to the experimental condition under which the attribute originated and to the research hypotheses, independently derived content categories from the list. Agreement between the two coders was 94%. As can be seen in Table 2, there was almost no overlap between the attributes mentioned by participants in the two experimental conditions. Participants in the conventional attachment condition portrayed their image of the current attributes of their group. They mentioned positive attributes such as in-group solidarity, geography and folklore, and Judaism. These attributes are related to the perceived meaning and traits of being an Israeli. Those in the critical attachment condition were prompted to list the attributes in which the current image of their in-group is discrepant from the ideal image. They mentioned ideal–actual discrepancies in areas such as interpersonal relations, welfare state, quality of governance, and peace and security. These discrepancies are related to hopes and desires for a better society in which the welfare of all is taken into account.

Group-Based Guilt in the Two Experimental Conditions To test the hypothesis that priming critical attachment induces stronger group-based guilt than priming conventional attachment, we compared the mean level of group-based guilt associated with Israel’s current treatment of its Arab minority in the two groups. As predicted, the mean score of group-based guilt was significantly higher in the critical attachment condition (M ⫽ 3.57, SD ⫽ 0.78) than in the conventional attachment condition (M ⫽ 3.24, SD ⫽ 0.74), t(87) ⫽ 2.05, p ⬍ .05. The results of Study 2 provide some insight into how critical attachment enhances group-based guilt: Contemplating attachment

Table 2 Attributes of Israel Mentioned by Participants in the Conventional and Critical Attachment Experimental Conditions Conventional attachment Category

No. of mentions

Solidarity

13

Positive attributes of Israelis

20

Geography and folklore

18

Links to Judaisma

33

Critical attachment

Example attributes

Category

No. of mentions

Solidarity in hard times; people care for each other People are warm; people are openminded Good food and warm weather; East and West are intermixed

Quality of governance

15

Peace

16

Welfare state

18

Social justicea Positive interpersonal relations

30 10

The only Jewish state; Jewish heritage

Example attributes Clean government; truthful government Peace with our neighbors; ending the occupation A socioeconomic system that takes care of everybody; a health system Care for minorities; equal rights More politeness; more civility

Note. Ten attributes in the conventional attachment condition and 15 in the critical attachment condition did not correspond to any of the listed categories. An attribute in this category was also mentioned once in the other experimental condition.

a

IDENTIFICATION AND GROUP-BASED GUILT

to the in-group as it ideally should be brings to mind “group basic values,” “basic human values,” and “group interests in the long run” (Staub, 1997, p. 214). In fact, our critical attachment participants mainly listed long-term group goals and values such as social justice, equality, peace, and quality of governance. These abstract long-term goals express universalistic values and interest in the welfare of all people (Schwartz, 1992, 1994). Thus, longterm universalistic goals became more salient for participants in the critical attachment condition than for those in the conventional attachment condition. Subsequently, when asked about the possibility that their group is treating members of an internal minority group unfairly, actions which violate universalistic goals, critical attachment participants were more inclined to experience moral distress than were conventional attachment participants. Stated differently, critical attachment to the group can be regarded as moral attachment. It is noteworthy that the “ideal group” manipulation affected group-based guilt without any explicit appeal to criticize the group’s current actions. Both critical and conventional attachment participants listed positive attributes of their national group. However, mere contemplation of the ideal (rather than actual) group as the target of attachment raised group-based moral responsibility. These findings may have practical implications for persuasive attempts intended to encourage people to oppose moral violations committed by their group, especially at times of open and violent intergroup conflict. Often group members who criticize their group’s actions are seen as being disloyal to the group, and consequently, their criticism is dismissed or is not influential. It follows that instead of directly criticizing the group’s behaviors, persuasive attempts can encourage group members to view their attachment to their group in terms of its fundamental long-term goals and values.

General Discussion In these two studies we sought to achieve a better understanding of group-based guilt in the context of an ongoing violent conflict. Study 1 was conducted during two phases of the Israeli– Palestinian conflict: a period of relative calm and a period of violent escalation. For both these times we examined reactions to information indicating that the in-group had harmed members of another group, enabling us to explore the effects of intensification of the conflict. In both periods participants expressed moral condemnation of the in-group’s past wrongdoings. The intensity of the conflict, however, had a clear effect: Endorsement of cognitions exonerating the in-group was higher and group-based guilt was lower in the sample examined during the escalation of the conflict. This is consistent with other field studies indicating that outbreaks of hostilities between countries coincide with the adoption of more negative national stereotypes (e.g., Haslam et al., 1992). The current results extend these findings by showing that escalation of a conflict also moves group members away from acknowledging their group’s wrongdoings in the past. A major goal of the current investigation was to clarify the seemingly paradoxical relations between identification with the national group and group-based guilt. On the one hand, identification is a prerequisite for feeling personal responsibility for the in-group’s harmful acts. On the other hand, identification leads to

707

viewing the group in the best possible light and hence it is associated with the rejection of group guilt (Doosje et al., 1998). The solution to this paradox, pursued in this article, is based on integrating insights from several diverse approaches to group identification: the nationalism–patriotism distinction in political psychology (e.g., Kosterman & Feshbach, 1989; Staub, 1997), the horizontal–vertical collectivism distinction in cross-cultural psychology (Triandis & Gelfand, 1998), and the social identity perspective (e.g., Brown, 2000). On the basis of these diverse theoretical perspectives, we suggest a distinction between two modes of identification: glorification and attachment. The core finding across our studies was that the two partly overlapping modes of identification have opposing relations to reactions to information about group transgressions: Glorification of the in-group is associated with lower feelings of guilt. Attachment to the group is seemingly unrelated to guilt. But when the glorification component is controlled for, attachment is consistently associated with stronger feelings of guilt. Thus it is attachment with low level of glorification—referred here as critical attachment—that is conducive to group-based guilt. These findings were replicated across two phases of the conflict. In the United States, Schatz et al. (1999) measured this critical mode of identification (i.e., attachment without glorification) directly with items such as “If I criticize the United States, I do so out of love for my country” and “I express my love for America by supporting efforts at positive change.” It might be argued that these are compound items that simultaneously probe extent of attachment and willingness to criticize the in-group. We supplement this methodology in two ways. In Study 1 we measured attachment and glorification. Then we isolated attachment without glorification by statistically controlling glorification scores when examining the effects of attachment. In Study 2 we experimentally induced attachment without glorification by focusing the participants’ attachment on the way the group should ideally be rather than on the current image they have of the group. Again findings supported our reasoning that it is attachment with low glorification that is most conducive to group-based guilt.

The Role of Exonerating Cognitions Exonerating cognitions were strongly negatively related to feelings of guilt. This finding is consistent with recent research that has measured or manipulated use of exonerating cognitions, showing their influence on guilt (Miron, Branscombe, & Schmitt, 2006). Thus there is growing evidence that such cognitions are related to moral disengagement from immoral acts committed by one’s in-group. This finding is of importance because it could point to possible ways to help conflicting groups in their path toward reconciliation. So far little research has been done on resistance to moral disengagement, but an experiment on this topic led to encouraging results: McAlister (2001) showed that hearing brief communications about the meaning of moral disengagement versus communications in support of military action affected opinions about U.S. bombing campaigns in Iraq and Yugoslavia. Thus people can be made aware of the significance of communications that encourage moral disengagement from the in-group’s wrongdoings and lead them to greater resistance to accepting acts that conflict with their own moral values.

708

ROCCAS, KLAR, AND LIVIATAN

In our studies, glorification and attachment effects on groupbased guilt were mediated by the use of exonerating cognitions. Glorification was positively related to the use of exonerating cognitions. Thus, people who glorify their in-group (e.g., blind patriots, nationalists) can avoid feeling morally responsible for their in-group’s wrongdoings because they interpret the harmful events in ways that justify the in-group’s actions, just as suggested by Doosje et al. (1998). In contrast, attachment to the in-group does not lead to such group-enhancing cognitions. Attachment (when glorification was controlled for) was correlated to lower justification of the in-group’s transgressions. This relationship suggests that those who are cognitively and emotionally involved with their group are to some extent also concerned with its moral standing and are reluctant to justify moral violations on the part of in-group members.

Mapping Glorification and Attachment to Definitions of Identification Stemming From the Social Identity Perspective The social identity perspective (including social identity theory and self-categorization theory) is the theoretical psychological framework in which group processes are most intensely studied. How does our multidimensional model of identification map with current research in the social identity perspective? Consistent with social identity theory and self-categorization theory (Tajfel & Turner, 1986; Turner, 1999), we view national identification as a process of depersonalization that emphasizes commonalities among members of a nation and bases connections between individuals on their collective identity rather than on interpersonal ties. Tajfel (1978) defined social identity as “that part of an individual’s self-concept which derives from his knowledge of his membership in a group together with the value and emotional significance attached to the membership” (p. 63). This definition has been interpreted as being made up of three aspects— cognitive, affective, and evaluative. There have been many empirical attempts to differentiate these three aspects with variable degrees of success (e.g., Brown & Williams, 1984; Ellemers, Kortekaas, & Ouwerkerk, 1999; Hinkle, Taylor, Fox-Cardamone, & Crook, 1989; Jackson, 2002; Karasawa, 1991). In line with social identity theory, we view identification as multifaceted. Our concept of attachment to the group is similar to the cognitive and affective components in Tajfel’s (1978) definition. People who are highly attached to their nation attribute high importance to their national identity as part of their self-concept (cognitive aspect), love the group, and wish to contribute to its welfare (the affective component). Our concept of glorification is similar to the evaluative component described by Tajfel because it refers to the belief that one’s nation is better and more worthy than other nations. But it also includes allegiance to the group’s symbols and authority. To the best of our knowledge, no previous study has examined the associations of identification and group-based guilt using a multifaceted approach to identification. Studies stemming from the social identity perspective have found, however, that different dimensions of identification are differentially related to attitudes toward the out-group: For example, Ellemers et al. (1999) found that commitment alone was related to in-group favoritism in the context of a laboratory created group. Jackson (2002), examining

real-life groups, found that affective ties to the in-group were more strongly correlated to in-group bias than attraction to the in-group and extent of self-categorization. Thus, these studies support our reasoning that distinguishing between different modes of identification is useful in seeking a better understanding of intergroup relations.

Incongruent Patterns of Identification The positive correlation between glorification and attachment suggests that many people have a congruent pattern of identification with the national group: They are high or low on both attachment and glorification. Indeed, out of the 381 participants in Study 1, 120 were above the median on both modes of identification, and 121 were below the median on both modes of identification. However incongruent patterns of identification are not rare: In our sample 62 participants were above the median on glorification but below the median on attachment, and 61 were below the median on glorification but above the median on attachment. In this article we have elaborated on the meaning of being attached to the group and yet having low levels of glorification. We now turn to clarifying the meaning of glorifying the group and yet having low attachment to it. We are not aware of any empirical study that has focused on the characteristics of people with this identification profile. Research on social dilemmas, however, can contribute to its understanding. In an analysis of the prototypical problems of cooperation and competition within and between groups, Bornstein (2003) noted that pride in the group is a public good that is available to all members of a group. Any group member can think that his group is superior to other groups regardless of his willingness to contribute to the group. Bornstein further noted that group members may have an incentive to “free ride” on the contributions of others: They feel proud of the group, without feeling committed to contributing to its welfare. From this perspective, a person who glorifies the group without being attached to it is a free rider: He or she derives a positive social identity from the perceived group superiority without paying the price inherent to a strong commitment to it. The concept of basking in reflected glory (Cialdini et al., 1976) may exemplify the free rider nature of a person who derives positive identity from an in-group’s success without showing consistent attachment to the group. In a series of field studies, Cialdini et al. (1976) showed that some sports fans are happy to support group symbols following their team’s success but rescind their identification following the team’s failure. Going back to identification with national group, a person who expresses high glorification of the nation accompanied by low attachment can be typified as wishing to benefit from the perceived superiority of the group without being willing to commit to the group. More research is needed to determine the antecedents and consequences of incongruent modes of identification.

Glorification, Attachment, and Moral Outrage To return to the consequences of the different modes of identification, it is worth inquiring whether glorification and attachment affect only the sense of group-based guilt. How do these two modes of identification relate to emotions toward members of the

IDENTIFICATION AND GROUP-BASED GUILT

in-group who committed moral infractions? We are currently investigating these issues. In our initial study (Klar, Roccas, & Liviatan, 2006), we found results similar in direction to those found in the studies reported here. We presented our Jewish Israeli participants with contrived newspaper headlines (but similar to real events taking place during the armed conflict between Israelis and Palestinians) such as “Israelis [Palestinians] prevented the passage of an ambulance carrying a severely injured woman.” We assessed moral outrage directed at the offenders when the same act was committed by the in-group (in one experimental condition) or the out-group (in another experimental condition). Glorification was positively related to moral outrage toward out-group perpetrators and negatively related to moral outrage toward in-group offenders. Attachment, on the other hand, was negatively related to moral outrage toward the out-group perpetrators and positively to the amount of outrage toward in-group perpetrators. Thus, just as guilt about the in-group’s wrongdoings has opposite effects, glorification and attachment may have opposite effects in relating to moral outrage toward members of the in-group who acted wrongly. In conclusion, national identification is a multifaceted construct. Although the different modes of identification are highly correlated, they can have opposing relationships with group-related phenomena.

References Adorno, T., Frankel-Brunswik, E., Levinson, D., & Sanford, N. (1950). The authoritarian personality. New York: Harper. Augoustinos, M., & LeCouteur, A. (2004). On whether to apologize to Indigenous Australians: The denial of White guilt. In N. R. Branscombe & B. Doosje (Eds.), Collective guilt: International perspectives (pp. 236 –261). New York: Cambridge University Press. Bandura, A. (1999). A social cognitive theory of personality. In L. Pervin & O. John (Eds.), Handbook of personality (2nd ed., pp 154 –196). New York: Guilford. Barkan, S. E., & Snowden, L. S. (2001). Collective violence. Boston: Allyn & Bacon. Bar-Tal, D. (1998). Societal beliefs in times of intractable conflict: The Israeli case. International Journal of Conflict Management, 9, 22–50. Bentler, P. M. (2002). EQS 6 structural equations program manual. Encino, CA: Multivariate Software. Bizman, A., Yinon, Y., & Krotman, S. (2001). Group-based emotional distress: An extension of self-discrepancy theory. Personality and Social Psychology Bulletin, 27, 1291–1300. Bornstein, G. (2003). Intergroup conflict: Individual, group, and collective interests, Personality and Social Psychology Review, 7, 129 –145. Branscombe, N. R. (2004). A social psychological process perspective on collective guilt. In N. R. Branscombe & B. Doosje (Eds.), Collective guilt: International perspectives (pp. 320 –334). New York: Cambridge University Press. Branscombe, N. R., & Doosje, B. (2004). International perspectives on the experience of collective guilt. In N. R. Branscombe & B. Doosje (Eds.), Collective guilt: International perspectives (pp. 3–15). New York: Cambridge University Press. Branscombe, N. R., Doosje, B., & McGarty, C. (2002). Antecedents and consequences of collective guilt. In D. M. Mackie & E. R. Smith (Eds.), From prejudice to intergroup emotions: Differentiated reactions to social groups (pp. 49 – 66). Philadelphia: Psychology Press. Branscombe, N. R., & Miron, A. M. (2004). Interpreting the ingroup’s negative actions toward another group: Emotional reactions to appraised harm. In L. Z. Tiedens & C. W. Leach (Eds.), The social life of emotions (pp. 314 –335). New York: Cambridge University Press.

709

Branscombe, N .R., Slugoski, B., & Kappen, D.M. (2004). The measurement of collective guilt: What it is and what it is not. In N. R. Branscombe & B. Doosje (Eds.), Collective guilt: International perspectives (pp. 16 –34). New York: Cambridge University Press. Brown, R. (2000). Social identity theory: Past achievements, current problems and future challenges. European Journal of Social Psychology, 30, 745–778. Brown, R. J., & Williams, J. (1984). Group identification: The same thing to all people? Human Relations, 37, 547–564. Chiou, J. (2001). Horizontal and vertical individualism and collectivism among college students in the United States, Taiwan, and Argentina. Journal of Social Psychology, 141, 667– 678. Cialdini, R., Borden, R., Thorne, A., Walker, M., Freeman, S., & Sloan, L. (1976). Basking in reflected glory; Three (football) field studies. Journal of Personality and Social Psychology, 34, 366 –375. Cohen, S. (2001). States of denial: Knowing about atrocities and suffering. Cambridge, England: Polity. Doosje, B., & Branscombe, N. R. (2003). Attributions for the negative historical actions of a group. European Journal of Social Psychology, 33, 235–248. Doosje, B., Branscombe, N. R., Spears, R., & Manstead, A. S. R. (1998). Guilty by association: When one’s group has a negative history. Journal of Personality and Social Psychology, 75, 872– 886. Doosje, B. J., Branscombe, N. R., Spears, R., & Manstead, A. S. R. (2004). Consequences of national ingroup identification for responses to immoral historical events. In N. R. Branscombe & B. J. Doosje (Eds.), Collective guilt: International perspectives (pp. 95–111). New York: Cambridge University Press. Ellemers, N., Kortekaas, P., & Ouwerkerk, J. (1999). Self categorization, commitment to the group and social self esteem as related but distinct aspects of social identity. European Journal of Social Psychology, 28, 371–398. Ganim, A., & Ghanem, A., (2001). The Palestinian-Arab minority in Israel, 1948–2000: A political study. New York: State University of New York Grussendorf, J., McAlister, A., Sandstrom, P., Udd, L., & Morrison, T. C. (2002). Resisting moral disengagement in support for war: Use of the “Peace Test” scale among student groups in 21 nations. Peace and Conflict: Journal of Peace Psychology, 8, 73– 83. Haslam, S. A., Turner, J. C., Oakes, P. J., McGarty, C., & Hayes, B. K. (1992). Context dependent variation in social stereotyping 1: The effects of intergroup relations as mediated by social change and frame of reference. European Journal of Social Psychology, 22, 3–20. Herbert, T. L., & Dunkel-Schetter, C. (1992). Negative social reactions to victims: An overview of responses and their determinants. In L. Montada, S. H. Filipp, & M. J. Lerner (Eds.), Life crises and experiences of loss in adulthood (pp. 497–518). Hillsdale, NJ: Erlbaum. Higgins, E. T. (1989). Self-discrepancy theory: What patterns of selfbeliefs cause people to suffer? In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 22, pp. 93–136). New York: Academic Press. Higgins, E. T. (1999). When do self-discrepancies have specific relations to emotions? The second-generation question of Tangney, Niedenthal, Covert, and Barlow (1998). Journal of Personality and Social Psychology, 77, 1313–1317. Hinkle, S., Taylor, L. A., Fox-Cardamone, D. L., & Crook, K. F. (1989). Intragroup identification and intergroup differentiation: A multicomponent approach. British Journal of Social Psychology, 28, 305–317. Iyer, A., Leach, C. W., & Crosby, F. J. (2003). White guilt and racial compensation: The benefits and limits of self-focus. Personality & Social Psychology Bulletin, 29, 117–129. Jackson, J. W. (2002). Intergroup attitudes as a function of different dimensions of group identification and perceived intergroup conflict. Self and Identity, 1, 11–33.

710

ROCCAS, KLAR, AND LIVIATAN

Janis, I. (1982). Groupthink (2nd ed.). Boston: Houghton Mifflin. Karasawa, M. (1991). Toward an assessment of social identity: The structure of group identification and its effect on in-group evaluations. British Journal of Social Psychology, 30, 293–307. Karasawa, M. (2002). Patriotism, nationalism, and internationalism among Japanese citizens: An etic-emic approach. Political Psychology, 23, 645– 666. Klar, Y., Roccas, S., & Liviatan, I. (2006). When my group is inflicting and is being inflicted pain: Symmetries and asymmetries in perceptions of group-based moral violations. Manuscript in preparation, Tel Aviv University, Tel Aviv, Israel. Kosterman, R., & Feshbach, S. (1989). Toward a measure of patriotic and nationalistic attitudes. Political Psychology, 10, 257–274. Landau, J. M. (1993). The Arab minority in Israel, 1967–1991. Oxford, England: Claredon Press. Li, Q., & Brewer, M. B. (2004). What does it mean to be an American? Patriotism, nationalism, and American identity after 9/11. Political Psychology, 25, 727–739. Licata, L., & Klein, O. (2004). Regards croise´s sur un passe´ commun: Anciens colonise´s et anciens coloniaux face a` l’action belge au Congo [Crossed outlooks on a common past: Former colonized and former colonizers facing the Belgian action in Congo]. In M. Sanchez-Mazas & L. Licata (Eds.), L’Autre: Regards psychosociaux (pp. 241–277). Grenoble, France: Presses Universitaires de Grenoble McAlister, M. (2001). Epic encounters: Culture, media and U.S. interests in the Middle East, 1945–2000. Berkeley, CA: University of California Press. Michman, D. (Ed.). (2002). Remembering the Holocaust in Germany, 1945–2000: German strategies and Jewish responses. New York: Peter Lang. Miron, A. M., Branscombe, N. R., & Schmitt, M. T. (2006). Collective guilt as distress over illegitmate intergroup inequality. Group Processes and Intergroup Relations, 9, 163–180. Mummendey, A., Klink, A., Mielke, R., Wenzel, M., & Blanz, M. (1999). Socio-structural characteristics of intergroup relations and identity management strategies: Results from a field study in East Germany. European Journal of Social Psychology, 29, 259 –285. Niven, B. (2002). Facing the Nazi past: United Germany and the legacy of the Third Reich. New York and London: Routledge. Opotow, S. (1990). Moral exclusion and injustice: An introduction. Journal of Social Issues, 46, 1–20. Petty, R. E., & Cacioppo, J. T. (1990). Involvement and persuasion: Tradition versus integration. Psychological Bulletin, 107, 367–374. Rensmann, L. (2004). Collective guilt, national identity and political processes in contemporary Germany. In N. R. Branscombe & B. Doosje (Eds.), Collective guilt: International perspectives (pp. 169 –192). New York: Cambridge University Press. Roccas, S., Sagiv, L., Schwartz, S. H., Halevy, N., & Eidelson, R. (2006). The nature of identification with groups: Theory and empirical investigations. Manuscript submitted for publication.

Rouhana, N. N. (1998). Israel and its Arab citizens: Predicaments in the relationship between ethnic states and ethnonational minorities. Third World Quarterly, 19, 277–296. Rouhana, N. N., & Bar-Tal, D. (1998). Psychological dynamics of intractable ethnonational conflicts: The Israeli–Palestinian case. American Psychologist, 53, 761–770. Rutland, A., & Brown, R. (2001). Stereotypes as justifications for prior intergroup discrimination: Studies of Scottish national stereotyping. European Journal of Social Psychology, 31, 127–141. Schatz, R. T., & Staub, E. (1997). Manifestations of blind and constructive patriotism: Personality correlates and individual– group relations. In D. Bar-Tal & E. Staub (Eds.), Patriotism in the lives of individuals and nations (pp. 229 –245). New York: Nelson-Hall. Schatz, R. T., Staub, E., & Lavine, H. (1999). On the varieties of national attachment: Blind versus constructive patriotism. Political Psychology, 20, 151–174. Schwartz, S. H. (1992). Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 25, pp. 1– 65). New York: Academic Press. Schwartz, S. H. (1994). Are there universal aspects in the content and structure of values? Journal of Social Issues, 50, 19 – 45. Staub, E. (1989). The roots of evil: The origins of genocide and other group violence. New York: Cambridge University Press. Staub, E. (1997). Blind versus constructive patriotism: Moving from embeddedness in the group to critical loyalty and action. In D. Bar-Tal & E. Staub (Eds.), Patriotism in the lives of individuals and nations (pp. 213–228). New York: Nelson-Hall. Stephan, W. G., & Renfro, C. L. (2002). The role of threat in intergroup relations. In D. M. Mackie & E. R. Smith (Eds.), From prejudice to intergroup emotions: Differentiated reactions to social groups (pp. 191–207). Philadelphia: Psychology Press. Tajfel, H. (Ed.). (1978). Differentiation between social groups: Studies in the social psychology of intergroup relations. London: Academic Press. Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin & S. Worchel, (Eds.), The social psychology of intergroup relations (pp. 33– 47). Monterey, CA: Brooks/Cole. Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup behavior. In S. Worschel & W. G. Austin (Eds.), The social psychology of intergroup relations (pp. 7–24). Chicago: Nelson Hall. Triandis, H. C. (1995). Individualism and collectivism. Boulder, CO: Westview Press. Triandis, H. C., & Gelfand, M. J. (1998). Converging measurement of horizontal and vertical individualism and collectivism. Journal of Personality and Social Psychology, 74, 118 –128. Turner, J. C. (1999). Some current issues in research on social identity and self-categorization theories. In N. Ellemers, R. Spears, & B. Doosje (Eds.), Social identity: Context, commitment, content (pp. 6 –34). Oxford, England: Blackwell.

IDENTIFICATION AND GROUP-BASED GUILT

711

Appendix Measure of Identification With Israel Mode of identification

Item

Attachment Glorification Attachment Glorification

I love Israel. Other nations can learn a lot from us. Being an Israeli is an important part of my identity. In today’s world, the only way to know what to do is to rely on the leaders of our nation. It is important to me to contribute to my nation. The IDF is the best army in the world. It is important to me to view myself as an Israeli. One of the important things that we have to teach children is to respect the leaders of our nation. I am strongly committed to my nation. Relative to other nations, we are a very moral nation. It is important to me that everyone will see me as an Israeli. It is disloyal for Israelis to criticize Israel. It is important for me to serve my country. Israel is better than other nations in all respects. When I talk about Israelis I usually say “we” rather than “they.” There is generally a good reason for every rule and regulation made by our national authorities.

Attachment Glorification Attachment Glorification Attachment Glorification Attachment Glorification Attachment Glorification Attachment Glorification

Note. Items were rated on a scale from 1 (strongly disagree) to 7 (strongly agree). IDF ⫽ Israeli Defense Forces.

Received August 4, 2004 Revision received March 15, 2006 Accepted April 2, 2006 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 712–729

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.712

Negotiation From a Near and Distant Time Perspective Marlone D. Henderson, Yaacov Trope, and Peter J. Carnevale New York University Across 3 experiments, the authors examined the effects of temporal distance on negotiation behavior. They found that greater temporal distance from negotiation decreased preference for piecemeal, singleissue consideration over integrative, multi-issue consideration (Experiment 1). They also found that greater temporal distance from an event being negotiated increased interest in conceding on the lowest priority issue and decreased interest in conceding on the highest priority issue (Experiment 2). Lastly, they found increased temporal distance from an event being negotiated produced a greater proportion of multi-issue offers, a greater likelihood of conceding on the lowest priority issue in exchange for a concession on the highest priority issue, and greater individual and joint outcomes (Experiment 3). Implications for conflict resolution and construal level theory are discussed. Keywords: construal, time, psychological distance, negotiation, integrative

1970; Kelley, 1966; Yukl, Malone, Hayslip, & Pamin, 1976). This classic work on issue consideration has shown that negotiators who are able to deal with multiple issues in a more simultaneous fashion are more willing to concede on low-priority issues in exchange for concessions on high-priority issues (logrolling) and, as a result, achieve higher joint outcomes. Considering issues in a more localized, discrete manner interferes with negotiators’ ability to distinguish between low-priority and high-priority issues when making concessions. As a result, negotiators who deal with issues separately are less likely to make appropriate trade-offs on lowpriority issues in exchange for concessions from the other party on high-priority issues (Froman, & Cohen, 1970; Pruitt, 1981). More recent research has continued to demonstrate the negative consequences of piecemeal issue consideration during negotiation, with special emphasis often placed on the interplay between issue consideration and other important negotiation variables (Mannix, Thompson, & Bazerman, 1989; Weingart, Bennett, & Brett, 1993). Notably, Weingart et al. (1993) examined the benefits of simultaneously rather than sequentially considering issues when individuals had different types of motivational orientation. A simultaneous approach to the issues led to better joint outcomes when individuals had an individualistic orientation; when negotiators had a cooperative orientation, negotiators were still able to achieve high joint outcomes even when issues were considered one at a time. This suggests that factors that promote a more integrative approach toward negotiation will mainly produce benefits for those with an individualistic orientation, as negotiators with a cooperative orientation seem ready and able to obtain mutually beneficial agreements.

A large amount of negotiation research has been dedicated to identifying the factors that either facilitate or hinder the resolution of interpersonal conflict (see Bazerman, Curhan, Moore, & Valley, 2000; Carnevale & Pruitt, 1992; De Dreu & Carnevale, 2003, for reviews). A general conclusion of much of this research is that integrative behaviors are a key means by which optimal negotiation outcomes are achieved— behaviors that include, for example, making trade-offs across low- and high-priority issues (Kelley & Schenitzki, 1972). Integrative outcomes are said to be optimal for the resolution of interpersonal conflict because they maximize utility, foster positive relationships between negotiators, and increase the chances that the parties will follow through on any agreement reached (Pruitt, 1981). Negotiation researchers have therefore recognized the importance of identifying the factors that facilitate integrative behavior. The current study examines the proposition that temporal distance from the realization of negotiated agreements promotes integrative behavior during the negotiation process.

Integrative Behavior Several lines of research have demonstrated that making offers in a piecemeal (one issue at a time) rather than a multi-issue fashion interferes with integrative agreements (Froman & Cohen,

Marlone D. Henderson, Yaacov Trope, and Peter J. Carnevale, Department of Psychology, New York University. This research was supported by a National Science Foundation Graduate Student Fellowship to Marlone D. Henderson, National Institute of Mental Health Grant 59030-06A1 to Yaacov Trope, and National Science Foundation Grant SES-0453301 to Peter J. Carnevale. We wish to thank Beryl J. Filton and Ryan D. Coganow for their assistance with data collection, Niall Bolger for his assistance with data analyses, and Susan M. Andersen and Robert B. Lount, Jr., for comments on an earlier draft of this article. Special thanks also go to Kentaro Fujita, Ido Liviatan, and Cheryl J. Wakslak for extensive discussion of ideas. Correspondence concerning this article should be addressed to Marlone D. Henderson, who is now at the Department of Psychology, University of Chicago, Chicago, IL 60637. E-mail: [email protected]

Temporal Distance in Negotiation Time has long been regarded as an important factor in negotiation (Carnevale, O’Connor, & McCusker, 1993; Druckman, 1994). However, there is a relatively small amount of research on the impact of temporal distance on negotiation. Most studies of time have examined the issue of time pressure, defined as the amount of time available to negotiate or the costs of continued negotiation (e.g., De Dreu, 2003; Lytle, Brett, & Shapiro, 1999; 712

TEMPORAL DISTANCE AND NEGOTIATION

Moore, 2004; Olekalns & Smith, 2003). A set of studies conducted by Okhuysen, Galinsky, and Uptigrove (2003) examined the effects of temporal distance from the realization of negotiated agreements on the quality of outcomes reached in negotiation. Okhuysen et al. found that dyads that negotiated over an agreement that was set to take effect in the distant future (1 year) reached better joint outcomes than dyads that negotiated over an agreement that was set to take effect in the near future (2 weeks). These researchers argued that as temporal distance increases, negotiators’ incentive to achieve an agreement that maximizes their positive outcomes or minimizes their negative outcomes should decrease, making it easier for them to concede on issues and reach integrative agreements. Okhuysen et al.’s findings are important because they demonstrate that increased temporal distance from the realization of negotiated outcomes make it easier for individuals to reach better joint outcomes. These findings also raise new questions regarding the specific aspects of the negotiation process that are affected by temporal distance and the role these aspects play as mediators of the effect of temporal distance on joint outcomes. Are dyads more likely or less likely to rely on single-issue or multiissue offers when negotiating distant future outcomes than near future outcomes? How does temporal distance affect the kind of concessions negotiators are willing to make? Is the willingness to make concessions on low-priority versus high-priority issues similarly or differently affected by temporal distance? The present research addresses these questions within the framework of construal level theory (Trope & Liberman, 2003).

Temporal Construal According to construal level theory (CLT; N. Liberman et al., in press; Trope & Liberman, 2003), people construe objects and events differently depending on their distance from them. From a distant perspective, people form high-level, more abstract construals of objects and events. Construing an object or event at a higher level involves extracting the perceived essence, gist, or summary of the given information about the object or event (Medin & Ortony, 1989; Schul, 1983). Because higher level processing entails extracting the core aspects of an object or event, it involves emphasizing primary, critical features over secondary, incidental features. To distinguish between the primary and secondary features of an object or event, higher level processing first involves a global consideration of the available information about the object or event (“seeing the forest rather than the trees”) and then a focus on the relations among the information (Hayes-Roth, 1977; Reder & Anderson, 1980; Schul, 1983). Therefore, high-level construals may be regarded as a by-product of integrating separate features of an object or event within a structured representation that emphasizes more primary than secondary information. From a proximal perspective, in contrast, people form more concrete, low-level construals of objects and events. These construals are less structured and fail to integrate separate aspects of objects and events. Therefore, low-level construals allow for a greater relative emphasis on secondary information about objects and events than highlevel construals. For example, an aspect such as “air conditioning in the employee lounge” might identify a secondary, low-level feature of an object (“work contract”). Conceptualizing the same object at a higher level construal (“job satisfaction”) renders such a feature less relevant and makes other aspects more prominent, such as “the amount of paid sick days.”

713

Consistent with this analysis, research has shown that greater temporal distance from objects and events promotes a more global consideration of information and more emphasis on essential rather than peripheral features of objects and events (Fo¨rster, Friedman, & Liberman, 2004; also see Smith & Trope, 2006). For example, Fo¨rster et al. (Studies 2 and 3) found that participants who envisioned their lives and imagined themselves engaging in a visual task in the distant future (a year later) as opposed to the near future (next day) subsequently exhibited increased perceptual integration and recognition of images out of a fragmented visual stimulus. In a similar vein, N. Liberman, Sagristano, and Trope (2002) found that a temporally distant perspective from an event fostered more inclusive processing of information related to the event. For example, N. Liberman et al. (Study 1) found that participants who imagined engaging in several activities (e.g., having a yard sale, going on a camping trip) in the distant future (next year) as opposed to the near future (upcoming weekend) used broader categories to classify objects related to the activities. Presumably, individuals with a temporally distant perspective exhibited a greater breadth of categorization because they disregarded the incidental features of each object, which typically isolate them from one another, and focused instead on the fewer essential or defining features of the objects, which unite them together. Several lines of research have also shown that increased temporal distance from events and objects activates higher levels of construal when making evaluations and judgments about those events and objects. For example, N. Liberman and Trope (1998, Study 1) found that when individuals thought about engaging in activities (e.g., making a list, washing clothes) in the distant future (sometime next year) as opposed to the near future (tomorrow), they preferred to identify activities in terms of the superordinate end-states that could be achieved from the activities (getting organized, removing odors from clothes) rather than the subordinate means that could be used to carry out the activities (writing things down, putting clothes into the machine). In another set of studies, Trope and Liberman (2000) had individuals decide for the near or distant future between tasks that differed in the valence associated with the primary features and secondary features of the tasks. Results showed that individuals that decided for the distant future gave more weight to the primary features over secondary features when reporting their evaluations and preferences between the tasks (Trope & Liberman, 2000).

Temporal Construal and Negotiation Overall, research in support of CLT suggests that decision makers with a temporally distant perspective are more likely to focus on the big picture as opposed to the incidental details when making a decision, resulting in more weight being given to primary and essential information when making a decision. It is important to note, however, that these consequences of a temporally distant perspective have only been studied within the context of individual decision making. Nevertheless, CLT is meant to be a general theory that applies to all types of real-life decisions in which the available options entail a trade-off between one’s primary and secondary interests, including decisions by dyads and groups (Trope & Liberman, 2000). In the current research, we relied on CLT to offer some insights into how temporal distance might affect negotiation behavior. The

714

HENDERSON, TROPE, AND CARNEVALE

present research is not a test of construal level theory per se, but rather we focus on whether the theory can provide some insights into negotiation behavior. Specifically, we investigated how temporal distance from the realization of a negotiated agreement promotes a more structured approach toward the issues during negotiation. It is important to note that temporal distance from the realization of a negotiated agreement can be activated in several ways. In particular, temporal distance is often manifest in the nearness of a negotiation session in time (e.g., a meeting today versus next month) or the closeness in time between a negotiation session and the point in time when the negotiated agreement is to be implemented. To illustrate this distinction, we offer the now classic negotiation example of a couple that is planning a vacation (Pruitt, 1981). A couple might have opposing preferences for the vacation and they might set a date in the near or distant future to sit down and try to resolve their differences for the vacation. Moreover, a couple might try settling their differences for a vacation that is set to occur in the near or distant future. In both cases, when temporal distance is increased, the realization of whatever agreement that is reached is also pushed farther into the future. As a result, a temporally distant perspective, regardless of how it is induced, should promote a more structured consideration of the issues. As negotiators’ construal of information about the agreement becomes more structured, they should be better able to integrate separate aspects or features of the agreement and emphasize more primary rather than secondary issues within the agreement. Therefore, we hypothesized that a temporally distant perspective from the realization of a negotiated agreement would promote less single-issue offers (separate consideration of issues) and more appropriate, systematic concessions that indicate concessions on low-priority issues in exchange for concessions on high-priority issues (cf. Kelley, 1966; Kelley & Schenitzki, 1972). Three experiments were designed to test these hypotheses. In Experiment 1, we tested whether greater temporal distance from a negotiation decreases preference for single-issue consideration relative to preference for multi-issue consideration during the negotiation. In Experiment 2, we tested whether increased temporal distance from an event being hypothetically negotiated increases willingness to make a concession on the lowest priority issue and decreases willingness to make a concession on the highest priority issue. In Experiment 3, we tested whether increased temporal distance from an event being negotiated through non-face-to-face discussion produces less single-issue offers (relative to multi-issue offers) and more complete concessions on the lowest priority issue in exchange for complete concessions from the other party on the highest priority issue.

Experiment 1: Does Temporal Distance Affect Preference for Issue Consideration? Negotiators with a temporally near rather than distant perspective are likely to process information related to an agreement in a more localized, fragmented, and unstructured manner. Consequently, negotiators with a near rather than distant perspective are likely to find it more difficult to craft multi-issue offers instead of single-issue offers. In light of this, we hypothesized that as negotiators’ temporal perspective was decreased, they would find an approach that did not revolve around integrating issues (piecemeal issue consideration) more appealing than an approach that did

revolve around integrating issues (simultaneous issue consideration). The current experiment tested this hypothesis.

Method Participants Participants were 43 students (12 men, 31 women) enrolled in a psychology course at New York University, who participated in partial fulfillment of a course requirement. We included the gender of the participant as a factor for all of the analyses and controlled for gender in all analyses, and the pattern of results was unchanged in both cases. Thus, the gender of participants is not discussed further.

Design Type of issue consideration (piecemeal vs. simultaneous) was the within-participant variable. Temporal perspective (near vs. distant) and order of the presentation for the dependent measure (piecemeal first vs. simultaneous first) were the between-participants variables. No order differences emerged for any of the analyses reported. There were 23 participants in the temporally near perspective cell and 20 participants in the temporally distant perspective cell.

Negotiation Task For this task, the role-playing exercise Towers Market was used. This task was developed for multi-issue group decision making. Previous researchers who have used this task (e.g., Weingart et al., 1993) have asked participants to assume the roles of representatives of four stores that are interested in opening a joint market in which each store is separate, but common areas are shared. The stores include a grocery, liquor store, florist, and bakery. Each merchant’s decision to join the market is contingent on how the market would be managed. There are five issues that remain to be agreed on at a meeting before any decision is made to form the joint business. Typically, each participant is assigned to a different role with pre-specified interests and priorities designated by the assignment of numerical values to each alternative for each issue. Participants are then required to negotiate over the issues described in the exercise. To control for exposure to different information in the exercise, we held constant the role assignment by having each participant adopt the same role (florist); participants were led to believe that only they had been assigned to their particular role.

Procedure Participants were invited into the lab to participate in a study on “interpersonal decision-making.” We ran participants in groups of 3 to 4 people.1 Participants assigned to the temporally near perspective condition were led to believe that they would engage in the live negotiation in the current session. Participants assigned to the temporally distant perspective condition were led to believe that they would engage in the live negotiation 1 month later. Participants were told that such role-playing exercises were useful for studying how individuals make decisions together. Participants were told that before they could participate in the live negotiation exercise, they had to make a decision about what the format would be during the

1

The instructions informed participants that the negotiation exercise was designed for 4 people. Occasionally, only 3 participants arrived for a session because of the failure of participants to show up for their scheduled experimental appointment. Anticipating this possibility, the instructions informed participants that any participants that were not present at the beginning of the session would arrive later in the session. Follow-up checks revealed that this was indeed effective at bolstering the cover story.

TEMPORAL DISTANCE AND NEGOTIATION

715

9

negotiation. Specifically, participants were told that they would have to indicate their preference between requiring all negotiating parties to consider and make offers in a piecemeal fashion or requiring all negotiating parties to consider and make offers in a simultaneous fashion.

8 7

After reviewing the information about the role-playing exercise, participants were asked to make a choice between two format options for the negotiation exercise—piecemeal issue consideration or simultaneous issue consideration. Specifically, we borrowed the descriptions that were used by Weingart et al. (1993) to manipulate the type of issue consideration during the negotiation. However, unlike in the Weingart et al. experiments, in which the type of issue consideration was treated as an independent variable, we presented all participants in this experiment with descriptions of both types of issue consideration. The piecemeal issue format was labeled “vote on individual issues” and required all negotiators to discuss only one issue at a time. If negotiators adopted this format option, negotiators would not be allowed to discuss a new issue until the prior issue was decided by a formal vote. The simultaneous issue format was labeled “vote on a proposal incorporating all the issues” and required negotiators to discuss all of the issues together. If negotiators adopted this format option, negotiators would have to vote for or against proposals that included all of the issues (see Weingart et al., 1993, p. 508, for a complete description of each format option). After reading the description of each format option, participants were asked to indicate their preference for using that particular format during the negotiation by means of the following 9-point rating scale: “Please indicate how much you would prefer this negotiation format (when you come back in a month)?” The answer scale ranged from 1 (definitely do not prefer) to 9 (definitely do prefer). In addition, we gave participants a forced choice by having them indicate which type of issue consideration they would prefer during the negotiation task. Responses to the preference items and responses to the forced choice item served as our dependent measures. Finally, we verified that our experimental groups did not differ in mood by means of the following 9-point rating scale: “How do you feel right now, at this very moment?” The answer scaled ranged from 1 (very bad) to 9 (very good).

Results Preference for Issue Consideration Participants’ preference ratings toward piecemeal and simultaneous issue consideration were analyzed using a 2 (temporal perspective: near vs. distant) ⫻ 2 (type of issue consideration: piecemeal vs. simultaneous) repeated measures analysis of variance (ANOVA), with the first factor as a between-participants variable and the last factor as a within-participant variable. Neither the main effect of temporal perspective (F ⬍ 1) nor the main effect of type of issue consideration, F(1, 41) ⫽ 1.40, p ⫽ .24, were significant. Analyses did reveal a significant Temporal Perspective ⫻ Type of Issue Consideration interaction effect, F(1, 41) ⫽ 7.59, p ⬍ .01 (see Figure 1). Specific comparisons revealed, as expected, that participants in the temporally near perspective condition evidenced a greater preference for engaging in piecemeal issue consideration than participants in the temporally distant perspective condition (M ⫽ 6.26, SD ⫽ 2.00 vs. M ⫽ 4.95, SD ⫽ 2.06), t(41) ⫽ 2.11, p ⬍ .05, d ⫽ 0.64. Specific comparisons also revealed, as expected, that participants in the temporally distant perspective condition evidenced a greater preference in engaging in simultaneous issue consideration than participants in the temporally near perspective condition (M ⫽ 5.80, SD ⫽ 2.21 vs. M ⫽ 4.13, SD ⫽ 1.91), t(41) ⫽ 2.65, p ⬍ .05, d ⫽ 0.83.

Preference Rating

Measures

6 near distant

5 4 3 2 1 piecemeal

simultaneous

Figure 1. Mean preference ratings as a function of temporal perspective (near vs. distant) and type of issue consideration (piecemeal vs. simultaneous; Experiment 1).

We also analyzed the within time comparisons for the preference for issue consideration. Specific comparisons revealed, as expected, that participants in the temporally near perspective condition evidenced a greater preference in engaging in piecemeal issue consideration than simultaneous issue consideration, t(41) ⫽ 2.11, p ⬍ .05, d ⫽ 0.64. In contrast, participants in the temporally distant perspective condition preferred to engage in simultaneous issue consideration to the same extent as piecemeal issue consideration (t ⬍ 1).

Choice of Issue Consideration We analyzed the proportion of participants who chose to engage in piecemeal versus simultaneous issue consideration. We found that 12 of the 20 participants (60%) in the temporally distant perspective group chose to engage in simultaneous issue consideration, whereas only 5 of 23 (22%) in the temporally near perspective group chose to engage in simultaneous issue consideration. This pattern of results produced a significant difference, ␹2(1, N ⫽ 43) ⫽ 6.55, p ⫽ .01, between the groups.

Mood Participants’ mood rating was entered into a one-way ANOVA, with temporal perspective condition (near vs. distant) as the between-participants variable. No significant difference emerged between the temporally near perspective group (M ⫽ 5.30, SD ⫽ 1.55) and temporally distant perspective group (M ⫽ 5.10, SD ⫽ 1.65; F ⬍ 1). Moreover, adjusting for this variable as a covariate did not change the pattern of the results reported in the previous paragraphs, suggesting that mood does not mediate the effect of temporal distance on preference for issue consideration.

Discussion According to CLT, when individuals experience greater temporal distance from a negotiated agreement, two consequences

716

HENDERSON, TROPE, AND CARNEVALE

should emerge. First, information related to any agreement that might be reached in the distant future should be considered in a more global manner. Second, information related to the primary aspects or features of any agreement should receive more weight than information related to the secondary aspects or features of the agreement. Accordingly, strategies that center on considering issues in a more packaged format should seem less attractive to negotiators with a temporally near rather than distant perspective, because individuals with a near perspective are likely to have a more fragmented and unstructured representation of the issues compared with those with a distant perspective. Given that negotiators with a temporally near perspective find it more difficult to consider issues simultaneously rather than separately, we expected they would find the prospect of being required to construct multiissue offers throughout a negotiation less appealing than negotiators with a temporally distant perspective. Results confirmed our expectation, with participants in the temporally near perspective group exhibiting a greater preference for considering issues in piecemeal rather than simultaneous fashion. In this experiment, we examined the effects of temporal distance from the realization of a negotiated agreement on interest in single-issue and multi-issue consideration by varying when negotiators expected to meet to try to reach an agreement. As noted earlier, temporal distance from the realization of a negotiated agreement can also vary as a function of when negotiators expect to implement or carry out any agreement that is reached. Generally, we assume that when individuals negotiate over issues pertaining to an event that is not expected to occur until the distant future (e.g., opening of a new market, start of a new work contract), any agreement that is reached on the issues will not be implemented or experienced until the event itself occurs (e.g., until the market opens, until the work contract begins). Just as greater temporal distance from when a negotiation is set to occur should activate a higher level construal of the issues (e.g., more global consideration of issues, more weight on primary issues), so should greater temporal distance from when the event surrounding the negotiation is set to occur. In the next experiment, we examine the effects of temporal distance on concession behavior in a negotiation setting by exploring the consequences of greater temporal distance from when the event being negotiated over is set to occur. Specifically, in the next experiment we examined the weight that individuals with a temporally distant versus near perspective place on high-priority versus low-priority issues when making concessions during negotiation.

Experiment 2: Does Temporal Distance Affect Preference for Appropriate Concessions? Whereas Experiment 1 demonstrated the increased inclination to approach a negotiation in a more integrative fashion as temporal perspective increased from a negotiated agreement, it did not address whether this approach emphasized more primary than secondary concerns. That is, we showed that individuals who had a temporally near rather than distant perspective were less likely to consider the issues in relation to each other (i.e., they preferred to consider issues separately). However, we did not examine whether such individuals would be less focused on their higher priority concerns. Because negotiators with a temporally distant perspective from the negotiated agreement are expected to construe the

issues at a higher level, they should put more weight on their primary concerns than negotiators with a temporally near perspective. As a result, negotiators with a temporally distant rather than near perspective should exhibit less interest in reaching an agreement that requires a concession on a high-priority issue. Because negotiators with a temporally near perspective from the negotiated agreement are expected to construe the issues at a lower level, they should put more weight on their secondary concerns than negotiators with a temporally distant perspective. As a result, individuals with a temporally near rather than distant perspective from a negotiated agreement should exhibit less interest in reaching an agreement that requires a concession on a low-priority issue. The current experiment tested these hypotheses.

Method Participants Participants were 62 students (13 men, 49 women) at New York University, who participated for $5 or in partial fulfillment of a course requirement. We included the gender of the participant as a factor for all of the analyses and controlled for gender in all analyses, and the pattern of results was unchanged in both cases. Thus, the gender of participants is not discussed further.

Design Temporal perspective (near vs. distant) and type of concession (lowpriority issue vs. high-priority issue) were the between-participants variables. There were 15 or 16 participants in each of the cells. We borrowed a manipulation of motivational orientation used in previous negotiation research (O’Connor & Carnevale, 1997; Pruitt & Lewis, 1975; Weingart et al., 1993)2 and instructed all participants to adopt an individualistic orientation.3

Negotiation Task As in Experiment 1, the task for this experiment was the Towers Market role-playing exercise. Again, participants were told to take the perspective of one of the characters in the exercise and the role assignment was held constant in this experiment by assigning each participant to the same role (florist). The usage of the Towers Market exercise was ideal for our interest in this experiment because it has certain issues built into the description of the exercise that are of primary concern for each role in the exercise. All

2 We held the motivational orientation constant in this experiment and in Experiment 3 because prior research (Weingart et al., 1993) has demonstrated that having negotiators adopt an integrative approach during negotiation mainly serves to alleviate the negative consequences of having an individualistic orientation. Given that this prior work suggests that factors that promote a more integrative approach during negotiation are likely to benefit mainly those with an individualistic orientation and that many people approach a negotiation with such an orientation (Bazerman & Neale, 1992; Pruitt & Carnevale, 1993), we decided to limit our examination of the effects of temporal perspective on concession behavior and outcomes to negotiators with an individualistic orientation. 3 After reading over the description of the negotiation exercise, participants read the following: “Some advice for you: Any agreement that is reached about the Market before it is announced in 2 days (2 years) will have a major impact on your store’s profitability. The points on the previous page are indicators of profitability for just you. When making your decisions, don’t worry about how well the other parties are doing. Your primary objective is to maximize your own outcome.”

TEMPORAL DISTANCE AND NEGOTIATION participants in the current experiment read statements throughout the description of the Towers Market exercise that conveyed that the highest priority issue for their role (florist) was how the clerks for the market would be trained. These statements included, “Your business is built on the expertise of your personnel,” and “It is essential to hire and train your own clerks.” As such, the issue of clerks had the highest potential point value assigned to it to convey its primary status (100). Participants did not read any statements that conveyed any special priority status to the issue of how the advertisement would be done for the market, and thus, the issue of advertisement had the lowest potential point value assigned to it to convey its secondary status (40). The issue of temperature, maintenance, and position fell in between the issue of clerks and advertisement in terms of their priority status, and thus, the appropriate potential point values (60, 60, 80) were assigned to convey their intermediate status. The maximum outcome score possible was 340 (see Appendix A).

717

Maintenance

Option b

Position

Option a

For participants in the high-priority issue concession condition, the fifth remaining issue (clerks) contained one of the participants’ nonpreferred options, amounting to a loss of 80 points from the total outcome score (see Appendix A). Participants assigned to this high-priority issue concession condition were asked the following: Imagine you received the following proposal for the Market that will open in 2 days (2 years): Temperature

Option a

Advertising

Option b

Procedure

Clerks

Option c

As in Experiment 1, participants were invited to participate in a study on “interpersonal decision-making.” Participants were told that they would read a hypothetical negotiation exercise. The instructions explained that the experimenter was interested in individuals’ reactions to potential outcomes from a negotiation. Unlike Experiment 1, participants were not led to believe that they would engage in a live negotiation. Instead, participants read the exercise and were told that they would be presented with a possible proposal that they might have received had they actually engaged in a live negotiation. Participants were randomly assigned to condition. As in Experiment 1, we varied the temporal perspective in the negotiation task. However, unlike in Experiment 1, we used a manipulation of temporal perspective within the description of the negotiation exercise itself, by varying when the event (“the opening of the market”) that the negotiation revolved around was set to occur. All participants read about a group of four business owners in negotiation about opening a joint store together. As described in the procedure section of Experiment 1, participants learned that five issues needed to be negotiated before any decision could be made about whether to open the joint business. Participants assigned to the temporally near perspective condition read repeated statements throughout the description of the exercise that the opening of store would occur in 2 days. Participants assigned to the temporally distant perspective condition read repeated statements throughout the description of the exercise that the opening of the store would occur in 2 years. For example, some statements included in the description of the exercise were “The Market will have an open plan, with a common de´cor and will open in 2 days (2 years),” “You and the other merchants have agreed to meet to try to resolve these issues so that everyone can prepare to open the Market in 2 days (2 years),” and “P’s grocery, a successful East Side establishment, plans to open Towers Market in 2 days (2 years).” Afterwards, participants were told to imagine that they received an offer that contained an option for each of the five issues on the table. The proposal contained participants’ most preferred option for four out of the five issues. For participants assigned to the low-priority issue concession condition, the fifth remaining issue (advertising) contained one of the participants’ nonpreferred options, amounting to a loss of 40 points from the total outcome score (see Appendix A). Specifically, participants assigned to the low-priority issue concession condition were asked the following:

Maintenance

Option b

Position

Option a

Imagine you received the following proposal for the Market that will open in 2 days (2 years): Temperature

Option a

Advertising

Option a

Clerks

Option e

Measures We measured participants’ interest in accepting the presented proposal by means of the following 6-point rating scales: “How unsatisfied/satisfied would you be with this proposal?” The answer scale ranged from 1 (very unsatisfied) to 6 (very satisfied). “How unlikely/likely would you be to reject this proposal?” (reverse scored). The answer scale ranged from 1 (very unlikely) to 6 (very likely). “If you accepted this proposal, how much would you regret it later?” The answer scale ranged from 1 (completely regret it) to 6 (completely not regret it). “If you accepted this proposal, how much would you feel that the other parties took advantage of you?” (reverse scored). The answer scale ranged from 1 (not at all) to 6 (very much). We created an index of interest in accepting the offer by averaging each participant’s responses to the four items (␣ ⫽ .87). This served as our dependent measure.

Results Participants’ interest in accepting the offer was analyzed using a 2 (temporal perspective: near vs. distant) ⫻ 2 (type of concession: low-priority issue vs. high-priority issue) ANOVA. None of the main effects were significant (Fs ⬍ 1). Analyses did reveal, however, a significant Temporal Perspective ⫻ Type of Concession interaction effect, F(1, 58) ⫽ 11.39, p ⫽ .001. Specific comparisons revealed, as expected, that participants in the temporally distant perspective condition evidenced more interest in accepting the offer that required a concession on the lowest priority issue than participants in the temporally near perspective condition (M ⫽ 5.30, SD ⫽ 0.65 vs. M ⫽ 4.61, SD ⫽ 1.01), t(30) ⫽ 2.29, p ⫽ .03, d ⫽ 0.84. Specific comparisons also revealed, as expected, that participants in the temporally distant perspective condition evidenced less interest in accepting the offer that required a concession on the highest priority issue than participants in the temporally near perspective condition (M ⫽ 4.57, SD ⫽ 0.75 vs. M ⫽ 5.17, SD ⫽ 0.47), t(28) ⫽ ⫺2.62, p ⫽ .01, d ⫽ 0.99. We also analyzed the within time comparisons for the different types of concessions. Specific comparisons revealed, as expected, that participants in the temporally distant perspective condition indicated less interest in accepting the offer that required a concession on the highest priority issue rather than a concession on the lowest priority issue (M ⫽ 4.57 vs. M ⫽ 5.30), t(29) ⫽ ⫺2.90, p

HENDERSON, TROPE, AND CARNEVALE

718

⫽ .007, d ⫽ 1.08. In contrast, specific comparisons revealed that participants in the temporally near perspective condition indicated less interest in accepting the offer that required a concession on the lowest priority issue rather than a concession on the highest priority issue (M ⫽ 4.61 vs. M ⫽ 5.17), t(29) ⫽ ⫺1.94, p ⫽ .06, d ⫽ 0.72.

Discussion Overall, the findings from the current experiment are consistent with our hypotheses. On the basis of the tenets of CLT, we reasoned that individuals with a temporally distant perspective from a negotiated agreement would adopt a high-level construal of the issues, resulting in more weight being given to primary concerns. The pattern of results confirmed our expectations. Individuals who had a temporally distant perspective expressed less interest in accepting an agreement that involved a concession on a high-priority issue than individuals who had a temporally near perspective. We also reasoned that individuals with a temporally near perspective from a negotiated agreement would adopt a low-level construal of the issues, resulting in more weight being given to secondary concerns. Again, the pattern of results confirmed our expectations. Individuals who had a temporally near perspective expressed less interest in accepting the agreement that involved a concession on a secondary issue than individuals who had a temporally distant perspective. In fact, individuals with a temporally near perspective exhibited a greater interest in making the more costly concession (concession on the high-priority issue rather than concession on the low-priority issue). Such a preference reversal is consistent with the notion that negotiators with a temporally near perspective are likely to think about the issues surrounding their negotiation in such a fragmented manner that they find it difficult to maintain the relative priorities attached to the respective issues. In contrast, negotiators with a temporally distant perspective are likely to think about the issues surrounding the negotiation in such a structured manner that they clearly distinguish between their central, primary concerns and their incidental, secondary concerns (see Druckman & Rozelle, 1975, for a related discussion). The results from the current experiment provide preliminary evidence that the type of temporal perspective that negotiators adopt within a negotiation impacts the extent to which they are willing to make appropriate concessions. Our findings suggest that individuals who have a temporally distant perspective within a negotiation are more willing to accept concessions on low-priority issues as long as concessions on higher priority issues are not called for. The results from this experiment complement the findings from the first experiment, as several studies have found that considering issues in a localized, piecemeal fashion rather than in a more integrative, packaged fashion leads to inappropriate concessions and poorer joint outcomes (e.g., Erickson, Holmes, Frey, Walker, & Thibaut, 1974; Froman & Cohen, 1970; Kelley, 1966; Pruitt & Lewis, 1975; Weingart et al., 1993). Negotiators who resolve issues sequentially tend to make compromise concessions irrespective of whether an issue is of low or high value, reflecting a more fragmented approach toward the negotiation. In contrast, negotiators who resolve issues through global trade-offs make concessions on issues of low importance in return for similar concessions from the other party, reflecting a more structured understanding of

the negotiation (Pruitt, 1981). In a sense, considering issues one at a time makes people lose sight of the big picture, leading to poorer outcomes. Indeed, Weingart et al. (1993) found that participants who engaged in sequential consideration of issues argued more about specific positions on each issue while focusing less on the task of discovering joint outcomes. Taken together, the results from Experiments 1 and 2 suggest that a temporally distant perspective during a negotiation may be beneficial for reaching satisfactory agreements. When the issues being negotiated have integrative potential, our results suggest that a temporally distant perspective may facilitate more logrolling or concession making on low-priority issues in exchange for favorable outcomes on high-priority issues, resulting in more beneficial outcomes for both negotiators. Our third experiment uses live negotiation to investigate temporal perspective, the preference for piecemeal versus multi-issue consideration by negotiators, and the consequences for the types of concessions made and outcomes reached.

Experiment 3: Does the Impact of Temporal Distance on Issue Consideration and Concession Behavior Affect Outcomes? Whereas the previous set of experiments provide converging evidence that negotiators who have a temporally distant perspective within a negotiation are more likely to adopt an integrative, structured approach toward their negotiation (i.e., less single-issue consideration and less concern with secondary issues), we have yet to demonstrate that this approach manifests itself in live negotiation. At this point, we can only conclude from Experiment 1 that a temporally distant perspective within negotiation decreases individuals’ preference for considering issues one at a time when given the choice. And although this finding is certainly meaningful because it adds to our understanding of what factors influence negotiators’ attraction to certain strategies when several alternatives are available (Tinsley, 2001; Weingart, Hyder, & Prietula, 1996), it does not directly speak to whether a distant perspective spontaneously inhibits the consideration of issues in a fragmented, piecemeal manner during negotiation. Furthermore, although Experiment 2’s findings clearly suggest that individuals who have a temporally distant perspective reach more integrative agreements than individuals who have a temporally near perspective, Experiment 2’s findings derive from responses to an imaginary scenario, which limits inferences to actual negotiation behavior. Therefore, to address these limitations, Experiment 3 tested hypotheses in the context of live negotiation. On the basis of the tenets of CLT, we hypothesized that individuals who have a temporally distant rather than near perspective during live negotiation should exhibit a greater relative degree of multi-issue offers, more logrolling (concession on lowest priority issue while holding firm on highest priority issue), and better outcomes from the negotiation.

Method Participants Participants were 50 students enrolled in a psychology course at New York University, who participated in partial fulfillment of a course requirement as in Experiments 1 and 2. Twenty-five dyads took part in the experiment (13 female–female dyads, 4 male–male dyads, 8 female–male dyads). Data from 4 other dyads were collected but not included in the

TEMPORAL DISTANCE AND NEGOTIATION analyses because at least one of the individuals in the dyad had previously participated in a similar study. We included the gender composition of the dyad (same gender vs. mixed gender) as a factor for all of the analyses and controlled for gender composition in all analyses, and the pattern of results was unchanged in both cases. Thus, the gender composition of dyads is not discussed further.

Design Participants were randomly assigned to either the temporally near or distant perspective condition. There were 12 dyads in the temporally near perspective cell and 13 dyads in the temporally distant perspective cell. As in Experiment 2, we instructed all participants to adopt an individualistic orientation (see Footnote 2). As used in previous negotiation research (e.g., O’Connor & Carnevale, 1997; Thompson, 1991), all participants were told that 10 cash prizes worth $10 each would be awarded to individuals at the end of the semester. Specifically, participants were told that they had the opportunity to win a minimum of 1 and a maximum of all 10 cash prizes on the basis of their negotiation performance. The probability of winning these prizes was thus related to the number of points the negotiators earned.4

Negotiation Task For this experiment, we created a role-playing exercise that had the basic elements of a simulated bilateral negotiation as developed by Pruitt and Lewis (1975). This task asked pairs of participants to assume the roles of two students who are interested in doing an extra credit class presentation together. The two students are enrolled in a class in which the instructor has agreed to set aside time in the class schedule to allow pairs of students to give presentations to the class for extra credit. Each student’s decision to go through with the presentation is contingent on how several issues surrounding the presentation are resolved. There are four issues that remain to be agreed on before any decision is made about whether to do the presentation. To rule out the possibility that any effects from our previous experiments were due to the content associated with the specific issues described in the exercises, in the current experiment we left unspecified the content of the issues and the content of each option for each issue. Although we provided examples of the types of issues that might be negotiated for a class presentation (who will say what, what the color of the background will be, the order of the things presented, whether each of them should receive the same amount of extra credit, what to wear during the presentation), the instructions stressed that the issues did not refer to these examples. Participants were told that the content of the issues was not relevant to do the negotiation task. That is, they were told that they did not need to know what the issues were in order to do the negotiation. Participants were simply told to assume that the four issues represented things that would be relevant for doing an extra credit presentation. Participants were told they and the other person would be making offers and counteroffers back and forth between each other during the negotiation about which options should be adopted for the issues. Participants were told that they should try to come to an agreement about which options should be adopted for the extra credit presentation. That is, participants were told that they should try to come to a mutual agreement on the issues for the presentation. Specifically, participants were told to assume that if they did not reach an agreement on all of the issues, they would not be able to do the presentation. For this negotiation exercise, we constructed a point-scoring scheme to illustrate participants’ position on these issues. Participants were told that the use of points might seem a bit artificial but that it would allow them to compare the value of possible agreements with that of other alternative agreements. Specifically, participants were told that the points indicated how desirable each option was for them for each issue. The two point schedules that were presented to each dyadic member are presented in

719

Appendix B. Associated with each issue were five possible options, with an associated payoff. In this negotiation exercise, each participant was assigned to a different role with prespecified interests and different priorities designated by the assignment of numerical values to each option for each issue. On one of the negotiator’s schedules, Issue 2 had the highest potential for payoff (120) and Issue 4 had the lowest potential (40); these priorities were reversed for the other negotiator. Thus, the task had integrative (logrolling) potential for negotiators (some issues were of differing importance to negotiators), and therefore, high joint outcomes could be achieved if the negotiators completely exchanged concessions on their high- and low-priority issues. Issues 1 and 3 were distributive issues (of equal priority to negotiators). A compromise agreement (the midpoint on each issue) yielded a joint outcome score of 260 points (130 points for each negotiator). The maximum joint outcome score possible was 390 (190 points for one negotiator and 200 points for the other negotiator). Participants were not given any information about opponent payoffs. Note that we did not tell participants that they should not mention the payoff schedules during negotiation, nor did we tell participants that they should mention the payoff schedule during negotiation. Rather, we simply let participants decide for themselves how they should go about discussing the issues and reaching an agreement. As Yukl et al. (1976) pointed out, in such situations, neither person can know for sure whether the other side is telling the truth when information is communicated about payoffs or priorities, as it would be possible to give false information to mislead the opponent if a negotiator desired to do so (e.g., see O’Connor & Carnevale, 1997).5

Procedure As in our earlier experiments, participants were invited into the lab to participate in a study on non-face-to-face “interpersonal decision-making.” Each participant arrived separately and was assigned to a private room during the negotiation. Each participant was told that they would be engaged in a role-playing negotiation task over America Online (AOL) Instant Messenger with another participant. Participants received a screen name during the negotiation that was either “leftbooth” or “rightbooth.” We had participants engage in the negotiation over AOL for several reasons. First, we wanted to control for any variation in negotiation behavior due to visual cues of the other negotiator (e.g., gender, perceived physical attractiveness). Second, we wanted to ensure that all communication during the negotiation was coded properly, which is facilitated by having a computer transcript. Participants were randomly assigned to condition. Each participant was given a folder that contained a brief description of the extra credit presen-

4 Participants read the following: “It is EXTREMELY important to remember that your job is to get the most points out of this negotiation. Do not be at all concerned with the needs and welfare of the other person. The needs and welfare of the other negotiator are unimportant to you. In other words, your task is to maximize your own point winnings, disregarding how many points the other negotiator gets from any agreement that is reached for the presentation that is tomorrow (5 months from now). As an incentive for maximizing your points, individuals will be entered into a lottery for 10 cash prizes worth $10 each. The more points you receive, the more lottery tickets you will receive. Therefore, your probability of winning one of the cash prizes is related to the number of points you earn as a negotiator.” 5 Results revealed no differences in the likelihood of mentioning the payoff schedules between dyads in the temporally near (4 of 12) and distant (3 of 13) groups, ␹2(1, N ⫽ 25) ⫽ .33, p ⫽ .56. After adjusting for this as a covariate, all of the reported results remained significant. Moreover, we reran the analyses after excluding those dyads that did mention the payoff schedules, and the results remained unchanged. Overall, this suggests that it does not mediate the effect of temporal distance on the negotiation process and outcome.

720

HENDERSON, TROPE, AND CARNEVALE

tation and a payoff schedule. They were told the experiment was a laboratory simulation of a negotiation that might occur in the real world. They were told that the negotiation would involve several issues that were related to an extra credit class presentation, but the content of the issues was left unspecified, and their main task was to try to reach an agreement and to get as many points as possible from their agreement, in non-faceto-face discussion, on those issues. As in Experiment 2, we varied the temporal perspective in the negotiation task by varying when the event that the negotiation supposedly revolved around (“the extra credit class presentation”) was set to occur. Similar to the manipulation that was used in Experiment 2, participants assigned to the temporally near perspective condition read repeated statements throughout the description of the exercise that the extra credit class presentation would occur the next day. Participants assigned to the temporally distant perspective condition read repeated statements throughout the description of the exercise that the extra credit class presentation would occur in 5 months. For example, some statements included in the description of the exercise were, “During this exercise, you will learn about your role, and read about an extra credit class presentation that you and another student are considering doing tomorrow (5 months from now),” and “The instructor for your class has agreed to set aside time in the class schedule tomorrow (5 months from now) to allow pairs of students to give presentations to the class for extra credit.” All other information in the exercise was the same for both conditions. Following the instructions, the experimenter, who was blind to condition, reiterated to participants that the content of the issues was unimportant and that their main objective for the task was to maximize their outcome (number of points) from any agreement that was reached during the negotiation. Afterwards, participants immediately responded to a two-item questionnaire that asked them to verify that their concern during the negotiation was only with their own outcome and to indicate on a time line when the hypothetical extra credit presentation was set to take place. This was done to reinforce the individualistic motivational orientation and temporal perspective manipulation during the negotiation. Participants were told they would have 20 min to arrive at a mutually acceptable agreement. If dyads were able to reach an agreement, then their names were entered into a lottery at the end of the semester. If a dyad was unable to reach an agreement, then neither participant was entered into the lottery. Following the negotiation, participants answered questions about the negotiation. After the negotiation, participants were debriefed and thanked.

agreements were resolved by the expert coder, yielding one set of codes for each transcript. Subjective process measures. For the subjective measures of the negotiation process, participants responded to several items regarding their negotiation. First, we measured participants’ self-reported degree of singleissue consideration and multi-issue consideration by means of the following 7-point rating scales: “During your negotiation, how often did you make single-issue offers (i.e., offers that included an option for only one issue)?” “During your negotiation, how often did you make multi-issue offers (i.e., offers that included an option for more than one issue)?” Second, we measured participants’ self-reported degree of concession behavior on low- and high-priority issues by means of the following 7-point rating scales: “During your negotiation, how often did you make offers that involved you giving in or making a compromise on an issue that was worth a large amount of points to you?” “During your negotiation, how often did you make offers that involved you giving in or making a compromise on an issue that was worth a small amount of points to you?” The answer scales for all of the items ranged from 1 (not very often) to 7 (very often). Outcome measures. Each dyad had three outcome measures. First, we created an index of whether dyads reached a fully logrolling agreement by counting the number of dyads in which both negotiators completely conceded on their lowest priority issue in exchange for a complete concession on their highest priority issue. This served as an indicator of whether dyads took advantage of the integrative potential of the negotiation. Second, we examined the number of points that each participant earned in his or her agreement (individual gain). Finally, we examined the number of points that each dyad earned in their agreement (joint gain).

Results Length of Negotiation The length of negotiation for dyads with a temporally near perspective (M ⫽ 11.70 min, SD ⫽ 5.61) versus temporally distant perspective (M ⫽ 14.10 min, SD ⫽ 4.66) was not significantly different, t(23) ⫽ ⫺1.17, p ⫽ .26. All of the following analyses were rerun controlling for the length of negotiation, and the pattern of results were unchanged.

Objective Process Measures Measures The primary dependent measures were derived from the negotiation process and outcome. We obtained objective as well as subjective measures of the negotiation process and objective measures of the negotiation outcome. Objective process measures. The online interaction of the dyad was coded from a digital transcript of the negotiation. Our main interest was with the relative frequency of single-issue and multi-issue offers. We counted the number of offers that were made that involved a single issue (e.g., “How about we go with Option E for Issue 2?”) and the number of offers that were made that involved multiple issues (e.g., “How about we do E for Issue 4 and go with A for Issue 1?”). On the basis of this count, we created an index of multi-issue offers by taking the number of multiissue offers that occurred within a dyad and dividing them by the total number of offers that occurred within that dyad. We also counted the maximum number of issues that were involved in any offers that were made. This allowed us to index the extent to which dyads processed the issues in a global manner. Each transcript was coded twice for the number of single-issue and multi-issue offers made, and an index of multi-issue offers was created from each set of codes. Both coders were blind to the experimental conditions. Agreement between the two raters was high (r ⫽ .92). Dis-

First, the relative frequency of multi-issue offers made during negotiation was examined. Results showed, as expected, that dyads with a temporally distant perspective made a greater proportion of multi-issue offers (M ⫽ .56, SD ⫽ .25) than dyads with a temporally near perspective (M ⫽ .28, SD ⫽ .25), t(23) ⫽ 2.81, p ⫽ .01, d ⫽ 1.17. Second, we analyzed the number of single-issue and multi-issue offers that were made during negotiation using a 2 (temporal perspective: near vs. distant) ⫻ 2 (type of issue consideration: single-issue vs. multi-issue) repeated measures ANOVA, with the first factor as a between-participants variable and the last factor as a within-participant variable. The main effect of temporal perspective was not significant, F(1, 23) ⫽ 1.17, p ⫽ .29. However, the main effect of type of issue consideration was significant, F(1, 23) ⫽ 7.57, p ⫽ .01, with dyads making a greater number of single-issue offers (M ⫽ 6.84, SD ⫽ 4.50) than multi-issue offers (M ⫽ 4.24, SD ⫽ 3.67). Importantly, this main effect was qualified by a significant Temporal Perspective ⫻ Type of Issue Consideration interaction effect, F(1, 23) ⫽ 8.46, p ⫽ .008. Specific comparisons revealed that participants in the temporally near perspective condition made a greater number of singleissue offers (M ⫽ 9.00, SD ⫽ 3.91) than participants in the

TEMPORAL DISTANCE AND NEGOTIATION

temporally distant perspective condition (M ⫽ 4.85, SD ⫽ 4.18), t(23) ⫽ 2.56, p ⫽ .02, d ⫽ 1.07. Specific comparisons also revealed that participants in the temporally distant perspective condition made a greater number of multi-issue offers (M ⫽ 5.00, SD ⫽ 3.63) than participants in the temporally near perspective condition (M ⫽ 3.42, SD ⫽ 3.68), although this latter difference failed to reach statistical significance, t(23) ⫽ 1.08, p ⫽ .29, d ⫽ 0.45. We also analyzed the within time comparisons for the number of different types of offers made. Specific comparisons revealed that participants in the temporally near perspective condition made a greater number of single-issue offers than multi-issue offers, t(11) ⫽ 3.28, p ⫽ .007, d ⫽ 1.98; no significant difference emerged for participants in the temporally distant perspective condition (t ⬍ 1). The maximum number of issues that were involved in at least one offer during the negotiation was analyzed next (see Figure 2). On average, the maximum number of issues that were involved in an at least one offer was greater when dyads had a temporally distant perspective (M ⫽ 3.46, SD ⫽ 0.88) rather than a temporally near perspective (M ⫽ 2.17, SD ⫽ 1.03), t(23) ⫽ 3.39, p ⫽ .002, d ⫽ 1.41. When we analyzed the number of dyads that dealt with a maximum of one issue, two issues, three issues, or four issues at least once during negotiation, the following pattern emerged: Out of the 13 dyads that had a temporally distant perspective, 3 made an offer that dealt with two issues, 1 made an offer that dealt with three issues, and 9 made an offer that dealt with all four issues. Out of the 12 dyads that had a temporally near perspective, 3 never made an offer that dealt with more than one issue, 6 made an offer that dealt with two issues, 1 made an offer that dealt with three issues, and 2 made an offer that dealt with all four issues. Overall, this difference between conditions was significant, ␹2(3, N ⫽ 25) ⫽ 8.43, p ⫽ .04.

Subjective Process Measures Next, we examined the subjective measures of the negotiation process. Participants’ self-reported degree of single-issue and multi-issue consideration was analyzed using a 2 (temporal perspective: near vs. distant) ⫻ 2 (type of issue consideration:

P ercen tag e o f D yad s

80 70 60 50

near

40

distant

30 20 10 0 1 issue

2 issues

3 issues

4 issues

Maximum Number of Issues Dealt With Figure 2. Percentage of dyads that dealt with a maximum of one, two, three, or four issues together at least once while making an offer during negotiation as a function of temporal perspective (near vs. distant; Experiment 3).

721

single-issue vs. multi-issue) repeated measures ANOVA, with the first factor as a between-participants variable and the last factor as a within-participants variable. The main effect of temporal perspective was not significant, F(1, 23) ⫽ 2.28, p ⫽ .15. However, the main effect of type of issue consideration, F(1, 23) ⫽ 8.71, p ⫽ .007, was significant, with dyads reporting a greater degree of multi-issue consideration (M ⫽ 4.86, SD ⫽ 1.34) than single-issue consideration (M ⫽ 3.50, SD ⫽ 1.51). Importantly, this main effect was qualified by a significant Temporal Perspective ⫻ Type of Issue Consideration interaction effect, F(1, 23) ⫽ 7.13, p ⫽ .01. Specific comparisons revealed that participants in the temporally near perspective condition reported a greater degree of singleissue consideration (M ⫽ 4.33, SD ⫽ 1.56) than participants in the temporally distant perspective condition (M ⫽ 2.73, SD ⫽ 1.01), t(23) ⫽ 3.08, p ⫽ .005, d ⫽ 1.28. Specific comparisons also revealed that participants in the temporally distant perspective condition reported a greater degree of multi-issue consideration (M ⫽ 5.23, SD ⫽ 1.27) than participants in the temporally near perspective condition (M ⫽ 4.46, SD ⫽ 1.36), although this latter difference failed to reach significance, t(23) ⫽ 1.47, p ⫽ .16, d ⫽ 0.61. We also analyzed the within time comparisons for the selfreported degree of single-issue and multi-issue consideration. Although specific comparisons revealed that participants in the temporally distant perspective condition reported a greater degree of multi-issue consideration than single-issue consideration, t(12) ⫽ 5.40, p ⬍ .001, d ⫽ 3.12, no significant difference emerged for participants in the temporally near perspective condition (t ⬍ 1). Participants’ self-reported degree of concession behavior on low- and high-priority issues was analyzed using a 2 (temporal perspective: near vs. distant) ⫻ 2 (type of concession: lowpriority vs. high-priority) repeated measures ANOVA, with the first factor as a between-participants variable and the last factor as a within-participant variable. Neither the main effect of temporal perspective, F(1, 23) ⫽ 1.40, p ⫽ .25, nor the main effect of type of concession (F ⬍ 1) were significant. However, as expected, there was a significant Temporal Perspective ⫻ Type of Concession Interaction effect, F(1, 23) ⫽ 4.32, p ⫽ .05. Although participants in the temporally distant perspective condition reported more offers that involved a concession on an issue that was worth a small amount of points (M ⫽ 4.85, SD ⫽ 0.63) than participants in the temporally near perspective condition (M ⫽ 4.38, SD ⫽ 1.46), the difference was not significant, t(23) ⫽ 1.06, p ⫽ .30, d ⫽ 0.44. However, as expected, participants in the temporally distant perspective condition reported a fewer offers that involved a concession on an issue that was worth a large number of points (M ⫽ 3.96, SD ⫽ 1.13) than participants in the temporally near perspective condition (M ⫽ 5.08, SD ⫽ 1.36), t(23) ⫽ ⫺2.25, p ⫽ .03, d ⫽ 0.94. We also analyzed the within time comparisons for the self-reported degree of concession behavior. Specific comparisons revealed that participants in the temporally distant perspective condition reported fewer offers that involved a concession on an issue that was worth a large amount of points than a concession on an issue that was worth a small amount of points, t(12) ⫽ ⫺2.07, p ⫽ .06, d ⫽ 1.96; no significant difference emerged for participants in the temporally near perspective condition, t(11) ⫽ 1.09, p ⫽ .30, d ⫽ 0.66.

HENDERSON, TROPE, AND CARNEVALE

722 Outcome Measures

Twenty-three out of 25 of the dyads reached an agreement. The 2 dyads in the temporally distant perspective condition that did not reach an agreement indicated it was because of time constraints rather than because of impasse.6 For the 23 dyads that did reach an agreement, we first analyzed the number of cases in which the dyad reached a fully logrolling agreement (both negotiators completely conceded on their lowest priority issue in exchange for a complete concession on their highest priority issue). Although 10 out of the 11 dyads (90.9%) with a temporally distant perspective reached a fully logrolling agreement, only 6 of the 12 dyads (50%) with a temporally near perspective reached a fully logrolling agreement. This pattern of results produced a significant difference, ␹2(1, N ⫽ 23) ⫽ 4.54, p ⫽ .03, between the groups. Second, we analyzed the individual outcome of negotiators. Results revealed that negotiators with a temporally distant perspective earned a greater number of points (M ⫽ 182.50, SD ⫽ 19.93) than negotiators with a temporally near perspective (M ⫽ 162.29, SD ⫽ 34.23), t(44) ⫽ 2.42, p ⫽ .02, d ⫽ 0.73. Lastly, we analyzed the joint outcome of negotiators. Results revealed that dyads with a temporally distant perspective earned a greater number of points (M ⫽ 365.00, SD ⫽ 33.47) than dyads with a temporally near perspective (M ⫽ 324.58, SD ⫽ 56.91), t(21) ⫽ 2.05, p ⫽ .05, d ⫽ 0.89.7

Tests of Mediation To test whether the proportion of multi-issue offers mediated the effects of temporal perspective on joint outcome, we performed a multiple regression analysis following Kenny, Kashy, and Bolger (1998). First, temporal perspective significantly predicted joint outcome, ␤ ⫽ .41, t(21) ⫽ 2.05, p ⫽ .05, and the proportion of multi-issue offers, ␤ ⫽ .50, t(21) ⫽ 2.67, p ⫽ .01. Second, the proportion of multi-issue offers significantly predicted joint outcome, ␤ ⫽ .74, t(20) ⫽ 4.43, p ⬍ .001, with temporal perspective held constant. Finally, temporal perspective no longer significantly predicted joint outcome (␤ ⫽ .03, t ⬍ 1). These results indicate that the proportion of multi-issue offers mediated the effects of temporal perspective on joint outcome. A Sobel test of mediation (Sobel, 1982) confirmed that the proportion of multi-issue offers significantly mediated the relationship between experimental condition and subsequent joint outcome (Sobel test ⫽ 2.37, p ⬍ .02). The full mediational model is displayed in Figure 3.

Discussion The results from the current experiment conceptually replicate and extend the results from our first two experiments, in the context of live negotiation. Results based on the actual dialogue between negotiators as well as negotiators’ self-report of the negotiation process indicated that those who had a temporally distant perspective from the event being negotiated exhibited a greater relative degree of multi-issue offers than dyads that had a temporally near perspective. That is, the results showed that the overwhelming preference for single-issue consideration over multi-issue consideration often found in negotiation was virtually eliminated when negotiators had a temporally distant rather than near perspective. Moreover, the results further showed that dyads that had a temporally distant perspective reported more effort

toward making offers that involved a concession on low- rather high-priority issues. Such appropriate concession behavior exhibited by dyads with a temporally distant perspective culminated in better outcomes for the respective parties. Both the objective as well as subjective measures used in the current experiment indicated that dyads that had a temporally distant rather than near perspective were less likely to engage in single-issue, piecemeal consideration. It is interesting to point out, nevertheless, that the results across the two measures suggest somewhat of a discrepancy between negotiators’ behavior and their perceptions of their behavior. The objective measures were based on the dialogue between negotiators, whereas the subjective measures were based on negotiators’ self-report of their behavior during the negotiation. Results based on the objective measures showed that dyads with a temporally distant perspective made an equal number of single-issue and multi-issue offers, whereas dyads with a temporally near perspective made more single-issue offers. Results based on the subjective measures, however, showed that dyads with a temporally near perspective perceived the same degree of single-issue and multi-issue consideration, whereas dyads with a temporally distant perspective perceived a greater degree of multi-issue consideration. What might explain this difference? One possibility might be that dyads that had a temporally distant perspective merely dealt with a greater number of issues when making multi-issue offers than dyads that had a temporally near perspective. Such an interpretation would be consistent with our results that showed that dyads with a temporally distant perspective dealt with more issues at least once while making an offer. If this explanation is correct, then dyads’ self-report of multi-issue consideration may have reflected the extent to which they considered all of the issues together when they made their multi-issue offers, rather than how many multi-issue offers they made per se. Indeed, when we first weighted each multi-issue offer that was made by a dyad by the number of issues contained in each offer and then analyzed the number of multi-issue offers that were made during the negotiation, results did in fact reveal that dyads in the temporally distant perspective group (M ⫽ 13.15, SD ⫽ 7.82) exhibited a greater amount of multi-issue consideration than dyads in the temporally near perspective group (M ⫽ 7.25, SD ⫽ 7.71), t(23) ⫽ 1.90, p ⫽ .07. Therefore, the pattern of results from both the objective and subjective measures is consistent with our theoretical framework, as both measures showed that negotiators who had a temporally distant rather than near perspective approached the negotiation in a more integrative manner. It is also interesting to point out that although negotiators with a temporally distant perspective achieved better outcomes than 6 Included in this postnegotiation questionnaire was an item that was directed at dyads that were not able to reach an agreement. We included an item that asked those participants that were not able to reach an agreement if it was because they ran out of time or if it was because they reached a point during the negotiation in which neither party wanted to give in or change their mind (i.e., impasse). Only 2 dyads failed to reach agreement, and both indicated it was because of time constraints. 7 We also calculated the difference in outcomes between the high- and low-scoring dyadic members. Although the discrepancy in outcomes was smaller when dyads had a temporally distant perspective (M ⫽ 15.91) instead of a temporally near perspective (M ⫽ 26.25), the difference failed to reach significance, t(21) ⫽ 1.01, p ⫽ .33, d ⫽ 0.44.

TEMPORAL DISTANCE AND NEGOTIATION

723

Proportion of Multi-issue Offers

.50**

Temporal Perspective Condition

.76*** (.74***)

.41* (.03)

Negotiators’ Joint Outcome

Figure 3. Path diagram of hypothesized model, with standardized beta weights (Experiment 3). Temporal perspective conditions were dummy coded, such that 0 ⫽ temporally near perspective and 1 ⫽ temporally distant perspective. Direct effects (i.e., values without parentheses) represent the standardized regression coefficient (␤). The value inside the parentheses represents the standard corrected regression coefficient. *p ⫽ .05. **p ⫽ .01. ***p ⬍ .001.

negotiators with a near perspective, those with a near perspective still performed quite well relative to a simple compromise on each issue. That is, had negotiators simply accepted the middle ground on each issue, the outcome would have been 130 for each negotiator and the joint outcome would have been 260. Indeed, dyads with a temporally near perspective achieved a significantly better outcome (M ⫽ 324.58) as compared with the simple compromise outcome, t(11) ⫽ 3.93, p ⫽ .002, d ⫽ 2.34, suggesting that the difference in outcomes between negotiators with a temporally near and distant perspective is more a function of the positive impact of a distant perspective rather than the negative impact of a near perspective.

General Discussion Across the current set of experiments, the data indicate that a temporally distant perspective from the realization of a negotiated agreement promotes a more structured approach toward the issues. First, we demonstrated that individuals who had a temporally distant perspective from a negotiation were equally inclined to require that proposals be made in a multi-issue or single-issue fashion, whereas individuals with a temporally near perspective were overwhelming in favor of proposals being made in a singleissue fashion (Experiment 1). Second, we demonstrated that individuals who had a temporally distant rather than near perspective from an event being negotiated were more interested in conceding on a low-priority issue and less interested in conceding on a high priority (Experiment 2). Finally, we demonstrated that dyads who had a temporally distant rather than near perspective from an event being negotiated during live negotiation made a greater proportion of multi-issue offers, dealt with more issues simultaneously, exhibited a greater likelihood of completely conceding on the lowest priority issue in exchange for a complete concession on the highest priority issue, and achieved greater individual and joint outcomes (Experiment 3). According to CLT, as individuals experience greater temporal distance from an event, they should be more likely to form a high-level construal of the event. High-level construals are global

and structured representations that clearly distinguish between the primary, central features and the secondary, peripheral features of the event. In the current set of experiments, the event we focused on was the realization of a negotiated agreement. As expected, our results showed that individuals who had a temporally distant rather than near perspective from the realization of a negotiated agreement were less likely to consider the issues surrounding the agreement in a localized, piecemeal manner and more likely to weigh their preferences for primary issues over their preferences for secondary issues when making a concession. At this point, one may be left wondering why negotiation behavior and outcomes were affected by manipulating temporal perspective in the current set of experiments when all other information provided in our near and distant conditions was the same. CLT (Trope & Liberman, 2003) assumes that temporal construal is a generalized heuristic that evolves as a result of differences in what people typically know and do about near and distant future experiences. Usually, when people think about novel events that are expected to occur in the distant future, information about the secondary aspects of those events, including the context in which they will occur and alternative courses of action that might be taken in regard to them, only become reliable as people get closer in time to experiencing them. Thus, essential information about events typically receives the most attention when people are far from experiencing them, and most issues that revolve around primary goals are generally resolved long before people experience events. As a result, people are typically left devoting most of their attention to the many incidental details about events as they get closer to experiencing them, and most issues that revolve around secondary goals are not resolved until people get closer to experiencing events. For example, when people are planning a vacation, they typically resolve issues regarding their destination and mode of travel (primary concerns) long before the vacation is set to occur, whereas issues regarding meals and clothing (secondary concerns) are not resolved until they are actually on the vacation.

HENDERSON, TROPE, AND CARNEVALE

724

CLT assumes that an association forms between temporal distance and level of construal and that this association is then overgeneralized, causing people to continue to form high-level construals for distant future events and low-level construals for near future events, even when information about the secondary aspects is reliable. So, even in situations in which information about the secondary and primary aspects of an event (e.g., an agreement) is known to be reliable, individuals with a temporally distant perspective will still construe information about the event in a more abstract or integrative fashion, focusing more on the primary or essential aspects of the event.

Temporal Distance in Negotiation Although temporal distance is often alluded to in negotiation research, it is rarely tackled empirically or theoretically. Indeed, many of the effects that emerge in short studies that restrict the temporal perspective of participants might not bear up well or at least be moderated by the impact of increased temporal perspective. Notably, the current work does extend in several ways prior research by Okhuysen and his colleagues (2003) on the effects of time discounting in negotiation. First, it is important to stress that we view the effect of temporal construal as independent of the effects of time discounting. However, we recognize, as did Okhuysen and his colleagues, that the effects of construal and discounting are likely to converge on producing beneficial negotiation outcomes. In fact, despite the fact that we did not observe evidence of discounting in our experiments, we see the results from the current set of experiments as complementing those of Okhuysen and his colleagues. We suggest that when time discounting occurs in negotiation, the value attached to secondary, low-priority issues will be discounted at a steeper rate than the value attached to primary, high-priority issues. Moreover, because the effects of temporal construal are independent of the effect of time discounting, our findings provides a useful framework for understanding why the value attached to primary, high-priority issues might be augmented when the temporal horizon to the realization of the negotiated agreement increases. This latter point highlights the danger of universally recommending that individuals negotiate with a temporally distant perspective, as concessions on primary issues are likely to be more difficult to obtain. The findings from the current research also extend prior work on temporal distance in negotiation because we identify specific behaviors (issue consideration and logrolling) that are affected by greater temporal distance, which promote more integrative solutions in negotiation.

Possible Alternative Explanations In this section, we address possible alternative explanations of our findings (i.e., explanations that are outside of CLT’s framework).

Time Pressure Perhaps the results in the current set of experiments were due to participants with a temporally near perspective feeling greater time pressure. Several features of the experiments suggest that such an explanation is unlikely. First, in Experiments 1 and 2, nothing was said to the participants about a time limit or amount of time they

had to report their judgments and decisions. Second, in Experiment 3, in which a time limit was mentioned, temporal perspective was operationalized using distance from a hypothetical event, and thus, there was no need or reason to come to an agreement more quickly. Third, in Experiment 3, the time limit was held constant across both conditions. Fourth, in Experiment 3, negotiators with a temporally near perspective actually made more offers overall than negotiators with a distant perspective, suggesting that they did not feel a greater need to come to a quick decision. Lastly, even if participants with a temporally near perspective did feel more time pressure during negotiation, one could easily make the case that they might be more inclined to engage in multi-issue offers to deal with all of the issues as quickly as possible. There is evidence that time pressure can be beneficial or harmful for negotiation, depending on the circumstances (Carnevale et al., 1993). Therefore, time pressure as an alternative explanation for the current findings seems untenable.

Cognitive Effort Perhaps the results in the present research were due to differences in effortful processing between the experimental groups. For example, is it possible that participants in the current studies who had a distant perspective felt the negotiation to be less self-relevant and therefore engaged in less effortful processing than participants who had a near perspective? More important, could such a difference in effortful processing (if it even occurred) explain the results obtained? We suspect the answer to this question is no. First, it is not clear why less effortful processing would lead to a lower preference for piecemeal issue consideration relative to multi-issue consideration, as demonstrated by the group with a distant perspective (Experiments 1 and 3). Second, it is not clear why it would lead to more interest in high-priority issues relative to low-priority issues, as also demonstrated by the group with a distant perspective (Experiments 2 and 3). Third, such an interpretation is not consistent with the lack of differences in the amount of time it took to reach agreement in negotiation in Experiment 3. Presumably more effortful negotiation would take a longer time. Overall, it seems that any differences in effortful processing between individuals with a temporally near versus distant perspective are unlikely to account for any differences in issue consideration or concession behavior (see Smith & Trope, 2006, for a related discussion)

Implications for Negotiation The findings from the current set of studies have several interesting implications for negotiation behavior. At a more theoretical level, our results provide some insights into the process that underlies the impact of temporal distance on negotiation outcomes by identifying the key role of issue consideration. Moreover, our results offer some additional insights into why people prefer to consider issues in a more piecemeal versus multi-issue fashion by demonstrating the role of temporal distance and level of construal. At a more applied level, the findings from Experiment 1 have some implications for negotiating parties preparing ahead of time before they reach the bargaining table. The majority of our participants in Experiment 1 preferred to consider issues as packages when they thought about their negotiation from a distant perspective. These results suggest that having individuals prepare for their negotia-

TEMPORAL DISTANCE AND NEGOTIATION

tions far in advance may be at least one possible procedure, in addition to explicit instruction (Weingart et al., 1996), for increasing negotiators’ likelihood of adopting a multi-issue approach during negotiation. Of course, given that negotiators’ preference for single-issue consideration is likely to increase as their negotiation gets closer in time, any decision that is made when the negotiation is farther away about how the issues will be dealt with should be binding in order to observe the benefits of the distant perspective. It is interesting that recent findings by Fo¨rster et al. (2004) suggest that even engaging in mental imagery exercises immediately prior to negotiation might decrease negotiators’ preference for single-issue consideration. Fo¨rster and his colleagues found that having participants simply imagine themselves engaging in a task at a distant point in time prior to having them actually engage in the task, led participants to process information about the task in a more global, integrative manner. Similar consequences might emerge for negotiation behavior, and future work should address such a possibility. The findings from the current set of studies also have implications for how individuals handle different types of conflicts, including conflicts over minor versus major issues (Druckman & Rozelle, 1975) and conflicts over specific interests versus broad values and ideological differences (Druckman & Broome, 1991; Druckman, Broome, & Korper, 1988; Harinck & De Dreu, 2004). Specifically, our results suggest that the resolution of conflict over things (minor issues, interests) that tend to be relatively more concrete and accompanied by local consequences should be hindered when issues are construed at a lower level. Therefore, when conflicts do in fact revolve around differences on such issues, solutions to such conflicts are likely to be facilitated by having a temporally distant perspective during negotiation. Conversely, the resolution of conflict over things (major issues, values, and ideological differences) that tend to be relatively more abstract and accompanied by global consequences should be hindered when issues are construed at a higher level. Therefore, solutions to such conflicts are likely to be facilitated by having a temporally near perspective during negotiation. Indeed, future research should examine these possibilities.

Future Directions Negotiation Context Our examination of the consequences of temporal perspective in negotiation was restricted to interpersonal, dyadic negotiation. It is important to note that the consequences of a temporally distant rather than near perspective might in fact be different in the context of intergroup negotiation. Indeed, in the context of intergroup negotiation, other factors that impact the process and outcome of a negotiation, including social identity concerns (Eggins, Haslam, & Reynolds, 2002; Smyth, 2002) and concerns about saving face (Hoobler, 2003; Mosterd & Rutte, 2000), might temper some of the conclusions drawn here about the consequences of a temporally distant perspective in negotiation. Therefore, future research should investigate the impact of temporal perspective in both the interpersonal and intergroup domain.

Social Motives The effects of social motives have been widely studied in the context of negotiation (Carnevale & Pruitt, 1992; De Dreu,

725

Weingart, & Kwon, 2000; Pruitt, 1981). Our treatment of social motives was limited in the current set of experiments, as we focused on the benefits of a distant perspective for negotiators with an individualistic orientation. Given the impact social motives have on negotiators’ behavior and outcomes, it would be worthwhile to fully explore how temporal distance within a negotiation interacts, if at all, with social motives. It is interesting that because social values typically serve as abstract psychological guides for behavior (Rokeach, 1968; Schwartz & Bilsky, 1987), one might expect, for example, that negotiators’ social motives would have more impact when making decisions about events that are set to occur in the distant future. Specifically, one might predict that a temporally distant perspective would act as a kind of magnifying glass for negotiators’ motivational orientation, with those with an individualistic orientation exhibiting even more concern for themselves and those with a cooperative orientation exhibiting even more concerns for others (cf. De Dreu & Carnevale, 2003). Indeed, results reported by Eyal, Sagristano, Liberman, and Trope (2006) are consistent with this prediction.

Relationship Between Time and Construal The present research examined the impact of temporal distance in negotiation by treating time as a categorical variable (near vs. distant). Despite the fact that most research conducted within the framework of construal level theory has manipulated temporal distance at two levels or time points (see Pennington & Roese, 2003, for an exception), it remains an open question as to how exactly temporal distance relates to level of construal. Several studies have demonstrated that events that are purportedly temporally near versus distant elicit different levels of construal (see Trope & Liberman, 2003, for a review). Nevertheless, these studies did not identify the form of the time function. Is there a linear relationship in which construals continue to become more abstract with increasing time to a future event? Is there a positively accelerated relationship in which changes in construals are small at the beginning as temporal distance is initially increased but then large as temporal distance is further increased? Is there a negatively accelerated relationship in which changes in construals are large at the beginning as temporal distance is initially increased but then small as temporal distance is further increased, such as how pay impacts employee reactions (Worley, Bowen, & Lawler, 1992)? Or is the relationship one in which time impacts construal in a categorical fashion, such as how light impacts color perceptions (Bornstein & Korda, 1984) and sounds impact speech perceptions (A. M. Liberman, Harris, Hoffman, & Griffith, 1957). That is, maybe there is a special temporal cutoff point that varies by individual, at which events that occur prior to that point are construed at the same level of concreteness and events that are occur after that point are construed at the same level of abstractness. Such questions highlight how the continuous nature of time offers new possibilities for studying the impact of temporal distance on construal in many judgment and decisionmaking domains, including negotiation. Indeed, the final word on this issue will have serious consequences for negotiation, as the extent to which a temporally distant rather than near perspective facilitates integrative agreements will depend on the extent to which the relevant point in time triggers a higher level construal in the first place.

HENDERSON, TROPE, AND CARNEVALE

726 Type of Temporal Distancing

The primary focus of this article is on the consequences of a temporally distant perspective from a negotiated agreement on the way issues were dealt with and on the quality of agreements that were reached during negotiation. It is important to note that the temporal perspective of the participants in our first experiment was in fact linked to an actual event that they believed would occur in the future. Nevertheless, the present research was limited in that the remaining experiments examined the consequences of temporal perspective in negotiation by manipulating temporal distance from a hypothetical event. It would be worthwhile for future research to develop methods for studying temporal distancing from actual events, rather than events that are imagined to be near or distant, as imagined temporal distancing might produce weaker effects than actual temporal distancing (see Carnevale & De Dreu, 2006, for a review of methods for the study of time in negotiation). Notably, there are methodological challenges to implementing actual temporal distancing and possible confounds that are not present with hypothetical temporal distancing. Therefore, future research should be sensitive to the complexity surrounding this issue.

Psychological Distance In general, we are interested in what impact psychological distance has on the resolution of conflict between individuals. According to CLT (see N. Liberman, et al., in press; Trope & Liberman, 2003), the same general principles that apply to temporal distance should also hold for other distance dimensions. Most relevant to the negotiation setting, the amount of spatial distance that individuals experience during negotiation is also likely to promote a more structured approach toward the issues. Recent work has found, for example, that participants who experienced greater spatial distance from events used more abstract language to describe events and identified more overarching, superordinate goals for events (Fujita, Henderson, Eng, Trope, & Liberman, 2006). In addition, recent work has found that participants who experienced greater spatial distance from events relied more on abstract, global information when making predictions about the likelihood that events would occur (Henderson, Fujita, Trope, & Liberman, in press). Just as the amount of temporal distance within a negotiation can be a multifaceted component in a negotiation setting, so can the amount of spatial distance. Indeed, negotiators may come to an agreement in one setting (e.g., United Nations Headquarters in New York) and expect to implement or realize the agreement in another setting that is geographically close (United States) or far away (Africa). Moreover, with the ever-increasing online activity in our society, more negotiations are beginning to take place across great physical distances (Carnevale & Probst, 1997). In both cases, such distances may affect the way individuals consider issues during negotiation, the relative focus they place on primary concerns, and the general integrative approach they take toward reaching an agreement.

Conclusion Across three experiments, we have demonstrated that as individuals’ temporal perspective is increased during negotiation, there

is less constraint on type of negotiation process that can be applied. Specifically, we have provided the first evidence that increased temporal distance from a negotiated agreement facilitates less single-issue consideration and more appropriate concessions, resulting in more mutually beneficial outcomes during negotiation. Although interpersonal conflict is ubiquitous, the current research suggests that those who choose to resolve conflict through open discussion may benefit from having a temporally distant perspective during their discussion, as such a perspective promotes allocating more resources to resolving central, primary concerns rather than less important, incidental concerns. Future research should continue to explore what consequences psychological distance and level of construal have in a negotiation setting for the resolution of interpersonal conflict.

References Bazerman, M., & Neale, M. (1992). Negotiating rationally. New York: Free Press. Bazerman, M. H., Curhan, J. R., Moore, D. A., & Valley, K. L. (2000). Negotiation. Annual Review of Psychology, 51, 279 –314. Bornstein, M. H., & Korda, N. O. (1984). Discrimination and matching within and between hues measured by reaction times: Some implications for categorical perception and levels of information processing. Psychological Research, 46, 207–222. Carnevale, P. J., & De Dreu, C. K. W. (Eds.). (2006). Methods of negotiation research. Leiden, the Netherlands: Martinus Nijhoff. Carnevale, P. J., O’Connor, K., & McCusker, C. (1993). Time pressure in negotiator and mediator decision making. In O. Svenson & J. Maule (Eds.), Time pressure and stress in human judgment and decision making (pp. 117–127). Cambridge, England: Cambridge University Press. Carnevale, P. J., & Probst, T. M. (1997). Conflict on the Internet. In S. Kiesler (Ed.), Culture of the Internet (pp. 233–255). Mahwah, NJ: Erlbaum. Carnevale, P. J., & Pruitt, D. G. (1992). Negotiation and mediation. Annual Review of Psychology, 43, 531–582. De Dreu, C. K. W. (2003). Time pressure and closing of the mind in negotiation. Organizational Behavior and Human Decision Processes, 91, 280 –295. De Dreu, C. K. W., & Carnevale, P. J. (2003). Motivational bases of information processing and strategy in negotiation and social conflict. In M. P. Zanna (Ed.), Advances in experimental social psychology (pp. 235–291). New York: Academic Press. De Dreu, C. K. W., Weingart, L. R., & Kwon, S. (2000). Influence of social motives on integrative negotiation: A meta-analytic review and test of two theories. Journal of Personality and Social Psychology, 78, 889 – 905. Druckman, D. (1994). Determinants of compromising behavior in negotiation: A meta-analysis. Journal of Conflict Resolution, 38, 507–556. Druckman, D., & Broome, B. J. (1991). Value differences and conflict resolution: Familiarity or liking? Journal of Conflict Resolution, 35, 571–593. Druckman, D., Broome, B. J., & Korper, S. H. (1988). Value differences and conflict resolution: Facilitation or delinking? Journal of Conflict Resolution, 32, 489 –510. Druckman, D., & Rozelle, R. M. (1975). Performance evaluation as a determinant of willingness to defend a counterattitudinal position. Social Behavior and Personality, 3, 243–252. Eggins, R. A., Haslam, S. A., & Reynolds, K. J. (2002). Social identity and negotiation: Subgroup representation and superordinate consensus. Personality and Social Psychology Bulletin, 28, 887– 899. Erickson, B., Holmes, J. G., Frey, R., Walker, L., & Thibaut, J. (1974). Functions of a third party in the resolution of conflict: The role of a

TEMPORAL DISTANCE AND NEGOTIATION judge in pretrial conferences. Journal of Personality and Social Psychology, 31, 864 – 872. Eyal, T., Sagristano, M. D., Liberman, N., & Trope, Y. (2006). Resolving value conflicts in planning the future. Unpublished data. Fo¨rster, J., Friedman, R. S., & Liberman, N. (2004). Temporal construal effects on abstract and concrete thinking: Consequences for insight and creative cognition. Journal of Personality and Social Psychology, 87, 177–189. Froman, L. A., Jr., & Cohen, M. D. (1970). Compromise and logroll: Comparing the efficiency of two bargaining processes. Behavioral Science, 15, 180 –183. Fujita, K., Henderson, M. D., Eng, J., Trope, Y., & Liberman, N. (2006). Spatial distance and mental construal of social events. Psychological Science, 17, 278 –282. Harinck, F., & De Dreu, C. K. W. (2004). Negotiating interests or values and reaching integrative agreements: The importance of time pressure and temporary impasses. European Journal of Social Psychology, 34, 595– 611. Hayes-Roth, B. (1977). Evolution of cognitive structure and process. Psychological Review, 84, 260 –278. Henderson, M. D., Fujita, K., Trope, Y, & Liberman, N. (in press). Transcending the “here”: The effect of spatial distance on social judgment. Journal of Personality and Social Psychology. Hoobler, G. (2003). Management of issues and relationships during international conflict management: Or how (not?) to end a war. International Journal of Conflict Management, 14, 297–317. Kelley, H. H. (1966). A classroom study of dilemmas in interpersonal negotiations. In K. Archibald (Ed.), Strategic intervention and conflict (pp. 49 –73). Berkeley, CA: University of California, Institute of International Studies. Kelley, H. H., & Schenitzki, D. P. (1972). Bargaining. In C. McClintock (Ed.), Experimental social psychology (pp. 298 –337). New York: Holt, Rinehart, & Winston. Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social psychology. In D. T. Gilbert, & S. T. Fiske (Eds.), The handbook of social psychology (pp. 233–265). New York: McGraw-Hill. Liberman, A. M., Harris, K. S., Hoffman, H. S., & Griffith, B. C. (1957). The discrimination of speech sounds within and across phoneme boundaries. Journal of Experimental Psychology, 54, 358 –368. Liberman, N., Sagristano, M. D., & Trope, Y. (2002). The effect of temporal distance on level of mental construal. Journal of Experimental Social Psychology, 38, 523–534. Liberman, N., & Trope, Y. (1998). The role of feasibility and desirability considerations in near and distant future decisions: A test of temporal construal theory. Journal of Personality and Social Psychology, 75, 5–18. Liberman, N., Trope, Y., & Stephan, E. (in press). Psychological distance. In E. T. Higgins & A. W. Kruglanski (Eds.), Social psychology: Handbook of basic principles (Vol. 2). New York: Guilford Press. Lytle, A. L., Brett, J. M., & Shapiro, D. L. (1999). The strategic use of interests, rights and power to resolve disputes. Negotiation Journal, 15, 31– 49. Mannix, E. A., Thompson, L. L., Bazerman, M. H. (1989). Negotiation in small groups. Journal of Applied Psychology, 74, 508 –517. Medin, D. L., & Ortony, A. (1989). Psychological essentialism. In S. Vosniadou & A. Ortony (Eds.), Similarity and analogical reasoning (pp. 179 –195). Cambridge, England: Cambridge University Press. Moore, D. A. (2004). Myopic prediction, self-destructive secrecy, and the unexpected benefits of revealing final deadlines in negotiation. Organizational Behavior and Human Decision Processes, 94, 125–139. Mosterd, I., & Rutte, C. G. (2000). Effects of time pressure and account-

727

ability to constituents on negotiation. International Journal of Conflict Management, 11, 227–247. O’Connor, K. M., & Carnevale, P. J. (1997). A nasty but effective negotiation strategy: Misrepresentation of a common-value issue. Personality and Social Psychology Bulletin, 23, 504 –515. Okhuysen, G. A., Galinsky, A. D., & Uptigrove, T. A. (2003). Saving the worst for last: The effect of time horizon on the efficiency of negotiating benefits and burdens. Organizational Behavior and Human Decision Processes, 91, 269 –279. Olekalns, M., & Smith, P. L. (2003). Testing the relationships among negotiators’ motivational orientations, strategy choices, and outcomes. Journal of Experimental Social Psychology, 39, 101–117. Pennington, G. L., & Roese, N. J. (2003). Regulatory focus and temporal distance. Journal of Experimental Social Psychology, 39, 563–576. Pruitt, D. G. (1981). Negotiation behavior. New York: Academic. Pruitt, D. G., & Carnevale, P. J. (1993). Negotiation in social conflict. Pacific Grove, CA: Brooks/Cole. Pruitt, D. G., & Lewis, S. A. (1975). Development of integrative solutions in bilateral negotiation. Journal of Personality and Social Psychology, 31, 621– 633. Reder, L., M., & Anderson, J. R. (1980). A partial resolution of the paradox of interference: The role of integrating knowledge. Cognitive Psychology, 12, 447– 472. Rokeach, M. (1968). A theory of organization and change within valueattitude systems. Journal of Social Issues, 24, 13–32. Schul, Y. (1983). Integration and abstraction in impression formation. Journal of Personality and Social Psychology, 44, 45–54. Schwartz, S. H., & Bilsky, W. (1987). Toward a universal psychological structure of human values. Journal of Personality and Social Psychology, 53, 550 –562. Smith, P. K., & Trope, Y. (2006). You focus on the forest when you’re in charge of the trees: Power priming and abstract information processing. Journal of Personality and Social Psychology, 90, 578 –596. Smyth, L. F. (2002). Identity-based conflicts: A systemic approach. Negotiation Journal, 18, 147–161. Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. In S. Leinhardt (Ed.), Sociological methodology (pp. 290 –312). Washington, DC: American Sociological Association. Thompson, L. L. (1991). Information exchange in negotiation. Journal of Experimental Social Psychology, 27, 161–179. Tinsley, C. H. (2001). How negotiators get to yes: Predicting the constellation of strategies used across cultures to negotiate conflict. Journal of Applied Psychology, 86, 583–593. Trope, Y., & Liberman, N. (2000). Temporal construal and time-dependent changes in preference. Journal of Personality and Social Psychology, 79, 876 – 889. Trope, Y., & Liberman, N. (2003). Temporal construal. Psychological Review, 110, 403– 421. Weingart, L. R., Bennett, R. J., & Brett, J. M. (1993). The impact of consideration of issues and motivational orientation on group negotiation process and outcome. Journal of Applied Psychology, 78, 504 –517. Weingart, L. R., Hyder, E. B., & Prietula, M. J. (1996). Knowledge matters: The effect of tactical descriptions on negotiation behavior and outcome. Journal of Personality and Social Psychology, 70, 1205–1217. Worley, C. G., Bowen, D. E., & Lawler, E. E. (1992). On the relationship between objective increases in pay and employees’ subjective reactions. Journal of Organizational Behavior, 13, 559 –571. Yukl, G. A., Malone, M. P., Hayslip, B., & Pamin, T. A. (1976). The effects of time pressure and issue settlement order on integrative bargaining. Sotiometry, 39, 277–281.

(Appendixes follow)

HENDERSON, TROPE, AND CARNEVALE

728

Appendix A Point Value Assigned to Options for Each Issue (Experiment 2) Option

Point Value Temperature

A. 77° B. 74° C. 71° D. 68° E. 65°

60 30 10 0 0 Advertising

A. combined campaign, advertising for market as a whole, costs to be divided equally among the market merchants B. combined campaign, advertising for market as a whole, to be paid according to percentage of the market’s gross profits contributed by the merchant C. combined campaign, advertising the stores as individual units but on the same flyers, with each member given (and paying for) 1/4th of ad D. separate campaign for each member, 6% of expected gross profits to be spent on advertising E. separate campaign for each member, amount spent up to individual merchants

0 40 30 10 20

Clerks A. hire by group, train by group, distribute equally, paid for by group B. hire by group, train by group, distribute according to floor space, paid for by group C. hire by group, train individually, distribute according to demand for service, paid for by group D. hire individually, train individually, distribute according to demand for service, each merchant to pay from individual profits E. hire individually, train individually, each merchant to decide how many clerks, each merchant to pay from individual profits

0 0 20 60 100

Maintenance A. shared, each responsible for 1/4th of total costs B. shared, each responsible for percentage according to floor space occupied C. shared, each responsible for percentage according to floor space occupied, but with bakery paying double its percentage because of the nature of its carry-out business D. separate, each responsible for own floor space, plus common area maintenance cost as a function of floor space occupied E. separate, each responsible for own floor space plus equal contributions for common area maintenance

0 60 20 30 10

Position A. spontaneous purchases near entrance B. smaller departments near entrance C. common area near entrance D. convenient location for dept. with highest volume of sales E. merchants stocking heavier products should be located near entrance/exits

80 60 0 0 0

TEMPORAL DISTANCE AND NEGOTIATION

729

Appendix B Point Value Assigned to Options for Each Issue (Experiment 3) Issue Negotiator and option

1

2

3

4

A B C D E

60 30 10 5 0

40 30 20 10 0

80 60 40 30 10

120 80 60 40 10

A B C D E

0 5 10 30 60

10 40 60 80 120

10 30 40 60 80

0 10 20 30 40

1

2

Received August 10, 2005 Revision received April 7, 2006 Accepted May 3, 2006 䡲

PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES

Psychological Resilience, Positive Emotions, and Successful Adaptation to Stress in Later Life Anthony D. Ong and C. S. Bergeman

Toni L. Bisconti

University of Notre Dame

University of New Hampshire

Kimberly A. Wallace University of Montana In 3 studies, the authors investigated the functional role of psychological resilience and positive emotions in the stress process. Studies 1a and 1b explored naturally occurring daily stressors. Study 2 examined data from a sample of recently bereaved widows. Across studies, multilevel random coefficient modeling analyses revealed that the occurrence of daily positive emotions serves to moderate stress reactivity and mediate stress recovery. Findings also indicated that differences in psychological resilience accounted for meaningful variation in daily emotional responses to stress. Higher levels of trait resilience predicted a weaker association between positive and negative emotions, particularly on days characterized by heightened stress. Finally, findings indicated that over time, the experience of positive emotions functions to assist high-resilient individuals in their ability to recover effectively from daily stress. Implications for research into protective factors that serve to inhibit the scope, severity, and diffusion of daily stressors in later adulthood are discussed. Keywords: adaptation, positive emotion, recovery, resilience

negative emotions are interconnected in times of stress (Zautra, 2003; Zautra, Affleck, Tennen, Reich, & Davis, 2005) or how adaptive outcomes in later life can be reached by a variety of different pathways (Bergeman & Wallace, 1999; Ryff, Singer, Love, & Essex, 1998). In short, we know relatively little about the essential nature of successful adaptation to stress, how it unfolds over time and across contexts, and still less about its significance in late life. In this article, we examine how different protective factors shape and modify the unfolding experience of daily stress and emotion in later adulthood. The everyday challenges that accumulate in late life provide a natural context in which to investigate the mechanisms that underlie successful adaptation in the face of adversity (see Kling, Seltzer, & Ryff, 1997; Smider, Essex, & Ryff, 1996). Building on prior investigations of later life resilience (e.g., Ryff et al., 1998; Staudinger, Marsiske, & Baltes, 1993, 1995), we argue that there are multiple routes through which successful adaptation to stress might occur. First, adaptation might be reflected in the capacity to maintain positive outcomes in the face of untoward life events (Ryff & Singer, 1998; Staudinger et al., 1995). This conceptualization of adaptation converges with several distinct lines of recent work on the nature of affective relationships under stress (Cacioppo, Larsen, Smith, & Berntson, 2004; Zautra, Smith, Affleck, & Tennen, 2001), suggesting that the capacity to maintain and preserve the boundaries between positive and negative emotional states may represent one potential pathway

Although emotions have long been viewed as serving an adaptive function in times of stress (Folkman & Lazarus, 1985; Frijda, 1986, 1987, 1988; Levenson, 1988), the vast majority of research on emotions has focused on how processes go awry and lead to illness, dysfunction, and disorder (for reviews, see Fredrickson, 1998, 2001). With little exception (i.e., Folkman, 1997; Fredrickson, Tugade, Waugh, & Larkin, 2003; Zautra, Johnson, & Davis, 2005), there remain few countervailing studies of the role of positive emotions in the stress process, particularly in later adulthood in which assessments of psychopathology have been the norm (Ong & Bergeman, 2004a; Ryff, 1989, 1995). Rarer still are studies that shed light on the many ways in which positive and

Anthony D. Ong and C. S. Bergeman, Department of Psychology, University of Notre Dame; Toni L. Bisconti, Department of Psychology, University of New Hampshire; Kimberly A. Wallace, Department of Psychology, University of Montana. Preparation of this article was supported in part by National Institute on Aging Grant 1 RO3 AG18570-01, National Institute of Health Grant 1 RO3 MH53895-01, and a grant from the William Kirby Endowment for Research, University of Notre Dame. We thank Alex Zautra, Mario Mikulincer, Scott Maxwell, and Sy-Miin Chow for their thoughtful comments on previous versions of this article. Correspondence concerning this article should be addressed to Anthony D. Ong, who is now at the Department of Human Development, G77 Martha Van Rensselaer Hall, Cornell University, Ithaca, NY 14853-4401.

Journal of Personality and Social Psychology, 2006, Vol. 91, No. 4, 730 –749 Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.730

730

PSYCHOLOGICAL RESILIENCE AND ADAPTATION

underlying flexible adaptation. Successful adaptation may also be reflected in the capacity to recover more quickly from environmental stressors (Carver & Scheier, 1999; Davidson, 2000). In this view, stress is thought to evoke elevated negative emotional arousal that lingers for variable amounts of time, and certain homeostatic mechanisms function to speed the return to baseline levels of arousal. Finally, both resistance to and recovery from stress may, in turn, be linked to selective individual difference variables (Davis, Zautra, & Smith, 2004; Zautra, Affleck, et al., 2005). This integrative perspective suggests that equally important to delineating the diverse processes that lead to successful adaptation is identifying the broad protective factors that facilitate or contribute to sustaining the adaptive process (for reviews, see Ong & Bergeman, 2004b; Reich, Zautra, & Davis, 2003). This article examines the question of how psychological resilience and daily positive emotions influence the experience of negative emotions in times of stress. Using a multilevel daily process design, we examine data from three samples to address (a) the moderating and mediating role that positive emotions can play in strengthening daily resistance to and recovery from stress and (b) the contribution of psychological resilience in shaping daily resistance and recovery processes. Throughout, we argue that an integrative approach to positive adaptation in later adulthood necessitates an understanding of how certain individuals are able to maintain and recover emotional well-being despite the presence of daily challenge and adversity.

Why Positive Emotions Facilitate Adaptation to Stress Multiple studies have shown that positive emotions have a wide range of effects on individuals (for reviews, see Lyubomirsky, King, & Diener, 2005; Pressman & Cohen, 2005). Both theoretical and empirical work indicate that positive emotions promote flexibility in thinking and problem solving (Fredrickson & Branigan, 2005; Isen, Daubman, & Nowicki, 1987), counteract the physiological effects of negative emotions (Fredrickson & Levenson, 1998; Ong & Allaire, 2005), facilitate adaptive coping (Folkman & Moskowitz, 2000a, 2004), build enduring social resources (Fredrickson & Branigan, 2001; Keltner & Bonanno, 1997), and spark upward spirals of enhanced well-being (Fredrickson, 2000; Fredrickson & Joiner, 2002). Notably, positive emotions can cooccur with negative emotions with relatively high frequency, even in the midst of personally significant stress (Moskowitz, Folkman, Collette, & Vittinghoff, 1996; Ong, Bergeman, & Bisconti, 2004). For instance, in a study of AIDS-related caregiving and bereavement, Folkman (1997) reported that with the exception of the period immediately before and after their partner’s death, the positive emotion scores of men whose partners had died of AIDS did not reliably differ from their negative emotion scores, and at 3 months postloss had returned to prebereavement levels. Similarly, Keltner and Bonanno (1997) observed that Duchenne laughter and smiling were exhibited at least once by a majority of conjugally bereaved participants as they discussed their interpersonal loss. One way by which positive emotions may play a pivotal role in adaptation has been proposed by Zautra, Smith, Affleck, and Tennen (2001) in their dynamic model of affect (DMA). In contrast to other models of stress and coping, which view emotional adaptation entirely in terms of regulating psychological distress,

731

the DMA takes into account both negative and positive emotions in the stress process. The model predicts that under ordinary circumstances, positive and negative emotions are relatively independent, whereas during stressful encounters an inverse correlation between positive and negative emotions increases sharply (for a review, see Reich et al., 2003). One implication of the DMA is that positive emotions are more likely to diminish negative emotions on days of elevated stress. The model also predicts that a relative deficit in positive emotional experience should leave individuals more vulnerable to the effects of stress. Supportive evidence for the DMA comes from research demonstrating that during stressful periods, emotions are experienced along a single continuum in adults coping with chronic health conditions (Potter, Zautra, & Reich, 2000; Zautra et al., 2001), laboratory manipulations of stress (Zautra, Reich, Davis, Potter, & Nicolson, 2000), as well as everyday life events (Ong & Bergeman, 2004a; Zautra, Affleck, et al., 2005). Taken together, these prior investigations suggest that the experience of positive emotions amid challenge and adversity may contribute to stress resistance, and hence adaptation, by interrupting the ongoing experience of negative emotions during times of stress. In addition to offsetting the immediate adverse consequences of stress, positive emotions may also play an important role in recovery processes. Fredrickson’s (1998, 2001) broaden-and-build model of positive emotions raises the possibility that positive emotions are important facilitators of adaptive recovery, quieting or undoing the autonomic arousal generated by negative emotions. In several laboratory studies in which positive and negative emotions were experimentally induced, Fredrickson and colleagues (Fredrickson & Levenson, 1998; Fredrickson, Mancuso, Branigan, & Tugade, 2000) found that positive emotions were linked to faster cardiovascular recovery from negative emotional arousal. More recent investigations confirm the importance of positive emotions in fostering recovery from stressful life events (Fredrickson et al., 2003; Tugade & Fredrickson, 2004; see also Zautra, Johnson, & Davis, 2005). Taken together, theoretical and empirical work indicate that positive emotions may have both a protective and restorative function, guarding individuals from negative emotions as well as quelling the aftereffects of such emotions.

How Positive Emotions Arise in the Context of Stress What psychological traits are implicated in the generation and maintenance of positive emotions in the face of stress? An emerging adult literature suggests that individual differences in psychological resilience may account for the adaptive ways in which life stressors are encountered, managed, and transformed. Theoretical writings indicate that psychological resilience is a relatively stable personality trait characterized by the ability to overcome, steer through, and bounce back from adversity (J. Block & Kremen, 1996; J. H. Block & Block, 1980). Recent research, moreover, suggests that positive emotions are a crucial component of trait resilience (Tugade & Fredrickson, 2004; Tugade, Fredrickson, & Barrett, 2004). Rather than being a simple by-product of resilience, however, the experience of positive emotion is thought to have adaptive benefits in the coping process (for reviews, see Folkman & Moskowitz, 2000a, 2004). Empirical support for this prediction comes from research demonstrating that resilient individuals tend to draw on positive emotion-eliciting coping strategies such as

732

ONG, BERGEMAN, BISCONTI, AND WALLACE

benefit finding and positive reappraisal (Affleck & Tennen, 1996; Folkman & Moskowitz, 2000b), humor and infusing ordinary events with positive meaning (Folkman, Moskowitz, Ozer, & Park, 1997; Ong et al., 2004), and goal-directed problem-focused coping (Billings, Folkman, Acree, & Moskowitz, 2000; Folkman, 1997) to regulate negative emotional experiences. Taken as a whole, these findings indicate that traits (e.g., psychological resilience) with functional properties associated with positive emotions may serve to strengthen resistance to stress by affording greater access to positive emotional resources (Ong & Bergeman, 2004a; Tugade et al., 2004), which, in turn, may help to provide a momentary respite from ongoing stressful experiences (Folkman & Moskowitz, 2000a; see also, Zautra, Johnson, & Davis, 2005). In addition to promoting stress resistance, a growing number of studies suggest that individual differences in psychological resilience predict accelerated recovery from stressful situations. In a series of coordinated experimental and individual difference studies, Fredrickson and colleagues (Fredrickson et al., 2003; Tugade et al., 2004) found that high-resilient individuals exhibited faster physiological and emotional recovery from stress. In one study (Tugade et al., 2004), higher trait resilience was linked to quicker cardiovascular recovery following a laboratory stressor. In another study (Fredrickson et al., 2003), higher trait resilience was associated with lower subsequent depressive symptoms. Most notably, the effect of trait resilience on duration of cardiovascular reactivity and depressive symptoms was mediated by subjective reports of positive emotion (Fredrickson et al., 2003; Tugade et al., 2004).1 Although far from definitive, the available empirical evidence suggests that psychological resilience is associated with resistance to and recovery from stressful life events, and positive emotions may be the underlying mechanism by which high-resilient individuals achieve their adaptive outcomes. A number of unresolved questions remain, however. One major question is whether previous findings generalize to older populations. Older individuals are at higher risk for many diseases, both acute and chronic (for reviews, see Hawkley & Cacioppo, 2004; Smith, 2003). In addition, older adults may be especially likely to experience certain psychosocial stressors, such as spousal caregiving and bereavement (Moss, Moss, & Hansson, 2001). Because negative life events and chronic life conditions are more likely to accumulate with age, studies are needed that clarify how certain older adults are able to maintain and regain emotional health in the face of ongoing stress (Ong & Bergeman, 2004b). A related gap in the literature is the relative dearth of daily process studies that track the real-world adaptational processes of individuals, particularly older adults, intensively over time (for a discussion, see Almeida, 2005; Mroczek, Spiro, & Almeida, 2003). Additionally needed, therefore, are studies that sharpen understanding of the ways in which older adults effectively negotiate stressors in their everyday lives (Mroczek et al., 2003). Finally, extant studies of resilience, from childhood to old age, have given limited attention to the dynamic interplay between process and trait conceptualizations of resilience (cf. J. Block & Kremen, 1996; Luthar, Cicchetti, & Becker, 2000). Crucially needed are empirical investigations that further elucidate how stable personality traits influence and support meaningful shortterm adaptation to stress (Fleeson, 2004).

Overview of Research What role do daily positive emotions play in fostering resistance to and recovery from stress? What psychological traits influence the capacity to maintain and regain emotional well-being in the face of stress? The current research was designed to address these questions. Study 1a used diary data to explore the moderating and mediating roles that positive emotions play in promoting daily resistance to and recovery from stress and the contribution of psychological resilience in shaping daily resistance and recovery processes. Study 1b was an empirical replication of Study 1a using an independent sample and different measures of trait resilience and daily emotions. Study 2 provided a critical extension of the relationships observed in Studies 1a and 1b to a sample of recently bereaved older widows. Throughout, we predicted that (a) daily variations in positive emotions would promote both resistance to and recovery from stress and (b) the adaptive benefits that ensue from daily positive emotions are rooted in individual differences in psychological resilience.

Study 1a Study 1a was designed to provide an initial examination of the daily emotional processes associated with psychological resilience. Recent reviews of the resilience literature have underscored the need for greater operational precision in the (a) measurement of threat or challenge to the individual, (b) specification of criteria by which adaptation is judged to be successful, and (c) identification of attributes of the individual or ecological context that may help to shed light on the pathways through which effective negotiation of adversity is differentially expressed (for a discussion, see Luthar & Cicchetti, 2000). Throughout this investigation, the appraisal of threat or harm to the individual was considered an important indicator of the subjective experience of stress, the maintenance and recovery of emotional well-being were judged as markers of successful adaptation, and psychological resilience was examined as a potentially important individual difference factor that contributes to flexible adaptation to stress. Following previous research, we hypothesized that elevations in positive emotions during times of heightened stress would be particularly important in the regulation of negative emotions (Zautra, Johnson, & Davis, 2005, Zautra et al., 2001). On the basis of findings from previous laboratory investigations of positive emotions (Fredrickson & Levenson, 1998; Fredrickson et al., 2000), we also predicted that positive emotions would aid in the recovery from daily stress. Because resilient individuals are characterized by high positive emotionality (Tugade & Fredrickson, 2004), we further predicted that the experience of positive emotions would be an important resource that contributes to stress resistance, assisting high-resilient individuals in their ability to effectively regulate negative emotional arousal in the face of ongoing stress. Finally, 1 Increasing evidence from neuropsychological studies of brain activity (e.g., Davidson, 2000; Davidson, Jackson, & Kalin, 2000) further suggest that individuals who recover more quickly from emotional challenge are those who show less activation in the amygdala and more activation in the left prefrontal cortex, a focal area in the brain implicated in the experience of positive emotion. These effects, moreover, are present within the first year of life (Davidson & Fox, 1982).

PSYCHOLOGICAL RESILIENCE AND ADAPTATION

because positive emotions have been shown to play a mediating role between psychological resilience and stress recovery (Fredrickson et al., 2003; Tugade et al., 2004), we predicted that the effect of psychological resilience on emotional recovery from stress would be transmitted at least partially through the experience of daily positive emotions.

Method Participants Participants were randomly selected from a proband sample of 226 individuals who had previously participated in the Notre Dame Family Study of Aging. Forty-five participants were contacted and invited to participate in a study of daily stress and emotion. Twenty-seven participants, ages 62– 80 years (M ⫽ 72.09, SD ⫽ 5.29), agreed to take part in the 45-day study. Nearly half of the participants were women (48%; men, 52%) and married (52%) at the time of the study. Participants were predominantly European American (95.7%; African American, 4.3%) and half (52%) were educated through high school. Income was approximately normally distributed with 22.7% reporting family income less than $14,999, 18.2% between $15,000 and $24,999, 45.5% between $25,000 and $40,000, and 13.6% reporting income greater than $40,000. The characteristics of the sample, in general, reflect the Northern Indiana area. There were no significant differences in age, gender, or educational status for those who did not complete the study. Participants received a $5 gift certificate for each week of assessment completed, for a total of $30.

Procedure Prior to the daily assessment phase of the study, participants completed a trait measure of psychological resilience. The daily data are from a 45-day study in which participants received a packet of diaries every 2 weeks. Each diary contained 14 days of response sheets. Each response sheet contained 28 emotion items traditionally assessed in dimensional measures of positive and negative affect (e.g., Watson, Clark, & Tellegen, 1988). In addition, participants completed a single item on the most stressful event of the day and then rated their perceptions of how stressful the event was. Participants were instructed to respond to the daily items in the evening and return the completed diaries at the end of each 2-week period. The total number of days participants were in the study ranged from 35 to 42 (M ⫽ 37.4, SD ⫽ 3.6). The total number of days in the study for all participants was 1,215 (27 participants ⫻ 45 days). The total number of days of data the participants provided was 1,118 (92% complete).

Measures Psychological resilience. The Ego-Resilience Scale (J. Block & Kremen, 1996) was used to assess psychological resilience, defined as “the capacity of the individual to effectively modulate and monitor an everchanging complex of desires and reality constraints” (J. Block & Kremen, 1996, p. 359). The scale consists of 14 items, each responded to on a 4-point Likert scale, ranging from 1 (strongly disagree) to 4 (strongly agree). Sample items include “I get over anger with someone reasonably quickly” and “I enjoy dealing with new and unusual situations.” For this sample, the Cronbach’s alpha reliability was .72. J. Block and Kremen’s (1996) reported alpha was .76. Positive and negative emotions. Daily positive and negative emotions were measured with the daily form of the Positive and Negative Affect Schedule (PANAS; Watson et al., 1988). Participants were asked to indicate the extent to which they had experienced a range of emotions throughout the day. Ratings were made on a 5-point scale, ranging from 1 (very slightly or not at all) to 5 (extremely). The original PANAS consists of 10 items from the Negative Activation subscale (afraid, ashamed, distressed,

733

guilty, hostile, irritable, jittery, nervous, scared, upset) and 10 items from the Positive Activation subscale (active, alert, attentive, determined, enthusiastic, excited, inspired, interested, proud, strong). In addition to the original PANAS items, we included eight additional low-arousal items (cheerful, satisfied, relaxed, self-assured, depressed, worried, lonely, miserable) from selected octants of the mood circumplex (Feldman, 1995b). The final 28-item daily emotion measure represents a broad range of prototypical pleasant and unpleasant emotional states. Over all daily reports, moderate intercorrelations were observed between negative and positive emotion scales (r ⫽ ⫺.23, p ⬍ .001). Stress. In addition to reporting on their daily emotions, participants completed a single item on the most stressful event of the day and then rated their perceptions of how stressful the event was on a 5-point scale, ranging from 1 (not very stressful) to 5 (very stressful).

Results and Discussion Descriptive Statistics Preliminary analyses were conducted to obtain descriptive statistics and correlations among the person- and day-level variables. The daily variables were centered within each participant and aggregated across time. In comparison with negative emotion scores (M ⫽ 1.35, SD ⫽ 0.78), positive emotion scores were higher and more variable (M ⫽ 2.97, SD ⫽ 0.93). Overall, higher stress was associated with lower positive emotion (r ⫽ ⫺.35, p ⬍ .05) and higher negative emotion (r ⫽ .44, p ⬍ .05).2 Trait resilience, moreover, was significantly correlated with positive emotion (r ⫽ .41, p ⬍ .05) and stress (r ⫽ ⫺.38, p ⬍ .05) but was unrelated to negative emotion (r ⫽ ⫺.11, ns).

Overview of Multilevel Level Modeling Analyses We tested our hypotheses using multilevel random coefficient modeling (MRCM; Raudenbush & Bryk, 2002). The flexibility of MRCM provides a number of advantages. First, MRCM is appropriate for diary data. In the current study, the data have a hierarchical structure with up to 45 daily observations nested within each of 27 participants. Second, MRCM does not require that all individuals be measured at all occasions. We can use the data from participants who entered the study after it began and from participants who have missing data for some occasions of the study. Third, in MRCM, more reliable units of observation contribute more to the estimation of parameters than less reliable units, a process known as precision weighting (for a discussion, see Bryk & Raudenbush, 1992, pp. 32–57). By separating true and error variance, MRCM thus provides more accurate and robust estimates of parameters than ordinary least squares regression analyses. Finally, a multilevel-modeling approach allows for the simultaneous estimation of day- and person-level effects. Day-level effects address links between variables at the withinperson level and yield slope and intercept coefficients to index these relations (e.g., “On days in which individuals report high stress, do they also exhibit elevated negative emotions?”). In 2 Summary within-person correlations were converted to Fisher’s z⬘ equivalents (Cohen & Cohen, 1983), which were weighted on the basis of their estimated standard errors, averaged, and evaluated for significance. Reported values reflect the reconversion of averaged Fisher’s z⬘ scores back to r values to facilitate interpretation.

ONG, BERGEMAN, BISCONTI, AND WALLACE

734

Table 1 Parameter Estimates for Daily Negative Emotions Variable

B

t

df

p⬍

Intercept Stress Positive emotions Stress ⫻ Positive Emotions Stress ⫻ Trait Resilience Positive Emotions ⫻ Trait Resilience Stress ⫻ Positive Emotions ⫻ Trait Resilience

16.537 0.381 ⫺0.073 ⫺0.338 ⫺0.293 ⫺0.059 ⫺0.235

17.56 6.21 ⬍1 ⫺5.43 ⫺4.68 ⬍1 ⫺2.83

22 925 925 925 925 925 925

.001 .01 ns .01 .01 ns .05

Note. All day-level predictors were group–mean centered, and all person-level predictors were centered on sample means.

comparison, person-level effects address the relation between within-person coefficients and between-person variables (e.g., “Do high-resilient individuals also evidence a lower level of average negative emotion?”). In the current investigation, we also asked questions that assessed the interaction between our day-level variables (e.g., “On days in which people report high levels of positive emotion is there a weaker relation between stress and negative emotion?”). Finally, we assessed interactions across day and person levels (e.g., “Is the daily association between stress and negative emotion different in individuals who are low as opposed to high in psychological resilience?”). The first set of analyses examined the reliability of the day-level measure of negative emotion and other daily measures. These analyses are referred to as totally unconditional (J. D. Singer & Willett, 2003) because daily negative emotion was not modeled as a function of other day- or person-level variables. The basic day-level (within-person or Level 1) model is as follows: NEG ij ⫽ ␤0j ⫹ rij. In this model, ␤0j is a random coefficient representing the mean of daily negative emotion (NEG) for person j (across the i days for which each person provided data), rij represents the error associated with each measure of negative emotion, and the variance of rij constitutes the day-level residual (or error) variance. The basic person-level (between-person or Level 2) model is as follows: ␤ 0j ⫽ ␥00 ⫹ u0j. In this model, ␥00 represents the grand mean of the person-level means (␤0js) from the day-level model, u0j represents the error of ␤0j, and the variance of u0j constitutes the person-level residual variance. We first examined the unconditional model. Following recommendations by Raudenbush and Bryk (2002), all day-level variables were centered on individuals’ means, and all person-level variables were centered on sample means. This analysis estimated the mean level of daily negative emotion to be 1.35. The estimated within-person variance of daily negative emotion (the variance of rij) was 0.58, and the estimated between-person variance (the variance of u0j) was 1.12. The estimated within-person reliability (defined as the ratio of true to total variance) of daily negative emotion was .97 (for a discussion, see Bryk & Raudenbush, 1992, pp. 43– 44). These data thus indicated that the daily ratings of negative emotion were reliable and that there was sufficient variability at the day level to allow for the possibility of modeling

within-person relationships. The reliability estimates for daily positive emotion and daily stress were examined with a similar set of procedures. These analyses indicated that the coefficients for daily positive emotion (.95) and daily stress (.84) were also reliable.

Hypothesis 1: Positive Emotions Moderate the Effects of Stress To test the hypothesis that daily positive emotion moderates the effects of stress, the following day-level model was analyzed: NEG ij ⫽ ␤0j ⫹ ␤1j (Stress) ⫾ ␤2j (POS) ⫹ ␤3j (Stress ⫻ POS) ⫹ rij. In this model, ␤0j is a random coefficient representing the intercept of daily negative emotion (NEG) for person j (across the i days for which each person provided data); ␤1j (Stress) is a random coefficient, a slope, representing the day-level (within-person) relationship between stress and negative emotion for person j; ␤2j (POS) represents the relationship between positive emotion and negative emotion; ␤3j (Stress ⫻ POS) is the concurrent interaction between stress and positive emotion; and ri represents error.3 To examine whether day-level relationships were significantly different from 0 across the individuals in the study, the following person-level model was examined: ␤ 0j ⫽ ␥00 ⫹ u0j. ␤ 1j ⫽ ␥10 ⫹ u1j. ␤ 2j ⫽ ␥20 ⫹ u2j. ␤ 3j ⫽ ␥30 ⫹ u3j. In this model, the significance of ␥10 indicated if, on average, the within-person relationship between stress and negative emotion differed from zero; the significance of ␥20 indicated if, on average, the within-person relationship between positive emotion and negative emotion differed from zero; and the significance of ␥30 indicated if, on average, the within-person interaction between stress and positive emotion differed from zero. The results of these analyses are summarized in Table 1. 3 Because the associations between variables of interest may reflect the influence of linear trends, we included day of study as a control variable in all analyses.

PSYCHOLOGICAL RESILIENCE AND ADAPTATION

735

Across all participants, daily negative emotion scores tended to be higher on days when stress was higher, ␥10 ⫽ .381, t(925) ⫽ 6.21, p ⬍ .01. This within-person coefficient is functionally equivalent to an unstandardized regression coefficient and can be interpreted as such. Thus, for every unit increase in daily stress, mean daily negative emotion increased .38 units. The strength of this relationship was examined by comparing random parameter estimates, and strength was operationalized as the between-person variance in daily negative emotion accounted for by stress (for a discussion, see Bryk & Raudenbush, 1992, p. 65). Examination of the random parameter estimates indicated that inclusion of daily stress resulted in an 18% reduction of within-person variance in negative emotion. This corresponds to a correlation of .42 (the square root of .18) between daily stress and negative emotion.4 In support of Hypothesis 1, higher levels of positive emotion interacted with stress to weaken its influence on negative emotion, ␥30 ⫽ ⫺.338, t(925) ⫽ ⫺5.43, p ⬍ .01. For every unit increase in daily positive emotion, the stress–negative emotion slope decreased .34 units, a finding that is in line with predictions from the DMA (Davis et al., 2004; Zautra et al., 2001).

emotion mediates the relationship between stress and next day’s negative emotion. Alternatively, if daily positive emotion is added to the model and the lagged coefficient for daily stress remains significant, it can be concluded that (on average) some part of the covariation between stress and next day’s negative emotion is independent of the covariation between daily stress and positive emotion. Our analyses revealed that when positive emotion was included in the analysis of emotional recovery, the relationship between stress and next day’s negative emotion was reduced to nonsignificance (.08), whereas it was significant in an analysis without positive emotion (.31), suggesting that positive emotion mediates the relationship between stress and next day negative emotion. To the extent that such results can be used as a basis for making inferences about directionality of effects, it would appear that changes in emotional recovery from stress are due to changes in positive emotion. More specifically, part of the impact that stress may have on negative emotional recovery may be due to decreases in positive emotion brought about by stress. The presence of positive emotion, in contrast, functions to speed recovery from stress (Fredrickson et al., 2003; Tugade & Fredrickson, 2004).

Hypothesis 2: Positive Emotions Mediate the Effects of Stress Recovery

Hypothesis 3: Trait Resilience, Positive Emotions, and Stress Resistance

Our second hypothesis stated that positive emotions would mediate the effects of stress recovery. To analyze mediated relationships, lagged associations between daily stress and emotion were examined. These analyses require that data be provided on consecutive days. Of the total 1,043 days recorded in the study, 935 had data recorded for the days immediately preceding them and were included in the analyses. To rule out the possibility that any lagged effect of stress on negative emotion might be an artifact of the initial level of negative emotion, baseline negative emotion was included in the model as a control variable. In such a model, the dependent variable can be interpreted as the residual change in negative emotion scores from day t to day t ⫹1 (Kessler & Greenberg, 1981).5 The analysis model for changes in daily negative emotion for each individual can be expressed as follows: ⌬NEG t ⫹ 1 ⫽ ␤0j ⫹ ␤1j (NEGt) ⫹ ␤2j (Stresst) ⫾ ␤3j (POSt) ⫹ rt ⫹ 1, where ⌬NEGt⫹1 is the change in negative emotion scores between day t and day t ⫹ 1; ␤0j is a random regression intercept for person j; ␤1j is a random coefficient representing an individual’s level of negative emotion on day t (with the grand mean across all person– days subtracted); ␤2j ⫺ ␤3j represent the within-person associations of stress and positive emotion on next day’s negative emotion; and rt⫹1 is a residual component of change in negative emotion. To test the hypothesis that positive emotions mediate stress recovery, we used a product of coefficients test recently described by MacKinnon, Lockwood, Hoffman, West, and Sheets (2002). This test assesses the indirect effect of a mediating variable as the product of two regression coefficients, one linking the explanatory variable and the mediator and the other linking the mediator and the dependent variable. The significance of this cross-product is divided by its standard error and tested for significance with a specialized sampling distribution. If the inclusion of daily positive emotion (␤3j) renders the slope between stress and next day’s negative emotion (␤2j) nonsignificant (when it was significant in an analysis without ␤3j), then it can be concluded that positive

Our third hypothesis was that trait resilience would contribute to greater stress resistance or a weaker association between positive and negative emotions, particularly on days of heightened stress. To determine if day-level relationships between stress and emotion varied as a function of person-level variables (i.e., trait resilience), coefficients from the day-level models described in Hypothesis 1 were analyzed at the person level with the following models: ␤ 0j ⫽ ␥00 ⫹ ␥01(Trait Resilience) ⫹ u0j. ␤ 1j ⫽ ␥10 ⫹ ␥11(Trait Resilience) ⫹ u1j. ␤ 2j ⫽ ␥20 ⫹ ␥21(Trait Resilience) ⫹ u2j. ␤ 3j ⫽ ␥30 ⫹ ␥31(Trait Resilience) ⫹ u3j. In these models, each person’s day-level slopes are predicted by an intercept, trait resilience, and a random error component.6 For example, ␥10 can be interpreted as the predicted value of the stress–negative emotion association at average levels of trait resilience; ␥11 can be interpreted as the partial relationship between trait resilience and the stress–negative emotion relationship. The 4 Although some authors have suggested that calculations of estimated effect sizes in multilevel data structures should be viewed with caution (e.g., Kreft & De Leeuw, 1998, pp. 115–119), we have presented them to provide some indication of the strength of the relationship between daily negative emotion and other daily measures with strength operationalized in terms of shared variance. 5 We note that although lagged and cross-lagged correlations provide some indication of the lead–lag relationship between two constructs, they are by no means a tool for making causal inferences (Rogosa, 1979). 6 Throughout this investigation, age was assessed in the same multilevel models as trait resilience. Across studies, we did not find any significant variation across people in the size of the primary slope coefficients as a function of age.

736

ONG, BERGEMAN, BISCONTI, AND WALLACE

Figure 1. Study 1a: Concurrent relationship between daily positive and negative emotion as a function of trait resilience, one standard deviation above the mean in daily stress. High and low resilience were defined as one standard deviation from the mean.

analyses found that trait resilience moderated the relationship between daily stress and negative emotion, ␥11 ⫽ ⫺.293, t(925) ⫽ ⫺4.68, p ⬍ .01. Thus, every unit increase in trait resilience was associated with a .29 unit decrease in the stress–negative emotion slope. In support of Hypothesis 3, the individual slopes relating positive emotion to negative emotion on days of above average stress were also predictable from trait resilience, ␥31 ⫽ ⫺.235, t(925) ⫽ ⫺2.83, p ⬍ .05. To examine the form of this interaction, we used Aiken and West’s (1991) procedures to generate separate positive and negative emotion regression lines for individuals high (one standard deviation above the mean) and low (one standard deviation below the mean) on trait resilience. For comparison purposes, we present two-panel figures describing the interaction between trait resilience and positive emotion on high- and lowstress days. As shown in Figure 1, individuals low in trait resilience showed an inverse relationship between daily positive and negative emotion. A test of planned contrast (see Bryk & Raudenbush, 1992, pp. 48 –56) revealed that this relationship differed significantly across high-stress (–.22) and low-stress (–.09) days, ␹2(1, N ⫽ 27) ⫽ 8.12, p ⬍ .01. In comparison, the relationship between daily positive and negative emotions was negligible for high-resilient individuals and did not differ significantly across high-stress (⫺.06) and low-stress (⫺.02) days, ␹2(1, N ⫽ 27) ⫽ 1.43, p ⬎ .05 (cf. Figure 2). These findings thus provide further support for the DMA (Zautra et al., 2001) by identifying an important individual difference variable (i.e., trait resilience) that underlies the capacity for positive emotional engagement in the context of stress.

Hypothesis 4: Trait Resilience, Positive Emotions, and Stress Recovery Our final hypothesis stated that positive emotions would mediate the effects of trait resilience on stress recovery. In the context of our person- and day-level models, this hypothesis implies a process of mediated moderation (Muller, Judd, & Yzerbyt, 2005), whereby the magnitude of stress recovery is moderated by trait resilience, and daily positive emotions are responsible for this moderating effect. To test for mediated moderation, lagged coefficients from the day-level models described in Hypothesis 2 were analyzed as a function of trait resilience. These analyses found that

the effect of stress on next day’s negative emotion was moderated by trait resilience, ␥21 ⫽ ⫺.243, t(925) ⫽ ⫺3.46, p ⬍ .01. Thus, every unit increase in trait resilience was associated with a .24 unit decrease in the lagged stress–negative emotion slope. Consistent with Hypothesis 4, our analyses also revealed that when positive emotion was included, the moderation of the residual direct effect of trait resilience was reduced to nonsignificance (⫺.08), suggesting that positive emotion mediates the moderating relationship of trait resilience and stress on next day’s negative emotion. These findings thus strengthen the prediction that positive emotions may afford daily protective benefits by contributing to the ability of high-resilient individuals to recover more effectively from stressful experiences (Fredrickson et al., 2003; Tugade & Fredrickson, 2004; Tugade et al., 2004).7

Study 1b In Study 1a, we presented evidence that trait-resilient individuals have a tendency to (a) experience positive emotions even amid stressful events and (b) draw on such experiences to resourcefully rebound from daily negative emotional encounters. However, given that trait resilience measures may be negatively correlated with neuroticism (Maddi et al., 2002), any observed associations with daily stress and emotion may be due to this shared neuroticism component rather than any actual adaptive benefits of trait resilience. Thus, it would be useful to determine the extent to which the correlations between trait resilience and daily stress and 7

It is also possible that the mediating effect of daily positive emotions varies as a function of the overall moderating influence trait resilience. This would imply a process of moderation mediation (Muller et al., 2005). Although the various ways in which moderated mediation can occur in the context of multilevel data is beyond the scope of this article (for a discussion, see Bauer, Preacher, & Gil, 2006; Kenny et al., 2003), we note that in none of the models we evaluated was there evidence that the mediating process (i.e., daily positive emotions) was different for individuals who differed in trait resilience. Rather, daily positive emotion appears to be responsible for the overall moderating effect of trait resilience in the current research, and when this process is controlled, the residual moderation of trait resilience is markedly reduced.

PSYCHOLOGICAL RESILIENCE AND ADAPTATION

737

Figure 2. Study 1a: Concurrent relationship between daily positive and negative emotion as a function of trait resilience, one standard deviation below the mean in daily stress. High and low resilience were defined as one standard deviation from the mean.

emotion exist independently of their mutual associations with neuroticism. It would also be useful to determine whether the findings observed in Study 1a could be replicated on an independent sample with different measures of trait resilience and emotion. The goal of Study 1b, therefore, was to provide a conceptual replication of Study 1a while using different measures of trait resilience and daily emotion and controlling for the effects of neuroticism.

Method Participants An independent sample of older adults was randomly selected from the Notre Dame Longitudinal Study of Aging. Fifty participants were contacted and invited to participate in a study of daily stress and emotion. Forty participants, ages 60 – 85 years (M ⫽ 75.5, SD ⫽ 6.28), agreed to take part in the 30-day study. Half of the sample were women and half were men, and all were married at the time of the study. Participants were predominantly European American (97.5%, African American, 2.5%), and half were educated through high school. Income was normally distributed with 2.9% reporting family income less than $14,999, 28.6% between $15,000 and $24,999, 48.6% between $25,000 and $40,000, and 22.9% reporting income greater than $40,000. There were no significant differences in age, gender, income, or educational status for those who did not complete the study. Participants were not compensated for their participation in this study.

Procedure The procedure of Study 1b resembled that of Study 1a. After completing a broad range of mental health measures that included trait measures of psychological resilience and neuroticism, participants then took part in a daily study of stress and emotions. Participants received a diary containing a packet of daily response sheets. Each response sheet contained 24 emotion items and a single item on the most stressful event of the day. Participants were given a month’s supply of diary response sheets and were instructed to respond to the daily items in the evening and return the completed diaries at the end of the 30-day period. The total number of days participants were in the study ranged from 26 to 30 (M ⫽ 28.4, SD ⫽ 2.6). The total number of days in the study for all participants was 1,200 (40 participants ⫻ 30 days). The total number of days of data the participants provided was 1,155 (96% complete).

Measures Psychological resilience. Psychological resilience was assessed with a modified version of the Dispositional Resilience Scale (Bartone, Ursano, Wright, & Ingraham, 1989). The Dispositional Resilience Scale is composed of 45 items, with 15 items each assessing Commitment (e.g., “Most days life is interesting and exciting for me”), Control (e.g., “Planning ahead can help me avoid most future problems”), and Challenge (e.g., “Changes in routine are interesting to me”) aspects of psychological resilience. A 4-point Likert scale, ranging from 0 (not at all true) to 3 (completely true), was used. Reliability data indicated alphas of .72, .68, .59, and .86 for the Commitment, Control, and Challenge subscales and for the overall psychological resilience measure, respectively. Neuroticism. Neuroticism was assessed with a nine-item short form of the Eysenck Personality Inventory (Eysenck & Eysenck, 1975). Sample items include “I am often anxious” and “I am extra sensitive sometimes.” The scale score is based on the sum of yes–no responses to 9 items. Cronbach’s alpha for this sample was .71. Positive and negative emotions. Daily positive and negative emotions were measured with subscales from the Mental Health Inventory (MHI; Veit & Ware, 1983). Participants were assessed each day for 30 days on positive emotional states as well as symptom-specific indicators of anxiety and depression. In addition to being a widely used mental health assessment inventory, the MHI is sensitive to intraindividual change (see McHorney & Ware, 1995; Ware, Gandek, & Group, 1994). In the current study, participants were asked to indicate on a 4-point scale, ranging from 1 (not at all true) to 4 (completely true), the extent to which they had experienced a range of positive emotions and depression or anxiety symptoms on a daily basis. Positive emotions were measured with the 11-item subscale of the MHI. Example positive emotion items include “Today, I felt cheerful, lighthearted,” “Today, I felt calm and peaceful,” and “Today, I was a happy person.” Anxiety and depressive symptoms were assessed with the nineitem anxiety and four-item depression subscales of the MHI. Example items assessing anxiety are “Today, I was a very nervous person,” “Today, I was anxious and worried,” and “I had difficulty trying to calm down.” Example items measuring depression are “Today, I felt downhearted and blue,” “Today, I felt depressed,” and “Today, I had low or very low spirits.” A total daily negative emotion score was calculated for each individual by summing items from the Anxiety and Depression subscales, respectively. Stress. As in Study 1a, in addition to reporting on their daily emotions, participants completed a single item on the most stressful event of the day

ONG, BERGEMAN, BISCONTI, AND WALLACE

738

and then rated their perceptions of how stressful the event was on a 5-point scale, ranging from 1 (not very stressful) to 5 (very stressful).

Results and Discussion Descriptive Statistics Similar to results reported in Study 1a, daily positive emotion scores were higher and more variable (M ⫽ 2.81, SD ⫽ 0.98) than negative emotion scores (M ⫽ 1.25, SD ⫽ 0.83). Overall, daily positive emotion scores were inversely correlated with daily stress and negative emotions (r ⫽ ⫺.32, p ⬍ .05, and r ⫽ ⫺.27, p ⬍ .05, respectively). Trait resilience, moreover, was significantly correlated with daily positive emotion and stress (r ⫽ .37, p ⬍ .05, and r ⫽ ⫺.31, p ⬍ .05, respectively), but unrelated to daily negative emotion (r ⫽ ⫺.09, ns).

Hypothesis 1: Positive Emotions Moderate the Effects of Stress As in Study 1a, our first aim was to investigate the extent to which daily positive emotions moderate the effects of stress. To investigate this, we used MRCM because of the multilevel structure of the data. In Study 1b, there were up to 30 observations nested within each of 40 participants. The equations predicting daily concurrent and lagged negative emotion in Study 1b were identical to those of Study 1a. As in previous analyses, daily scores were group–mean centered to eliminate the influence of parameter estimates of individual differences. The final parameter estimates from our MRCM analyses are summarized in Table 2. As in Study 1a, higher levels of daily stress were concurrently associated with higher levels of daily negative emotion, ␥10 ⫽ .379, t(984) ⫽ 5.84, p ⬍ .01. It is important that our analyses also revealed significant reductions in the magnitude of the stress–negative emotion correlation on days in which greater positive emotion was present, ␥30 ⫽ ⫺.373, t(984) ⫽ ⫺5.61, p ⬍ .01.

Hypothesis 2: Positive Emotions Mediate the Effects of Stress Recovery Our second aim was to replicate findings observed in Study 1a, in which daily positive emotions were found to mediate the effects of stress recovery. We began by examining the time-lag dependency between daily stress and next day negative emotion scores. These analyses required that data be provided on consecutive days. Of the total 1,155 days recorded in the study, 992 had data recorded for the days immediately preceding them and were included in the analyses. Consistent with findings from Study 1a, the results of our lagged analyses revealed that stress on one day (day t) uniquely predicted negative emotion on the next day (day t ⫹ 1), above and beyond the relationship between negative emotion on day t and day t ⫹ 1. In support of Hypothesis 2, our analyses revealed that when positive emotion was included in the analysis of emotional recovery, the relationship between stress and next day’s negative emotion was reduced to nonsignificance (.03), whereas it was significant in an analysis without positive emotion (.27), suggesting that positive emotion mediates the relationship between stress and next day negative emotional recovery.

Hypothesis 3: Trait Resilience, Positive Emotions, and Stress Resistance As in Study 1a, we examined the extent to which trait resilience contributes to greater stress resistance by weakening the association between positive and negative emotions during times of stress. In addition to examining the daily emotional processes associated with trait resilience, however, we also evaluated the extent of reduction in these associations when neuroticism was statistically held constant. Table 2 shows the relationships between trait resilience and stress and emotion, with and without controlling for neuroticism. Although the coefficients for trait resilience, stress, and positive emotions and their interactions were smaller than they were in an analysis without neuroticism, Table 2 shows that all coefficients maintained their valence and remained statistically

Table 2 Parameter Estimates for Daily Negative Emotions, With and Without Controlling for Neuroticism Variable Without neuroticism Stress Positive emotions Stress ⫻ Positive Emotions Stress ⫻ Trait Resilience Positive Emotions ⫻ Trait Resilience Stress ⫻ Positive Emotions ⫻ Trait Resilience With neuroticism Stress Positive emotions Stress ⫻ Positive Emotions Stress ⫻ Trait Resilience Positive Emotions ⫻ Trait Resilience Stress ⫻ Positive Emotions ⫻ Trait Resilience

B

t

df

p⬍

0.379 ⫺0.162 ⫺0.373 ⫺0.325 ⫺0.124 ⫺0.257

5.84 ⬍1 ⫺5.61 ⫺4.83 ⬍1 ⫺2.97

984 984 984 984 984 984

.01 ns .01 .01 ns .05

0.269 ⫺0.163 ⫺0.338 ⫺0.315 ⫺0.114 ⫺0.246

3.27 ⬍1 ⫺5.24 ⫺4.73 ⬍1 ⫺2.81

984 984 984 984 984 984

.05 ns .01 .01 ns .05

Note. All day-level predictors were group–mean centered, and all person-level predictors were centered on sample means.

PSYCHOLOGICAL RESILIENCE AND ADAPTATION

739

Figure 3. Study 1b: Concurrent relationship between daily positive and negative emotion as a function of trait resilience, one standard deviation above the mean in daily stress. High and low resilience were defined as one standard deviation from the mean.

significant after neuroticism was controlled. Thus, trait resilience still accounted for variance in daily stress resistance, ␥30 ⫽ ⫺.246, t(984) ⫽ ⫺2.81, p ⬍ .05. As illustrated in Figure 3, individuals low in trait resilience showed an inverse relationship between positive and negative emotion, and this relationship differed significantly across high-stress (⫺.24) and low-stress (⫺.09) days, ␹2(1, N ⫽ 40) ⫽ 7.38, p ⬍ .01. Consistent with findings from Study 1a, the relationship between positive and negative emotions for high-resilient individuals was nonsignificant and did not differ across high-stress (⫺.07) or low-stress (⫺.04) days, ␹2(1, N ⫽ 40) ⫽ 1.17, p ⬎ .05 (cf. Figure 4).

Hypothesis 4: Trait Resilience, Positive Emotions, and Stress Recovery Our final aim was to replicate the mediational effect of daily positive emotions observed in Study 1a while also controlling for neuroticism. As in Study 1a, our MRCM analyses involved tests of mediated moderation (Muller et al., 2005). As was expected, these analyses found that the lagged effect of stress on negative emotion was moderated by trait resilience, ␥21 ⫽ ⫺.285, t(984) ⫽ ⫺3.47,

p ⬍ .05. That this effect was evident even after controlling for neuroticism is noteworthy. In addition, when positive emotion was included, the moderation of the residual direct effect of trait resilience was reduced to nonsignificance (⫺.05). These findings provide support for Hypothesis 4: The experience of daily positive emotions appears to aid resilient individuals in the ability to bounce back from daily stress. Overall, findings from Study 1b provide additional empirical support for the prediction that the functional benefits of positive emotions are (a) strongest in the context of stressful life events (Zautra, Johnson, & Davis, 2005; Zautra et al., 2001) and (b) contoured by broad individual differences in psychological resilience (Ong & Bergeman, 2004a; Tugade & Fredrickson, 2004). The results are thus concordant with findings from Study 1a, despite variations in the measures used to assess trait resilience and daily positive and negative emotion, respectively. Although findings from Studies 1a and 1b help to establish the adaptational significance of trait resilience and positive emotions, a number of methodological features limit the generality of these findings. Foremost, Studies 1a and 1b were limited to relatively minor

Figure 4. Study 1b: Concurrent relationship between daily positive and negative emotion as a function of trait resilience, one standard deviation below the mean in daily stress. High and low resilience were defined as one standard deviation from the mean.

ONG, BERGEMAN, BISCONTI, AND WALLACE

740

stressors, and as such, examining the generality of these findings to major life events is also necessary. Our primary goal in Study 2, therefore, was to further explore how psychological resilience and positive emotions shape the day-to-day regulation of ongoing negative emotions following the potentially traumatic life event of losing a spouse.

Study 2 Like the child and adolescent literature on resilience (Luthar et al., 2000), perspectives on resilience in adulthood and later life have emphasized the need to assess positive outcomes in response to major life challenges (Ryff et al., 1998). Few life events affect adults more than the death of a spouse or life partner (Bonanno & Kaltman, 1999; Stroebe & Stroebe, 1983). Despite the distress and grief that the death of a loved one brings, however, there is considerable variability in individuals’ responses to interpersonal loss; some individuals experience acute and enduring psychological distress, whereas others do not (Wortman & Silver, 1989, 1990). Accumulating evidence, in fact, indicates that a substantial minority of bereaved individuals experience and express positive emotions far more often than might have been previously anticipated (for a discussion, see Bonanno, 2004; Folkman, 2001). Thus, additional empirical work that addresses the adaptive processes through which individuals adjust to and recover from major life challenges is clearly needed. In Study 2, we explored how profiles of daily emotional responses to stress intersect with the significant adaptive pressures associated with conjugal loss. We tested the prediction that among high-resilient widows, the covariation in positive and negative emotions during times of stress differs in strength, duration, and significance. We predicted that (a) positive emotions would attenuate reactivity to and recovery from daily stress (Hypotheses 1 and 2) and (b) these within-person relationships would vary systematically as a function of psychological resilience (Hypothesis 3). Finally, on the basis of findings from Studies 1a and 1b, we predicted that the experience of daily positive emotions would mediate the effect of trait resilience on emotional recovery from stress (Hypothesis 4).

Method Participants Data for this study are from the Notre Dame Study of Adjustment to Widowhood, a longitudinal study of the effects of bereavement on the mental and physical health of older widows. Further details of the study are described elsewhere (Bisconti, Bergeman, & Boker, 2004; Ong et al., 2004). Briefly, the sample comprised 60 widows (age range ⫽ 57– 83 years), who were randomly assigned to a target or control group. All participants took part in a pre- and postinterview and completed self-report questionnaires at the initial and postinterviews as well as at 8, 12, 16, 20, 24, 36, and 48 months (60-month data are currently being collected). In addition to the interview and questionnaire data, the target widowed group (n ⫽ 34) was asked to keep a daily record of their stress and emotions. These daily assessments lasted for 98 days. The 34 widows who took part in the daily diary study ranged in age from 61 to 83 years of age (M ⫽ 71.94, SD ⫽ 6.11). The majority of participants had at least a high school education (97.06%). Additionally, 55.87% of the women had received some post-high school education or training. Income levels were difficult to assess immediately following the death of their spouse. However, during the follow-up interview, which was approximately 4 months postloss,

16.67% of the participants reported an annual income between $7,500 and $15,000. In addition, 46.67% of the participants reported a yearly income between $15,000 and $25,000, 13.33% reported an annual income between $25,000 and $40,000, and 23.33% reported making over $40,000 per year. The length of marriage ranged between 14 to 63 years (M ⫽ 46.97, SD ⫽ 12.26), and for 79.41% of the widows it was their first marriage. In addition, 61.76% of the widows expected the death of their husband to occur, and 91.18% of them were living alone following conjugal loss. Widows received $30 in return for their participation.

Procedure Participants received a battery of self-report questionnaires approximately 1 month postloss (M ⫽ 28 days, SD ⫽ 6). Participants then took part in a daily diary study of stress and emotion. Each daily packet was dated and mailed to the widows in bimonthly intervals. If a participant missed a day, she was instructed to leave that day’s response sheet blank. The first set was given to the participants at the initial interview and included a self-addressed, postage-paid envelope to return surveys. Participants were instructed to complete response sheets in the evening and return diaries by mail every 2 weeks. To remind participants to mail the packet of daily assessments, regular weekly phone calls were made. These conversations were also a way to keep in touch with the widows over the 3-month project. The total number of days participants were in the study ranged from 14 to 98 (M ⫽ 75.94, SD ⫽ 26.87). The total number of days in the study for all participants was 3,332 (34 participants ⫻ 98 days). The total number of days of data the participants provided was 2,590 (78% complete).

Measures Both the person-level (i.e., trait resilience) and day-level (i.e., stress, positive and negative emotions) measures were identical to Study 2. Reliability data indicated alphas of .77 and .74 for the trait resilience and neuroticism measures, respectively.

Results and Discussion Descriptive Statistics Across the 98-day assessment period, daily negative emotion scores were higher and more variable (M ⫽ 3.15, SD ⫽ 0.94) than positive emotion scores (M ⫽ 2.31, SD ⫽ 0.75). Similar to Studies 1a and 1b, the two positive indicators of well-being (trait resilience and daily positive emotions) correlated significantly with each other (r ⫽ .41, p ⬍ .01), as did the two negative indicators (daily negative emotions and daily stress; r ⫽ .45, p ⬍ .01).8 In addition, greater daily positive emotion was associated with less daily negative emotion and stress (r ⫽ ⫺.29, p ⬍ .05, and r ⫽ ⫺.35, p ⬍ .05, respectively). Trait resilience, in addition, was negatively correlated with stress and daily negative emotion (r ⫽ ⫺.38, p ⬍ .05, and r ⫽ ⫺.34, p ⬍ .05, respectively). 8

In addition to negative emotions, we also assessed daily grief responses using the Grief Resolution Index (GRI; Remondet & Hansson, 1987). Sample items from the GRI include “Accepted the death of my husband” and “Was able to think through what my husband’s death meant to me.” Although the correlates of the GRI mirrored the correlates obtained on the daily negative emotion measure, we chose not to include the GRI in our MRCM analyses because nearly half (48%) of the diary entries for this measure were missing across the study period.

PSYCHOLOGICAL RESILIENCE AND ADAPTATION

Hypothesis 1: Positive Emotions Moderate the Effects of Stress We predicted that elevations in positive emotions on days characterized by high stress would be particularly important in the regulation of negative emotional states (Zautra et al., 2001). As in Studies 1a and 1b, our hypotheses were tested with MRCM because the data have a hierarchical structure. In this study, there were up to 98 observations nested within each of 34 participants. The final parameter estimates from our MRCM analyses are summarized in Table 3. As with Studies 1a and 1b, stress was related to increased negative emotions, ␥10 ⫽ .436, t(2388) ⫽ 8.52, p ⬍ .01. Notably, positive emotions interacted with stress to interrupt its influence on negative emotions, ␥30 ⫽ ⫺.411, t(2388) ⫽ ⫺7.64, p ⬍ .01.

Hypothesis 2: Positive Emotions Mediate Stress Recovery Our hypothesis concerning the mediating role of positive emotions was also tested with MRCM. Of the total 2,590 days recorded in the study, 2,251 had data recorded for the days immediately preceding them and were included in the analyses. Similar to findings from Studies 1a and 1b, there was a strong lagged relationship between stress and negative emotion. In this study, stress continued to influence negative emotion as long as two lags later, ␥20 ⫽ .257, t(2388) ⫽ 4.36, p ⬍ .05. As was expected (Hypothesis 2), when daily positive emotion was added to the model predicting next day negative emotion, the lagged coefficient for stress was reduced to nonsignificance (.06), whereas it was significant in an analysis without positive emotion (.32), suggesting the relationship between stress and next day negative emotion is mediated by daily positive emotions.

Hypothesis 3: Trait Resilience, Positive Emotions, and Stress Resistance We next examined the extent to which trait resilience contributes to greater stress resistance by reducing the correlation be-

741

tween positive and negative emotions on days of elevated stress. As in Study 1b, we also examined the extent of reduction in these associations when neuroticism was held constant. Table 3 shows the relationships between trait resilience and stress and emotion, with and without controlling for neuroticism. As with Study 1b, after controlling for neuroticism, trait resilience still accounted for variance in daily stress resistance, ␥30 ⫽ ⫺.251, t(2388) ⫽ ⫺4.28, p ⬍ .05. As depicted in Figure 5, individuals low in trait resilience showed an inverse relationship between positive and negative emotion, and this relationship differed significantly across highstress (⫺.32) and low-stress (⫺.15) days, ␹2(1, N ⫽ 34) ⫽ 8.96, p ⬍ .01. In comparison, the positive and negative emotion relationship was nonsignificant for high-resilient individuals and did not differ across high-stress (⫺.09) or low-stress (⫺.06) days, ␹2(1, N ⫽ 34) ⫽ 1.59, p ⬎ .05 (cf. Figure 6).

Hypothesis 4: Trait Resilience, Positive Emotions, and Stress Recovery Finally, we examined whether select mediating findings observed in Studies 1a and 1b could also be observed in our study of bereaved widows. We argued that if the experience of positive emotions helps resilient individuals recover from everyday stressful events, then such experiences should have adaptive functions for those undergoing real-life stressors as well. On the basis of findings from Studies 1a and 1b, we predicted that among widows high in trait resilience, daily positive emotions may afford protective benefits by contributing to effective emotional recovery from stress. As in Studies 1a and 1b, our hypotheses were tested with multilevel models for mediated moderation (cf. Kenny, Korchmaros, & Bolger, 2003; Muller et al., 2005). After controlling for neuroticism, these analyses found that the lagged effect of stress on negative emotion was moderated by trait resilience, ␥21 ⫽ ⫺.269, t(2388) ⫽ ⫺4.38, p ⬍ .05. In support of Hypothesis 4, our analyses also revealed that when positive emotion was included, the moderation of the residual direct effect of trait resilience was reduced to nonsignificance (⫺.11).

Table 3 Parameter Estimates for Daily Negative Emotions, With and Without Controlling for Neuroticism Variable Without neuroticism Stress Positive Emotions Stress ⫻ Positive Emotions Stress ⫻ Trait Resilience Positive Emotions ⫻ Trait Resilience Stress ⫻ Positive Emotions ⫻ Trait Resilience With neuroticism Stress Positive emotions Stress ⫻ Positive Emotions Stress ⫻ Trait Resilience Positive Emotions ⫻ Trait Resilience Stress ⫻ Positive Emotions ⫻ Trait Resilience

B

t

df

p⬍

0.436 ⫺0.388 ⫺0.411 ⫺0.353 ⫺0.096 ⫺0.272

8.52 ⫺6.44 ⫺7.64 ⫺6.11 ⬍1 ⫺4.41

2388 2388 2388 2388 2388 2388

.01 .01 .01 .01 ns .05

0.283 ⫺0.327 ⫺0.366 ⫺0.324 ⫺0.071 ⫺0.251

4.63 ⫺5.85 ⫺6.27 ⫺5.62 ⬍1 ⫺4.28

2388 2388 2388 2388 2388 2388

.05 .01 .01 .01 ns .05

Note. All day-level predictors were group–mean centered, and all person-level predictors were centered on sample means.

ONG, BERGEMAN, BISCONTI, AND WALLACE

742

Figure 5. Study 2: Concurrent relationship between daily positive and negative emotion as a function of trait resilience, one standard deviation above the mean in daily stress. High and low resilience were defined as one standard deviation from the mean.

General Discussion We began this article by underscoring how little we know about the dynamic interplay between positive and negative emotional states in later life, and we sought an explanation of what are the individual and contextual factors that account for successful adaptation in the face of adversity. The results of the present set of studies converge on five broad conclusions: (a) The adaptive benefits of positive emotions are greatest when individuals are under stress; (b) positive emotions are more common among high-resilient individuals; (c) those low in psychological resilience tend to have difficulty regulating negative emotions and exhibit heightened reactivity to daily stressful life events; (d) when present, positive emotions are especially important for lowresilient individuals, particularly in the context of stress; and (e) over time, positive emotions serve to assist high-resilient individuals in their ability to effectively rebound from adversity. On the whole, these findings add to the growing number of longitudinal studies suggesting that positive emotional processes are a key component of what it means to be resilient (Bonanno, 2004; Fredrickson et al., 2003; Zautra, Johnson, & Davis, 2005).

In a recent review of the loss and trauma literature, Bonanno (2004) provocatively concluded that resilience is a “common phenomenon, distinct from the process of recovery, and can be reached by a variety of different pathways” (p. 26). In general, although our results provide corroborative support for each of these conclusions, they also suggest that our understanding of how these processes operate at the day-to-day level in later adulthood is at best elementary. First, across studies, our findings indicate that the concept of resilience has relevance not only to those undergoing significant life challenge but also to those experiencing daily stressors that spontaneously arise and subside in naturally occurring contexts. As such, the findings add to the growing body of empirical evidence indicating that human resilience is a normative process that operates across the life span (Bonanno, 2005; Masten, 2001; Staudinger et al., 1995). Second, the process of stress resistance was observed to be distinct from the process of recovery. In the present set of studies, the maintenance of emotional differentiation under stress was judged as a marker of stress resistance. It is important that daily stress resistance was seen to operate entirely in situ. That is, the systematic covariation between

Figure 6. Study 2: Concurrent relationship between daily positive and negative emotion as a function of trait resilience, one standard deviation below the mean in daily stress. High and low resilience were defined as one standard deviation from the mean.

PSYCHOLOGICAL RESILIENCE AND ADAPTATION

positive and negative emotions was found to reside largely in its concurrent interaction with stress. Individual differences in psychological resilience, moreover, appeared to influence the threshold at which individuals reacted to ongoing daily stressors. Further, our analyses of recovery processes revealed that for less resilient individuals, the unpleasant experience of one daily stressful event tends to follow on the heels of another, thereby ratcheting up subsequent stress levels even higher. Finally, our findings strongly support the assertion that flexible adaptation to adversity can be reached through a variety of protective pathways (Bonanno, 2004, 2005). Two such pathways were evident in the current research: one operating at the level of within-person variation (i.e., daily positive emotion) and the other at the level of betweenperson differences (i.e., psychological resilience). The breadth and convergence of evidence across studies raises important questions regarding the measurement and conceptualization of resilience in later adulthood.

The Adaptive Functions of Positive Emotions More than 2 decades ago, Lazarus, Kanner, and Folkman (1980) suggested that under intensely stressful conditions, positive emotions may provide an important psychological time-out, sustain continued coping efforts, and restore vital resources that have been depleted by stress. Until recently, there has been little empirical support for these ideas. Foundational evidence for the adaptive function of positive emotions is beginning to accrue, however (e.g., Bonanno & Keltner, 1997; Folkman, 1997; Fredrickson et al., 2003; Tugade & Fredrickson, 2004; Zautra, Johnson, & Davis, 2005). Taken together, the present findings add to and strengthen the generality of extant empirical work on positive emotions. In particular, the present investigation extends research by its empirical attention to the real-life challenges and stresses of later life. A primary finding emerging from this research is that a significant proportion of older adults manage to experience positive emotions, even in the midst of overwhelming loss. Despite variation in the types of stressors experienced, however, the results across three independent samples are remarkably consistent: Positive emotions have demonstrably beneficial effects when present during times of stress. Together with related work (e.g., Davis et al., 2004; Zautra et al., 2001), findings from the current research suggest that positive emotions may strengthen stress resistance by providing an important psychological breather when distress becomes particularly intense (Lazarus et al., 1980). Additionally, our findings join with past research (e.g., Fredrickson & Levenson, 1998; Fredrickson et al., 2003) in demonstrating that positive emotions may also protect against slow or prolonged recovery from stress. Fredrickson (1998, 2001) has argued that under stressful conditions, positive emotions may help to build and restore depleted personal resources. Consistent with this prediction, the present findings provide empirical support for the hypothesized building function of positive emotions. Moreover, by situating the study of positive emotions within existing dynamic models of stress and emotion (e.g., Zautra, Affleck, et al., 2005; Zautra et al., 2001), the current research provides an important conceptual and methodological link to previous laboratory studies. Overall, the data establish that positive emotions may function in the service of well-being not only by

743

interrupting the ongoing experience of stress but also by averting delays in adaptation to subsequent stressors.

The Emotional Underpinnings of Psychological Resilience As Mischel (2004) and others have noted, trait data become especially powerful explanatory constructs when they are viewed in connection with dynamic processes that activate and make salient selective individual difference variables. Which traits help some to maintain and regain emotional well-being whereas others languish in feelings of helplessness and hopelessness? Particular attention has recently been paid to psychological traits that generate and sustain positive emotions under stressful conditions (e.g., Ong et al., 2004; Tugade et al., 2004; Zautra et al., 2001). Findings across studies indicate that differences in adaptation to stress may follow from one’s habitual outlook on life; that is, how individuals react to, appraise, and interpret adverse life experiences. Overall, the present findings suggest that individual differences in psychological resilience may constitute an important route to understanding differential resistance to and recovery from daily stress in later adulthood. Throughout this research, psychological resilience accounted for meaningful differences in emotional responses to daily stressors. That these relationships held, even after controlling for variables thought to influence these daily processes (i.e., neuroticism) is noteworthy. Perhaps nowhere was this truer, however, than in our widowhood study. These individuals meet the two criteria on which resilience, as a process, is predicated (Luthar et al., 2000). First, they have been exposed to significant adversity. Second, they have achieved positive outcomes despite these adverse experiences. Over time, what are the key differences that distinguished high-resilient widows from their less well-functioning, vulnerable peers who similarly suffered interpersonal loss? There are relatively few, but those are revealing: High-resilient widows were more likely to experience a range of positive (e.g., cheerful, peaceful, happy) and negative (e.g., anxious, worried, depressed) emotions throughout the bereavement process. A key distinguishing feature of this experience, however, is the capacity to maintain partial separation of positive and negative emotional states while under stress, thereby preserving emotional complexity (cf. Ong & Bergeman, 2004a; Zautra et al., 2001). These findings provide additional empirical footing for the DMA. In previous work, Zautra and colleagues reported related effects for mood clarity, an aspect of emotional intelligence that reflects the capacity for emotional understanding (Salovey, Mayer, Goldman, Turvey, & Palfai, 1995), and cognitive complexity, a characteristic of selfconcept implicated in mood variability (Linville, 1985). Specifically, individuals with greater mood clarity (Zautra et al., 2001) and cognitive complexity (Potter et al., 2000) exhibited greater differentiation of positive and negative emotions. In the current investigation, a similar mechanism was found to underlie psychological resilience, suggesting that one adaptive outgrowth of resilience is an increase in emotional complexity during times of stress. In addition to evidencing greater emotional complexity, highresilient widows also showed greater control over their positive emotional experiences. Our mediational analyses revealed that high-resilient widows were also more likely to selectively mobilize positive emotions to recover and bounce back from daily stress. In the context of conjugal bereavement, these findings suggest that

744

ONG, BERGEMAN, BISCONTI, AND WALLACE

psychological resilience may help bereaved spouses sustain access to daily positive emotional resources, which in turn may lead to accelerated recovery from stress (for related work, see Moskowitz, Folkman, & Acree, 2003). Collectively, these findings link up with prior research (e.g., Fredrickson et al., 2003; Tugade & Fredrickson, 2004) in demonstrating positive emotions’ larger reach as well as its enduring connection to trait resilience.

Implications The present findings have a range of implications. First, if the ability to experience positive emotions in the context of stress is indeed adaptive, then interventions designed to bolster individuals’ capacity for seeing the complexity of emotions inherent in everyday stressful situations may prove to be beneficial. Zautra (2003) cited evidence that mindfulness-based approaches to stress reduction may offer a means of broadening emotional awareness and thus help to sustain positive emotional engagement under stressful conditions. In addition, interventions that facilitate the processing of emotions with greater complexity might also foster adaptive coping and adjustment to chronic stress and illness (Reich et al., 2003). Second, the evidence that positive emotions may be important in helping resilient individuals recover from stress suggests that the experience of positive emotions may be potentially modifiable by intervention. Several studies indicate that coping styles marked by situational meaning (Park & Folkman, 1997) and perspectivetaking (Fredrickson & Joiner, 2002) may facilitate adjustment to acute and persistent stress. Although we did not directly assess coping, our findings are in line with studies that indicate that changing the appraised personal significance of stressful conditions may be one mechanism by which to cultivate positive emotions in the midst of stress (Park & Folkman, 1997; Tugade & Fredrickson, 2004). Together with related research with younger adults (e.g., Fredrickson et al., 2003; Tugade & Fredrickson, 2004; Tugade et al., 2004), the present findings thus provide the basis for underscoring the importance of building positive emotional experiences into the ecology of older adults’ everyday lives. Finally, our findings are complemented by recent studies suggesting that the capacity for complex emotional experience may run in parallel with the progressive development of differentiated but integrated forms of cognition (for a discussion, see LabouvieVief, 2003, 2005).9 Discussing patterns in cognitive–affective differentiation from childhood to adulthood, Labouvie-Vief and Medler (2002) recently argued that age-related changes in emotional complexity are likely to be modified by relatively habitual individual differences in styles of emotion regulation. LabouvieVief and Medler (2002) further proposed that the most adaptive mode of emotion regulation is one that is characterized by high levels of positive emotion and integrative processing of positive and negative emotional information. This view of emotional development resonates with the general finding in the current research that differences in psychological resilience (an enduring self-regulatory capacity) can have an integrative function and promote overall adaptation by allowing individuals (a) to experience emotional complexity amid stressful experiences and (b) to capitalize on their positive emotions to successfully rebound from such experiences.

Caveats and Future Directions A number of limitations of this research warrant comment. In the following, we describe notable caveats, as well as issues that might profitably be considered in future research. First, we operationally defined successful adaptation in the current research as the ability to maintain and regain emotional health in the face of daily stress. The maintenance and recovery of emotional functioning are not the only indicators of successful adaptation, however. Future research should be broadened conceptually to include the role of psychological well-being (Keyes, Shmotkin, & Ryff, 2002), psychological thriving (Carver, 1998), and posttraumatic growth (Tedeschi & Calhoun, 2004) in response to later life challenges. We acknowledge, however, that there are situations in which an exclusive focus on emotional and psychological wellness may not necessarily be appropriate. For certain severe life adversities, such as direct exposure to the brutalities of war, the absence of mental and physical illness may be a more pressing and valid indicator of successful adaptation (Luthar & Cushing, 1999). Yet a third way of conceptualizing health-promoting processes follows from consideration of the conditions in which positive and negative processes may be simultaneously activated (Cacioppo, Gardner, & Berntson, 1999; Schimmack, Oishi, & Diener, 2002). Such states of coactivation have been posited to represent an important adaptive mechanism that confers benefits through its impact on the individual’s ability to directly engage and find meaning in challenge (Folkman et al., 1997; Larsen, Hemenover, Norris, & Cacioppo, 2003). When viewed together, these diverse representations of human resilience lend support to the notion of multifinality (Cicchetti & Rogosch, 1996), which emphasizes the dynamic and coordinated interplay between individual and contextually determined factors. We think this interplay is one of the most intriguing and promising areas for future study. Second, the results of the current research do not speak to the general restriction in the range of plasticity or reserve capacity (M. M. Baltes, Kuhl, & Sowarka, 1992; P. B. Baltes & Kliegl, 1992), or even possible hidden costs that may accompany profiles of adaptive functioning in later adulthood (Staudinger et al., 1993, 1995). Several lines of evidence primarily from studies of childhood psychopathology (e.g., Werner & Smith, 1992; Zucker, Wong, Puttler, & Fitzgerald, 2003) suggest that children labeled as resilient may fare well in certain domains (e.g., academic competence, prosocial behavior), but show noticeable deficits in others (e.g., social and emotional competence). Evidence of ontogenic variability across selective life domains has also been documented among older adults adapting to various age-related losses and transitions (P. B. Baltes, 1997; P. B. Baltes & Baltes, 1990). As 9 We note that our conceptualization of emotional complexity specifically refers to individual differences in the intraindividual covariation between positive and negative emotional states. This conceptualization follows from the nomothetic–idiographic tradition of studying emotional experience (e.g., Feldman, 1995a; Wessman & Ricks, 1966; Zautra et al., 2001). As such, this research needs to be distinguished from related cognitive-developmental work that follows from the tradition of Piaget and Kohlberg (for a review, see Labouvie-Vief, 1994), in which the capacity for emotional complexity is seen as cognitive skill that develops according to the classic principles of the development of cognition in general (Labouvie-Vief & Diehl, 2000; Labouvie-Vief & Medler, 2002).

PSYCHOLOGICAL RESILIENCE AND ADAPTATION

noted by others (e.g., Ryff et al., 1998), however, studies that include direct assessments of both flourishing and challenge in later life are limited. Future research should, therefore, extend these findings by examining the extent to which psychological resilience and positive emotions, when chronically mobilized in times of stress, exact tolls on other areas of functioning, thereby precipitating the inimical effects of allostatic load (McEwen, 1998). A third limitation concerns the methodological drawbacks that were shared by all three studies, including small sample sizes, lack of experimental control over confounding variables, and reliance on self-reports. Several investigators have noted, for instance, that interaction models with small sample sizes, in practice, are prohibitively difficult to test (e.g., Luthar et al., 2000). Our relatively small samples did not, however, obscure the presence of fairly complex cross-level interactions. That such effects were detectable across all three studies and over several temporal lags is noteworthy. It is also possible that other factors that could not be adequately controlled for in the current research were driving the results. For example, it is possible that other types of stressors might show comparatively different effects. Several studies, however, have independently manipulated stress (Zautra et al., 2000) and have found that the subjective experience of stress is reliably associated with diminished emotional experience and recovery. Nonetheless, the combination of randomized designs and intensive day-to-day monitoring of phenomena such as stress and emotion, we believe, presents a rich opportunity to explore the real-world effects of interventions based on a resilience paradigm. Finally, our analyses of daily stress and emotion relied heavily on selfreports from respondents. These measures were completed at the end of the day, and hours could have passed since the occurrence of the daily stressor. It is possible that negative mood could have resulted in a distorted recollection and appraisal of events (Marco & Suls, 1993). Similarly, we evaluated day-to-day stress by asking participants to focus on “the most stressful event of the day” and then to rate their stress level in relation to that event. Although this approach has the intuitive appeal of being less removed from the participant’s real experience, it assumes that the participant’s endof-day emotions are systematically influenced by the day’s most stressful event. It is possible, however, that many of the “most stressful events” participants responded to were not particularly stressful (see Tennen & Affleck, 2002). Future investigations should thus take a multimethod approach (for a discussion, see J. E. Singer & Davidson, 1991) to stress assessment by including not only self-reports from respondents but also physiological outcomes, biochemical assessments, and behavioral measures of stress. Similarly, because the occurrence of any life change requires some type of readjustment (Monroe & McQuaid, 1994), studies that include greater coverage in the range of daily events, both positive and negative, should be a high priority for future research (Zautra, Guarnaccia, & Dohrenwend, 1986). Fourth, a number of variables known to have an effect on the resilience process were not examined in the present research. In particular, we did not assess social support as a possible source of between-person differences affecting either daily stress or emotion. Converging life span evidence suggests that environmental supports routinely foster the development of successful adaptation among both children and adults under stress (cf. Luthar & Zelazo, 2003; Ryff et al., 1998). Our findings, moreover, relate only to

745

short-term stressors and cannot address the influence of psychological resilience on chronic strains of appreciable duration (e.g., caring for an ill spouse). Thus, it will be important for future studies to determine whether other protective resources, such as the social environment, contribute to adaptation to more protracted forms of later life challenge. A fifth limitation concerns our exclusive adoption of a variablefocused approach in the current research. Although this approach provides a sensitive strategy for detecting synergistic nonadditive effects, variable-centered approaches cannot fully capture the subgroup heterogeneity that may be reflected in qualitatively distinct subtype patterns or subpopulations. Person-centered approaches such as latent class analysis and growth mixture modeling, in comparison, allow for the consideration of whether hidden categorical variables (classes) explain the trajectories of individuals over time (Nagin, 1999). Such methods thus permit investigations of the different trajectories of resilience that may correspond to the different clusters of respondents within a sample (Bonanno, Wortman, & Nesse, 2004). Finally, we highlight the application of dynamical systems analysis (Boker & Nesselroade, 2002) as an innovative analytic technique for addressing process research questions within a resilience framework. One fundamental assumption underlying the process of resilience, for example, is that resilient functioning is characterized by quicker return to equilibrium (Curtis & Cicchetti, 2003; Davidson, 2000). Dynamical systems analysis allows one to directly test the extent to which a system of variables self-regulates over time and thereby show fluctuation about an equilibrium (Boker & Nesselroade, 2002). Using a subset of widows from the Notre Dame Study of Adjustment to Widowhood, Bisconti et al. (2004) assessed intraindividual variability in emotion regulation in the 3 months following the death of a spouse using dynamical system analysis. Results indicated that the trajectory of emotion regulation following the loss of a spouse resembles a pendulum with friction; that is, emotional responses during bereavement were consistent with an oscillating trajectory that dampens over time. Although this research was designed to assess the homeostatic mechanisms that are invoked following a major life event, an equally important line of research concerns tracking the processes that support allostasis, the process by which an organism remains stable in the face of challenge (McEwen, 1998). Because dynamical systems techniques can accommodate both person and situational changes, they are particularly well suited to exploring such processes. These methods may provide an important bridge between laboratory studies of the effects of stress exposure and reactivity, on the one hand, and field investigations of individual differences in stress recovery and the efficacy of restorative processes, on the other.

Concluding Remarks Although a great deal of progress has been made in our understanding of successful adaptation across the life span (cf. Luthar et al., 2000; Ryff et al., 1998), the role of short-term adaptation to stress remains vastly understudied, particularly in later adulthood. Although this body of work is still small, the results of the current research have uncovered important findings: Positive emotions in later adulthood fluctuate on a day-to-day basis, and both stress and negative emotion are robustly affected by these fluctuations. It is

ONG, BERGEMAN, BISCONTI, AND WALLACE

746

important that our results suggest that positive emotions are a prominent feature of psychological resilience in later life. For persons high in trait resilience, the experience of positive emotion tends to bracket the experience of daily stress and negative emotion. In a recent review of the resilience research conducted over the past 30 years, Curtis and Cicchetti (2003) concluded with the following statement: “If we are to grasp the true complexity of the concept of resilience, then we must investigate it with a commensurate level of complexity” (p. 803). We join with these and other investigators (e.g., Ryff et al., 1998; Tugade et al., 2004; Zautra, Johnson, & Davis, 2005) in emphasizing that the time has come for researchers to maximize the potential advantages of combining a variety of methodological (e.g., experimental, daily diary, life story narratives) and innovative data analytic (e.g., multilevel modeling, growth mixture modeling, dynamical systems analysis) techniques for tackling the complex theoretical questions surrounding the measurement and modeling of adaptive processes.

References Affleck, G., & Tennen, H. (1996). Construing benefits from adversity: Adaptational significance and dispositional underpinnings. Journal of Personality, 64, 899 –922. Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions: Thousand Oaks, CA: Sage. Almeida, D. M. (2005). Resilience and vulnerability to daily stressors assessed via diary methods. Current Directions in Psychological Science, 14, 64 – 68. Baltes, M. M., Kuhl, K. P., & Sowarka, D. (1992). Testing for limits of cognitive reserve capacity: A promising strategy for early diagnosis of dementia? Journal of Gerontology, 47, 165–167. Baltes, P. B. (1997). On the incomplete architecture of human ontogeny: Selection, optimization, and compensation as foundation of developmental theory. American Psychologist, 52, 366 –380. Baltes, P. B., & Baltes, M. M. (Eds.). (1990). Successful aging: Perspectives from the behavioral sciences. New York: Cambridge University Press. Baltes, P. B., & Kliegl, R. (1992). Further testing of limits of cognitive plasticity: Negative age differences in a mnemonic skill are robust. Developmental Psychology, 28, 121–125. Bartone, P. T., Ursano, R. J., Wright, K. M., & Ingraham, L. H. (1989). The impact of a military air disaster on the health of assistance workers: A prospective study. Journal of Nervous and Mental Disease, 177, 317– 328. Bauer, D. J., Preacher, K. J., & Gil, K. M. (2006). Conceptualizing and testing random indirect effects and moderated mediation in multilevel models: New procedures and recommendations. Psychological Methods, 11, 142–163. Bergeman, C. S., & Wallace, K. A. (1999). Resiliency in later life. In T. M. Merluzzi & T. L. Whitman (Eds.), Life span perspectives on health and illness (pp. 207–225). Mahwah, NJ: Erlbaum. Billings, D. W., Folkman, S., Acree, M., & Moskowitz, J. T. (2000). Coping and physical health during caregiving: The roles of positive and negative affect. Journal of Personality and Social Psychology, 79, 131–142. Bisconti, T. L., Bergeman, C. S., & Boker, S. M. (2004). Emotional well-being in recently bereaved widows: A dynamical systems approach. Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 59, P158 –P167. Block, J., & Kremen, A. M. (1996). IQ and ego-resiliency: Conceptual and empirical connections and separateness. Journal of Personality and Social Psychology, 70, 349 –361.

Block, J. H., & Block, J. (1980). The role of ego-control and ego-resiliency in the organization of behavior. In W. A. Collins (Ed.), The Minnesota Symposia on Child Psychology (Vol. 13, pp. 39 –101). Hillsdale, NJ: Erlbaum. Boker, S. M., & Nesselroade, J. R. (2002). A method for modeling the intrinsic dynamics of intraindividual variability: Recovering the parameters of simulated oscillators in multi-wave panel data. Multivariate Behavioral Research, 37, 127–160. Bonanno, G. A. (2004). Loss, trauma, and human resilience: Have we underestimated the human capacity to thrive after extremely aversive events? American Psychologist, 59, 20 –28. Bonanno, G. A. (2005). Clarifying and extending the construct of adult resilience. American Psychologist, 60, 265–267. Bonanno, G. A., & Kaltman, S. (1999). Toward an integrative perspective on bereavement. Psychological Bulletin, 125, 760 –776. Bonanno, G. A., & Keltner, D. (1997). Facial expressions of emotion and the course of conjugal bereavement. Journal of Abnormal Psychology, 106, 126 –137. Bonanno, G. A., Wortman, C. B., & Nesse, R. M. (2004). Prospective patterns of resilience and maladjustment during widowhood. Psychology and Aging, 19, 260 –271. Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: Sage. Cacioppo, J. T., Gardner, W. L., & Berntson, G. G. (1999). The affect system has parallel and integrative processing components: Form follows function. Journal of Personality and Social Psychology, 76, 839 – 855. Cacioppo, J. T., Larsen, J. T., Smith, N. K., & Berntson, G. G. (2004). The affect system: What lurks below the surface of feelings? In A. S. Manstead, N. H. Frijda, & A. H. Fischer (Eds.), Feelings and emotions: The Amsterdam conference (pp. 223–242). New York: Cambridge University Press. Carver, C. S. (1998). Resilience and thriving: Issues, models, and linkages. Journal of Social Issues, 54, 245–266. Carver, C. S., & Scheier, M. F. (1999). Stress, coping, and self-regulatory processes. In O. P. John & L. A. Pervin (Eds.), Handbook of personality: Theory and research (pp. 553–575). New York: Guilford Press. Cicchetti, D., & Rogosch, F. A. (1996). Equifinality and multifinality in developmental psychopathology. Development and Psychopathology, 8, 597– 600. Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum. Curtis, W., & Cicchetti, D. (2003). Moving research on resilience into the 21st century: Theoretical and methodological considerations in examining the biological contributors to resilience. Development and Psychopathology, 15, 773– 810. Davidson, R. J. (2000). Affective style, psychopathology, and resilience: Brain mechanisms and plasticity. American Psychologist, 55, 1196 – 1214. Davidson, R. J., & Fox, N. A. (1982, December 17). Asymmetrical brain activity discriminates between positive and negative affective stimuli in human infants. Science, 218, 1235–1237. Davidson, R. J., Jackson, D. C., & Kalin, N. H. (2000). Emotion, plasticity, context, and regulation: Perspectives from affective neuroscience. Psychological Bulletin, 126, 890 –909. Davis, M. C., Zautra, A. J., & Smith, B. W. (2004). Chronic pain, stress, and the dynamics of affective differentiation. Journal of Personality, 72, 1133–1159. Eysenck, H. J., & Eysenck, S. B. (1975). Manual of the Eysenck Personality Questionnaire. London: Hodder & Stoughton. Feldman, L. A. (1995a). Valence focus and arousal focus: Individual differences in the structure of affective experience. Journal of Personality and Social Psychology, 69, 153–166.

PSYCHOLOGICAL RESILIENCE AND ADAPTATION Feldman, L. A. (1995b). Variations in the circumplex structure of mood. Personality and Social Psychology Bulletin, 21, 806 – 817. Fleeson, W. (2004). Moving personality beyond the person–situation debate: The challenge and the opportunity of within-person variability. Current Directions in Psychological Science, 13, 83– 87. Folkman, S. (1997). Positive psychological states and coping with severe stress. Social Science and Medicine, 45, 1207–1221. Folkman, S. (2001). Revised coping theory and the process of bereavement. In M. S. Stroebe, R. O. Hansson, W. Stroebe, & H. Schut (Eds.), Handbook of bereavement research: Consequences, coping, and care (pp. 563–584). Washington, DC: American Psychological Association. Folkman, S., & Lazarus, R. S. (1985). If it changes it must be a process: Study of emotion and coping during three stages of a college examination. Journal of Personality and Social Psychology, 48, 150 –170. Folkman, S., & Moskowitz, J. T. (2000a). Positive affect and the other side of coping. American Psychologist, 55, 647– 654. Folkman, S., & Moskowitz, J. T. (2000b). Stress, positive emotion, and coping. Current Directions in Psychological Science, 9, 115–118. Folkman, S., & Moskowitz, J. T. (2004). Coping: Pitfalls and promise. Annual Review of Psychology, 55, 745–774. Folkman, S., Moskowitz, J. T., Ozer, E. M., & Park, C. L. (1997). Positive meaningful events and coping in the context of HIV/AIDS. In B. H. Gottlieb (Ed.), Coping with chronic stress (pp. 293–314). New York: Plenum Press. Fredrickson, B. L. (1998). What good are positive emotions? Review of General Psychology, 2, 300 –319. Fredrickson, B. L. (2000). Cultivating positive emotions to optimize health and well-being. Prevention and Treatment, 3, 1–25. Fredrickson, B. L. (2001). The role of positive emotions in positive psychology: The broaden-and-build theory of positive emotions. American Psychologist, 56, 218 –226. Fredrickson, B. L., & Branigan, C. (2001). Positive emotions. In G. A. Bonanno & T. J. Mayne (Eds.), Emotions: Current issues and future directions (pp. 123–151). New York: Guilford Press. Fredrickson, B. L., & Branigan, C. (2005). Positive emotions broaden the scope of attention and thought–action repertoires. Cognition and Emotion, 19, 313–332. Fredrickson, B. L., & Joiner, T. (2002). Positive emotions trigger upward spirals toward emotional well-being. Psychological Science, 13, 172– 175. Fredrickson, B. L., & Levenson, R. W. (1998). Positive emotions speed recovery from the cardiovascular sequelae of negative emotions. Cognition and Emotion, 12, 191–220. Fredrickson, B. L., Mancuso, R. A., Branigan, C., & Tugade, M. M. (2000). The undoing effect of positive emotions. Motivation and Emotion, 24, 237–258. Fredrickson, B. L., Tugade, M. M., Waugh, C. E., & Larkin, G. R. (2003). What good are positive emotions in crisis? A prospective study of resilience and emotions following the terrorist attacks on the United States on September 11th, 2001. Journal of Personality and Social Psychology, 84, 365–376. Frijda, N. H. (1986). The emotions. New York: Cambridge University Press. Frijda, N. H. (1987). Emotion, cognitive structure, and action tendency. Cognition and Emotion, 1, 115–143. Frijda, N. H. (1988). The laws of emotion. American Psychologist, 43, 349 –358. Hawkley, L. C., & Cacioppo, J. T. (2004). Stress and the aging immune system. Brain, Behavior and Immunity, 18, 114 –119. Isen, A. M., Daubman, K. A., & Nowicki, G. P. (1987). Positive affect facilitates creative problem solving. Journal of Personality and Social Psychology, 52, 1122–1131. Keltner, D., & Bonanno, G. A. (1997). A study of laughter and dissocia-

747

tion: Distinct correlates of laughter and smiling during bereavement. Journal of Personality and Social Psychology, 73, 687–702. Kenny, D. A., Korchmaros, J. D., & Bolger, N. (2003). Lower level mediation in multilevel models. Psychological Methods, 8, 115–128. Kessler, R. C., & Greenberg, D. F. (1981). Linear panel analysis: Models of quantitative change. New York: Academic Press. Keyes, C. L., Shmotkin, D., & Ryff, C. D. (2002). Optimizing well-being: The empirical encounter of two traditions. Journal of Personality and Social Psychology, 82, 1007–1022. Kling, K. C., Seltzer, M. M., & Ryff, C. D. (1997). Distinctive late-life challenges: Implications for coping and well-being. Psychology and Aging, 12, 288 –295. Kreft, I., & De Leeuw, J. (1998). Introducing multilevel modeling. London: Sage. Labouvie-Vief, G. (1994). Psyche and eros: Mind and gender in the life course. New York: Cambridge University Press. Labouvie-Vief, G. (2003). Dynamic integration: Affect, cognition, and the self in adulthood. Current Directions in Psychological Science, 12, 201–206. Labouvie-Vief, G. (2005). Self-with-other representations and the organization of the self. Journal of Research in Personality, 39, 185–205. Labouvie-Vief, G., & Diehl, M. (2000). Cognitive complexity and cognitive-affective integration: Related or separate domains of adult development? Psychology and Aging, 15, 490 –504. Labouvie-Vief, G., & Medler, M. (2002). Affect optimization and affect complexity: Modes and styles of regulation in adulthood. Psychology and Aging, 17, 571–587. Larsen, J. T., Hemenover, S. H., Norris, C. J., & Cacioppo, J. T. (2003). Turning adversity to advantage: On the virtues of the coactivation of positive and negative emotions. In U. M. Staudinger & L. G. Aspinwall (Eds.), A psychology of human strengths: Fundamental questions and future directions for a positive psychology (pp. 211–225). Washington, DC: American Psychological Association. Lazarus, R. S., Kanner, A. D., & Folkman, S. (1980). Emotions: A cognitive–phenomenological analysis. In R. Plutchik & H. Kellerman (Eds.), Theories of emotion (pp. 189 –217). New York: Academic Press. Levenson, R. W. (1988). Emotion and the autonomic nervous system: A prospectus for research on autonomic specificity. In H. L. Wagner (Ed.), Social psychophysiology and emotion: Theory and clinical applications (pp. 17– 42). Oxford, England: Wiley. Linville, P. W. (1985). Self-complexity and affective extremity: Don’t put all of your eggs in one cognitive basket. Social Cognition, 3, 94 –120. Luthar, S. S., & Cicchetti, D. (2000). The construct of resilience: Implications for interventions and social policies. Development and Psychopathology, 12, 857– 885. Luthar, S. S., Cicchetti, D., & Becker, B. (2000). The construct of resilience: A critical evaluation and guidelines for future work. Child Development, 71, 543–562. Luthar, S. S., & Cushing, G. (1999). Measurement issues in the empirical study of resilience: An overview. In J. L. Johnson & M. D. Glantz (Eds.), Resilience and development: Positive life adaptations (pp. 129 –160). Dordrecht, the Netherlands: Kluwer Academic. Luthar, S. S., & Zelazo, L. B. (2003). Research on resilience: An integrative review. In S. S. Luthar (Ed.), Resilience and vulnerability: Adaptation in the context of childhood adversities (pp. 510 –549). New York: Cambridge University Press. Lyubomirsky, S., King, L., & Diener, E. (2005). The benefits of frequent positive affect: Does happiness lead to success? Psychological Bulletin, 131, 803– 855. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7, 83–104. Maddi, S. R., Khoshaba, D. M., Persico, M., Lu, J., Harvey, R., & Bleecker, F. (2002). The personality construct of hardiness: II. Relation-

748

ONG, BERGEMAN, BISCONTI, AND WALLACE

ships with comprehensive test of personality and psychopathology. Journal of Research in Personality, 36, 72– 85. Marco, C. A., & Suls, J. (1993). Daily stress and the trajectory of mood: Spillover, response assimilation, contrast, and chronic negative affectivity. Journal of Personality and Social Psychology, 64, 1053–1063. Masten, A. S. (2001). Ordinary magic: Resilience processes in development. American Psychologist, 56, 227–238. McEwen, B. S. (1998). Stress, adaptation, and disease: Allostasis and allostatic load. In J. M. Lipton & S. M. McCann (Eds.), Neuroimmunomodulation: Molecular aspects, integrative systems, and clinical advances (Vol. 480, pp. 33– 44). New York: New York Academy of Sciences. McHorney, C. A., & Ware, J. E. (1995). Construction and validation of an alternate form general mental health scale for the Medical Outcomes Study Short-Form 36-Item Health Survey. Medical Care, 33, 15–28. Mischel, W. (2004). Toward an integrative science of the person. Annual Review of Psychology, 55, 1–22. Monroe, S. M., & McQuaid, J. R. (1994). Measuring life stress and assessing its impact on mental health. In I. H. Gotlib & W. R. Avison (Eds.), Stress and mental health: Contemporary issues and prospects for the future (pp. 43–73). New York: Plenum Press. Moskowitz, J. T., Folkman, S., & Acree, M. (2003). Do positive psychological states shed light on recovery from bereavement? Findings from a 3-year longitudinal study. Death Studies, 27, 471–500. Moskowitz, J. T., Folkman, S., Collette, L., & Vittinghoff, E. (1996). Coping and mood during AIDS-related caregiving and bereavement. Annals of Behavioral Medicine, 18, 49 –57. Moss, M. S., Moss, S. Z., & Hansson, R. O. (2001). Bereavement and old age. In M. S. Stroebe, R. O. Hansson, W. Stroebe, & H. Schut (Eds.), Handbook of bereavement research: Consequences, coping, and care (pp. 241–261). Washington, DC: American Psychological Association. Mroczek, D. K., Spiro, A. I., & Almeida, D. M. (2003). Between- and within-person variation in affect and personality over days and years: How basic and applied approaches can inform one another. Ageing International, 28, 260 –278. Muller, D., Judd, C. M., & Yzerbyt, V. Y. (2005). When moderation is mediated and mediation is moderated. Journal of Personality and Social Psychology, 89, 852– 863. Nagin, D. S. (1999). Analyzing developmental trajectories: A semiparametric, group-based approach. Psychological Methods, 4, 139 –157. Ong, A. D., & Allaire, J. C. (2005). Cardiovascular intraindividual variability in later life: The influence of social connectedness and positive emotions. Psychology and Aging, 20, 476 – 485. Ong, A. D., & Bergeman, C. S. (2004a). The complexity of emotions in later life. Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 59, P117–P122. Ong, A. D., & Bergeman, C. S. (2004b). Resilience and adaptation to stress in later life: Empirical perspectives and conceptual implications. Ageing International, 29, 219 –246. Ong, A. D., Bergeman, C. S., & Bisconti, T. L. (2004). The role of daily positive emotions during conjugal bereavement. Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 59, P168 – P176. Park, C. L., & Folkman, S. (1997). Meaning in the context of stress and coping. Review of General Psychology, 1, 115–144. Potter, P. T., Zautra, A. J., & Reich, J. W. (2000). Stressful events and information processing dispositions moderate the relationship between positive and negative affect: Implications for pain patients. Annals of Behavioral Medicine, 22, 191–198. Pressman, S. D., & Cohen, S. (2005). Does positive affect influence health? Psychological Bulletin, 131, 925–971. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks: Sage. Reich, J. W., Zautra, A. J., & Davis, M. C. (2003). Dimensions of affect

relationships: Models and their integrative implications. Review of General Psychology, 7, 66 – 83. Remondet, J. H., & Hansson, R. O. (1987). Assessing a widow’s grief: A short index. Journal of Gerontological Nursing, 13, 30 –34. Rogosa, D. (1979). Causal models in longitudinal research: Rationale, formulation, and interpretation. In J. R. Nesselroade & P. B. Baltes (Eds.), Longitudinal research in the study of behavior and development (pp. 263–302). New York: Academic Press. Ryff, C. D. (1989). In the eye of the beholder: Views of psychological well-being among middle-aged and older adults. Psychology and Aging, 4, 195–210. Ryff, C. D. (1995). Psychological well-being in adult life. Current Directions in Psychological Science, 4, 99 –104. Ryff, C. D., & Singer, B. (1998). The contours of positive human health. Psychological Inquiry, 9, 1–28. Ryff, C. D., Singer, B. H., Love, G. D., & Essex, M. J. (1998). Resilience in adulthood and later life: Defining features and dynamic processes. In J. Lomranz (Ed.), Handbook of aging and mental health: An integrative approach (pp. 69 –96). New York: Plenum Press. Salovey, P., Mayer, J. D., Goldman, S. L., Turvey, C., & Palfai, T. P. (1995). Emotional attention, clarity, and repair: Exploring emotional intelligence using the Trait Meta-Mood Scale. In J. W. Pennebaker (Ed.), Emotion, disclosure, & health (pp. 125–154). Washington, DC: American Psychological Association. Schimmack, U., Oishi, S., & Diener, E. (2002). Cultural influences on the relation between pleasant emotions and unpleasant emotions: Asian dialectic philosophies or individualism– collectivism? Cognition and Emotion, 16, 705–719. Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling change and event occurrence. London: Oxford University Press. Singer, J. E., & Davidson, L. M. (1991). Specificity and stress research. In R. S. Lazarus & A. Monat (Eds.), Stress and coping: An anthology (3rd ed., pp. 36 – 47). New York: Columbia University Press. Smider, N. A., Essex, M. J., & Ryff, C. D. (1996). Adaptation of community relocation: The interactive influence of psychological resources and contextual factors. Psychology and Aging, 11, 362–372. Smith, J. (2003). Stress and aging: Theoretical and empirical challenges for interdisciplinary research. Neurobiology of Aging, 24, S77–S80. Staudinger, U. M., Marsiske, M., & Baltes, P. B. (1993). Resilience and levels of reserve capacity in later adulthood: Perspectives from life-span theory. Development and Psychopathology, 5, 541–566. Staudinger, U. M., Marsiske, M., & Baltes, P. B. (1995). Resilience and reserve capacity in later adulthood: Potentials and limits of development across the life span. In D. J. Cohen & D. Cicchetti (Eds.), Developmental psychopathology: Risk, disorder, and adaptation (Vol. 2, pp. 801– 847). Oxford, England: Wiley. Stroebe, M. S., & Stroebe, W. (1983). Who suffers more? Sex differences in health risks of the widowed. Psychological Bulletin, 93, 279 –301. Tedeschi, R. G., & Calhoun, L. G. (2004). Target article: “Posttraumatic growth: Conceptual foundations and empirical evidence.” Psychological Inquiry, 15, 1–18. Tennen, H., & Affleck, G. (2002). The challenge of capturing daily processes at the interface of social and clinical psychology. Journal of Social and Clinical Psychology, 21, 610 – 627. Tugade, M. M., & Fredrickson, B. L. (2004). Resilient individuals use positive emotions to bounce back from negative emotional experiences. Journal of Personality and Social Psychology, 86, 320 –333. Tugade, M. M., Fredrickson, B. L., & Barrett, L. F. (2004). Psychological resilience and positive emotional granularity: Examining the benefits of positive emotions on coping and health. Journal of Personality, 72, 1161–1190. Veit, C. T., & Ware, J. E. (1983). The structure of psychological distress

PSYCHOLOGICAL RESILIENCE AND ADAPTATION and well-being in general populations. Journal of Consulting and Clinical Psychology, 51, 730 –742. Ware, J. E., Gandek, B., & Group, I. P. (1994). The SF-36 Health Survey: Development and use in mental health research and the IQOLA Project. International Journal of Mental Health, 23, 49 –73. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Werner, E. E., & Smith, R. S. (1992). Overcoming the odds: High risk children from birth to adulthood: Ithaca, NY: Cornell University Press. Wessman, A. E., & Ricks, D. F. (1966). Mood and personality. Oxford, England: Holt, Rinehart, and Winston. Wortman, C. B., & Silver, R. C. (1989). The myths of coping with loss. Journal of Consulting and Clinical Psychology, 57, 349 –357. Wortman, C. B., & Silver, R. C. (1990). Successful mastery of bereavement and widowhood: A life-course perspective. In M. M. Baltes & P. B. Baltes (Eds.), Successful aging: Perspectives from the behavioral sciences (pp. 225–264). New York: Cambridge University Press. Zautra, A. J. (2003). Emotions, stress, and health. London: Oxford University Press. Zautra, A. J., Affleck, G. G., Tennen, H., Reich, J. W., & Davis, M. C. (2005). Dynamic approaches to emotions and stress in everyday life: Bolger and Zuckerman reloaded with positive as well as negative affects. Journal of Personality, 76, 1511–1538.

749

Zautra, A. J., Guarnaccia, C. A., & Dohrenwend, B. P. (1986). Measuring small life events. American Journal of Community Psychology, 14, 629 – 655. Zautra, A. J., Johnson, L. M., & Davis, M. C. (2005). Positive affect as a source of resilience for women in chronic pain. Journal of Consulting and Clinical Psychology, 73, 212–220. Zautra, A. J., Reich, J. W., Davis, M. C., Potter, P. T., & Nicolson, N. A. (2000). The role of stressful events in the relationship between positive and negative affects: Evidence from field and experimental studies. Journal of Personality, 68, 927–951. Zautra, A. J., Smith, B., Affleck, G., & Tennen, H. (2001). Examinations of chronic pain and affect relationships: Applications of a dynamic model of affect. Journal of Consulting and Clinical Psychology, 69, 786 –795. Zucker, R. A., Wong, M. M., Puttler, L. I., & Fitzgerald, H. E. (2003). Resilience and vulnerability among sons of alcoholics: Relationship to development outcomes between early childhood and adolescence. In S. S. Luthar (Ed.), Resilience and vulnerability: Adaptation in the context of childhood adversities (pp. 76 –103). New York: Cambridge University Press.

Received June 15, 2004 Revision received March 15, 2006 Accepted April 18, 2006 䡲

PSYCHOLOGICAL RESILIENCE AND ADAPTATION and well-being in general populations. Journal of Consulting and Clinical Psychology, 51, 730 –742. Ware, J. E., Gandek, B., & Group, I. P. (1994). The SF-36 Health Survey: Development and use in mental health research and the IQOLA Project. International Journal of Mental Health, 23, 49 –73. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Werner, E. E., & Smith, R. S. (1992). Overcoming the odds: High risk children from birth to adulthood: Ithaca, NY: Cornell University Press. Wessman, A. E., & Ricks, D. F. (1966). Mood and personality. Oxford, England: Holt, Rinehart, and Winston. Wortman, C. B., & Silver, R. C. (1989). The myths of coping with loss. Journal of Consulting and Clinical Psychology, 57, 349 –357. Wortman, C. B., & Silver, R. C. (1990). Successful mastery of bereavement and widowhood: A life-course perspective. In M. M. Baltes & P. B. Baltes (Eds.), Successful aging: Perspectives from the behavioral sciences (pp. 225–264). New York: Cambridge University Press. Zautra, A. J. (2003). Emotions, stress, and health. London: Oxford University Press. Zautra, A. J., Affleck, G. G., Tennen, H., Reich, J. W., & Davis, M. C. (2005). Dynamic approaches to emotions and stress in everyday life: Bolger and Zuckerman reloaded with positive as well as negative affects. Journal of Personality, 76, 1511–1538.

749

Zautra, A. J., Guarnaccia, C. A., & Dohrenwend, B. P. (1986). Measuring small life events. American Journal of Community Psychology, 14, 629 – 655. Zautra, A. J., Johnson, L. M., & Davis, M. C. (2005). Positive affect as a source of resilience for women in chronic pain. Journal of Consulting and Clinical Psychology, 73, 212–220. Zautra, A. J., Reich, J. W., Davis, M. C., Potter, P. T., & Nicolson, N. A. (2000). The role of stressful events in the relationship between positive and negative affects: Evidence from field and experimental studies. Journal of Personality, 68, 927–951. Zautra, A. J., Smith, B., Affleck, G., & Tennen, H. (2001). Examinations of chronic pain and affect relationships: Applications of a dynamic model of affect. Journal of Consulting and Clinical Psychology, 69, 786 –795. Zucker, R. A., Wong, M. M., Puttler, L. I., & Fitzgerald, H. E. (2003). Resilience and vulnerability among sons of alcoholics: Relationship to development outcomes between early childhood and adolescence. In S. S. Luthar (Ed.), Resilience and vulnerability: Adaptation in the context of childhood adversities (pp. 76 –103). New York: Cambridge University Press.

Received June 15, 2004 Revision received March 15, 2006 Accepted April 18, 2006 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 750 –762

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.750

The Differential Effects of Intrinsic and Identified Motivation on Well-Being and Performance: Prospective, Experimental, and Implicit Approaches to Self-Determination Theory Kimberly D. Burton, John E. Lydon, David U. D’Alessandro, and Richard Koestner McGill University Self-determination theory research has demonstrated that intrinsic and identified self-regulations are associated with successful adaptation. However, few distinctions are typically made between these regulations and their outcomes. In the present studies, the associations between intrinsic and identified motivations and outcomes of psychological well-being and academic performance are compared in educational settings. In Study 1, intrinsic self-regulation predicted psychological well-being, independent of academic performance. In contrast, identified regulation predicted academic performance. Additionally, the more that students demonstrated an identified academic regulation, the more that their psychological well-being was contingent on performance. In Study 2a, priming intrinsic self-regulation led to greater psychological well-being 10 days later. In Study 2b, an implicit measure of identified regulation predicted academic performance 6 weeks later. Results indicate the need to address important distinctions between intrinsic and identified regulations. Keywords: motivation, self-determination, well-being, performance

Although intrinsic and identified regulations are correlated, these constructs are theoretically distinct. Despite this, little research has examined intrinsic and identified regulatory styles separately, and there is not much empirical differentiation between the outcomes associated with each type of motivation. We propose that, in addition to the benefits associated with autonomous motivation generally, theory and research may be advanced by examining the relative contributions of intrinsic and identified regulatory styles to the prediction of positive outcomes. Briefly, in the academic domain, among both elementary school and university students, we will show that intrinsic self-regulation is an important predictor of psychological well-being outcomes, controlling for identified self-regulation, and that identified regulation is an important predictor of performance outcomes, controlling for intrinsic self-regulation.

Most things that matter in life are not easy to achieve. How do we manage to attain our goals while being happy in the process? Does a certain type of motivation lead us to work toward doing well, and another type lead us to feel good as we work? Researchers have identified styles of self-regulation that may help to answer these questions. Self-determination theory built on the classic distinction between extrinsic and intrinsic motivation by developing a continuum model of motivation (Deci & Ryan, 2000), with points along the continuum representing distinct self-regulatory styles for behavior. Past research demonstrated that the autonomous end of the continuum, comprising intrinsic and identified self-regulations,1 is associated with positive outcomes, such as psychological wellbeing (e.g., Reis, Sheldon, Gable, Roscoe, & Ryan, 2000) and academic performance (e.g., Grolnick & Ryan, 1987). Intrinsic regulation is the most autonomous of the regulatory styles and exists when people freely choose to perform an activity out of a sense of interest. In contrast, the identified regulatory style involves an individual’s recognition and acceptance of the value and importance of a behavior and the integration of this into the self.

Self-Regulatory Style and Outcomes Autonomous reasons for engaging in a particular behavior are associated with beneficial psychological outcomes, such as feeling good about an activity, goal progress, and psychological wellbeing, as well as with positive behavioral outcomes, such as school performance. Autonomous regulation is positively correlated with children’s enjoyment of elementary school (Ryan & Connell, 1989), with university students’ enjoyment of an organic chemistry class (Black & Deci, 2000), as well as with positive mood (Reis et al., 2000), vitality (e.g., Nix, Ryan, Manly, & Deci, 1999; Reis et

Kimberly D. Burton, John E. Lydon, David U. D’Alessandro, and Richard Koestner, Department of Psychology, McGill University, Montreal, Que´bec, Canada. This work was supported by research grants from the Social Sciences and Humanities Research Council of Canada and the Fonds pour la Formation de Chercheurs et l’Aide a` la Recherche du Quebec. We thank Miri Rozenek for her help with the programming of the lexical decision task and Stephanie Blum for her assistance with data entry. Correspondence concerning this article should be addressed to John E. Lydon, Department of Psychology, McGill University, 1205 Docteur Penfield Avenue, Montreal, Que´bec H3A 1B1, Canada. E-mail: [email protected]

1 Subsequent theory and research described an integrated self-regulation that is highly correlated with identified regulation but that is thought to be relatively more autonomous.

750

EFFECTS OF INTRINSIC AND IDENTIFIED MOTIVATION

al., 2000; Ryan & Frederick, 1997), and positive coping strategies (Ryan & Connell, 1989). In addition, autonomous regulation is associated with greater conceptual learning of material (Grolnick & Ryan, 1987), with academic performance (Grolnick, Ryan, & Deci, 1991), and with grades (Miserandino, 1996) in elementary school children. Likewise, junior college students who persisted in a course were significantly more autonomous at the beginning of the semester than were students who later dropped out (Vallerand & Bissonnette, 1992). Overall, research has indicated that having an autonomous self-regulatory style is associated with psychological well-being and positive behavioral outcomes. The majority of the above research treats self-regulation in terms of the self-determination continuum. For example, a number of researchers report using autonomy summary scores that are often computed from the Self-Regulation Scale (Ryan & Connell, 1989) by subtracting the sum of the nonautonomous regulations (i.e., introjected and extrinsic) from the sum of intrinsic and identified regulations, thereby creating a Relative Autonomy Index (e.g., Black & Deci, 2000; Grolnick et al., 1991; Miserandino, 1996). Others compute the same index by attaching weights to the self-regulation subscales (i.e., extrinsic ⫽ ⫺2, introjection ⫽ ⫺1, identified ⫽ ⫹1, intrinsic ⫽ ⫹2) and use this composite to predict outcome variables (e.g., Grolnick & Ryan, 1987; Reis et al., 2000; Vallerand & Bissonnette, 1992). As a result of these indexing methods, distinctions between regulations that are close together on the self-determination continuum, such as the distinctions between intrinsic and identified regulations, may be overlooked (Koestner & Losier, 2002). Implicit in the combining of intrinsic and identified selfregulation scores is that the promotion of high levels of both regulations is an appropriate objective toward which socializing agents should strive. Indeed, self-determination theory suggests that intrinsic motivation and internalization work in a complementary fashion to encourage vitality, growth, and adaptation (Deci & Ryan, 2000). Intrinsic self-regulation promotes a focus on the task itself and yields energizing emotions such as interest and excitement, whereas identification keeps one oriented toward the longterm significance of one’s current pursuits and may foster persistence at uninteresting, but important, activities. Possessing high levels of both intrinsic motivation and identification would seem to allow one the flexibility to adapt to a wide array of situations. Researchers who do not use self-determination theory as their starting point also established the adaptive value of possessing both intrinsic and more instrumental goals. For example, in a longitudinal study of talented teenagers, Wong and Csikszentmihalyi (1991) distinguished between two forms of academic motivation: intrinsic motivation and work orientation. They argued that intrinsic motivation is based on the rewards of ongoing experience, whereas work orientation reflects an investment in long-term goals such as fulfilling one’s career expectations and meeting one’s psychological needs. Their results indicated that work orientation, which we would liken to identified regulation, was significantly associated with the amount of time that students spent studying but was unrelated to their experience while studying. By contrast, intrinsic motivation was related to enjoyable studying experiences but not to the amount of time spent studying. To achieve good grades and simultaneously enjoy the process would seem to require combining self-regulation strategies that focus both on im-

751

mediate experience (i.e., intrinsic motivation) and on long-term goals (i.e., identification). Educational settings provide a fertile environment in which to study distinct regulatory styles because students often report having high levels of both intrinsic and identified motivation for their academic involvement (Vallerand, Fortier, & Guay, 1997) and because academic involvement is central to the identities of many young people (Blais, Vallerand, Brie`re, Gagnon, & Pelletier, 1990). In the present studies, we examine the correlates of intrinsic and identified regulation among elementary school and university undergraduate students. Because the intrinsic self-regulatory style reflects the positive experience that individuals have regarding an activity, such as feelings of enjoyment and interest, we hypothesized that intrinsic regulation would predict positive affect and satisfaction with life, two of the primary indices of psychological well-being (Diener & Seligman, 2004). Moreover, we expected that the link between intrinsic motivation and psychological wellbeing would not be contingent on performance. That is, the positive psychological benefits of pursuing intrinsically regulated goals should be based on the activity itself and not dependent on associated outcomes. However, because success at school involves performing deliberate, effortful, and challenging exercises, such as studying and doing homework, we hypothesized that an identified regulation would be predictive of performance, as assessed by students’ grades. Internalizing a goal into the self, as is done when an individual has an identified self-regulation, is necessary for establishing the importance of the goal and also for developing and maintaining commitment to, and persistence at working toward, the goal (Austin & Vancouver, 1996). Research has indicated that the extent to which individuals identify with their goals is predictive of their goal commitment and progress, even in the face of adversity (e.g., Lydon, Burton, & Menzies-Toman, 2005; Lydon & Zanna, 1990). Therefore, we believed that students’ academic performance would be best predicted by their levels of identified self-regulation.

Including Experimental and Implicit Cognitive Methodologies Whereas our primary objective was to examine the distinct contributions of intrinsic and identified motives in accounting for psychological well-being and performance outcomes, we also sought to expand on the methodologies typically used in selfdetermination theory research by incorporating social– cognitive theory and its methods. To date, few studies have directly used cognitive experimental methods in the examination of motivational processes. Levesque and Pelletier (2003) demonstrated that priming intrinsic motivation, or temporarily increasing its cognitive accessibility, led participants to report greater enjoyment of a laboratory activity and to show better performance than those primed with extrinsic motivation. This interesting work appears to be one of the first published reports of the incorporation of cognitive principles into the study of self-determination theory. However, similar to others, Levesque and Pelletier (2003) compared only intrinsic and extrinsic motivation and so did not examine performance differences between intrinsic and identified regulations.

752

BURTON, LYDON, D’ALESSANDRO, AND KOESTNER

In addition to the benefits of experimental control that priming affords, we viewed social– cognitive theory as an approach that would expand our understanding of how autonomous motives may guide behavior. Although a relatively large amount of research exists that examines explicit self-reported autonomous motives and their correlates, to our knowledge there is little research that uses implicit methods of assessing individual differences in autonomous motivation in the self-determination framework (cf. McClelland, Koestner, & Weinberger, 1989). Social– cognitive theory suggests that, over time, an explicit, conscious motive can come to operate in an implicit, efficient, and effortless fashion (Bargh & Chartrand, 1999; 2000). In fact, recent research indicated that some types of motivation can be activated merely by associated cues (Ratelle, Baldwin, & Vallerand, 2005). In the context of our educational paradigm, this might mean that a student who consciously identifies with an academic goal may learn to associate environmental cues, such as a textbook, with his or her motivation. Consequently, the student’s identified motivation may become activated simply by the sight of a textbook, and this may occur even when he or she is tired, distracted, or facing other demands on conscious attention, for little effort is required for such activation. Thus, a measure of motivation that operates at an implicit level would add explanatory power to research based in selfdetermination theory.

Present Studies Our goal in this set of studies was to examine intrinsic and identified self-regulations in the academic domain, with respect to performance and psychological well-being outcomes.2 Study 1 explored three primary hypotheses with a sample of elementary school children. First, we predicted that pursuing an intrinsically motivating goal would be associated with higher psychological well-being, as measured by positive affect. Second, we hypothesized that the association between intrinsic regulation and psychological well-being would not be contingent on students’ performance, as measured by their report card grades. Specifically, we believed that intrinsic motivation, assessed 7 days prior to the receipt of report card grades, would be a significant predictor of positive affect 1 day after the receipt of grades, when controlling for baseline affect and report card grades. Third, we hypothesized that identified regulation would be predictive of academic performance. We statistically controlled for students’ grade expectations because students who did well in the past and expected to do well in the future may have developed identified motives. We sought to demonstrate that identified motives assessed 7 days prior to the receipt of report cards would be predictive of report card grades, over and above students’ grade expectations. In addition to our three primary hypotheses, we examined one exploratory hypothesis. Given that progress toward autonomous goals is associated with greater well-being (Sheldon & Elliot, 1999; Sheldon & Kasser, 1998) and our prediction that, specifically, identified regulations would be positively associated with grades, we speculated that identified motives would interact with grades to predict positive affect. Our rationale was that when students recognize the importance of the goal and incorporate it into the self, that is, when they have an identified self-regulation, performing well will be rewarding and result in an increase in psychological well-being. Conversely, when such students per-

form poorly, they may feel a strong sense of disappointment and therefore experience a decrease in well-being. For all analyses in Study 1, we deemed it important to include both the intrinsic and identified self-regulation styles to examine the possible distinct contributions that one style may make to a particular outcome, relative to the other style. In Study 2a, we created an experimental test of the first two hypotheses from Study 1, with a sample of university students who were initially assessed 10 days prior to, and then only mere hours after, writing a difficult midterm examination. We hypothesized that experimentally priming intrinsic motivation would increase students’ levels of psychological well-being, measured immediately after the exam. We also predicted that this effect would not be contingent on how well students believed that they had performed on the exam. Whereas most students might be expected to experience a change in well-being dependent on their perceived exam performance, we hypothesized that those in the intrinsic priming condition would experience an increase in their psychological well-being regardless of their perceived performance. In addition, we sought to examine the relation between active involvement in goal pursuit, as measured by the amount of time spent studying, and changes in well-being. We hypothesized that among those in the intrinsic regulation condition, the amount of time that students spent studying for the exam would represent the frequency of environmental cues associated with an intrinsically motivated goal and thereby would predict increases in psychological well-being. Finally, in Study 2b, we tested our third hypothesis from Study 1, with the addition of an implicit measure of motivation. We predicted that identified self-regulation, assessed both implicitly and explicitly, would predict performance on the course’s final examination, when controlling for intrinsic regulation and previous course grades. Lastly, an exploratory goal was to examine the possibility of interactive effects between explicit and implicit self-regulations. We believed that if it is identified regulation that is most critical for performance, then having high levels of explicit or implicit identification should contribute to better final examination grades. In contrast, we expected that those with low levels on both measures would perform more poorly than others.

Study 1 With a sample of elementary school children, our goal in this study was to demonstrate the importance of distinguishing between intrinsic and identified self-regulatory styles. Briefly, we hypothesized that intrinsic self-regulation would best predict psychological well-being outcomes, such as positive affect, when controlling for baseline levels, and that this association would not be contingent on students’ performance as assessed by their report card grades. In addition, we believed that such performance outcomes would be best predicted by identified self-regulation. Finally, we sought to examine the interaction between identified 2 Ryan and Deci (2003) noted that “identifications can be thought of as more versus less compartmentalized, and only those that are well integrated within the psyche represent the full endorsement of the self” (p. 258). The measures that we use to assess identification do not currently capture this important distinction.

EFFECTS OF INTRINSIC AND IDENTIFIED MOTIVATION

regulation scores and academic performance on students’ psychological well-being.

Method Participants Participants were 241 elementary school children ranging in age from 8 years to 13 years (127 girls, M for age ⫽ 11 years, 4 months), attending schools in Toronto, Ontario, Canada. Parents and children consented to the children’s participation as well as to the school’s disclosure of report card grades to the researchers. Participants completed scales 7 days before their report cards were distributed (Time 1) and 1 day after the distribution (Time 2).

Procedure Researchers visited the classrooms of participants at two different time points, during which children completed paper and pencil questionnaires. To aid comprehension, the researchers read each item of the questionnaires aloud to the class, and children then recorded their own responses on individual questionnaire packets. Although the measures were completed in group settings, participants responded individually, and privately, at their own desks. At Time 1, children completed 24 items from Ryan and Connell’s (1989) Self-Regulation Scale, representing three of the four domains from the original scale (reasons for class work, homework, and trying to do well in school). On a 4-point scale, participants endorsed various statements reflecting the different styles of self-regulation and the extent that they explained the reasons for their own behavior with regard to school. Five items assessed intrinsic regulation (␣ ⫽ .90), and five items assessed identified regulation (␣ ⫽ .75). An example of an identified self-regulation reason is “because I want to learn new things,” whereas an example of an intrinsic regulation reason is “because it’s fun.” Children also completed the Positive and Negative Affect Scale for Children (PANAS-C; Laurent et al., 1999), which is a child-oriented adaptation of the PANAS (Watson, Clark, & Tellegen, 1988). On a 5-point scale, they rated words describing their feelings and emotions (e.g., happy, calm, sad, gloomy) in the past few weeks. The 12 items from the positive affect subscale were used to measure subjective well-being (␣ ⫽ .88).3 Finally, participants indicated their grade expectations for their upcoming report cards. Because this was the first reporting period of the school year, there was no baseline measure of students’ academic performance available for that year. As an alternative, we used students’ grade expectations as the baseline measure. One week later, after report cards had been distributed, the researchers returned to the classrooms of participating children. At this Time 2 session, participants again completed the PANAS-C (Laurent et al., 1999) as the measure of their psychological well-being, reflecting their current feelings and emotions. As at Time 1, researchers read each item aloud, and participants indicated their responses in their own questionnaires. Following this session, the researchers received copies of children’s report cards from school administration and then converted students’ expected grades and report card grades to a 13-point scale ranging from 13 (A⫹) to 1 (F).

753

positive affect, rs(239) ⫽ .15–.17, ps ⬍ .02, and with Time 2 positive affect, rs(239) ⫽ .23–.26, ps ⬍ .001. However, when examining grades, only identified regulation was significantly correlated with actual report card grades, r(239) ⫽ .26, p ⬍ .001, and expected grades, r(239) ⫽ .24, p ⬍ .001. Finally, expected grades correlated with actual grades, r(239) ⫽ .42, p ⬍ .01.

Mean Differences We made a comparison to examine whether mean positive affect scores changed from Time 1 to Time 2; however, no change was observed, t(240) ⫽ 1, ns. We then made a similar comparison between children’s expected grades (M ⫽ 9.11) and their actual report card grades (M ⫽ 8.05). Although these two grade indices were significantly correlated, r(239) ⫽ .42, p ⬍ .001, children significantly overestimated what their grades would be, t(240) ⫽ 6.87, SD ⫽ 2.40, p ⬍ .001.

Predicting Positive Affect To test our first two hypotheses that intrinsic self-regulation would be independently predictive of psychological well-being and that this association would not be contingent on performance, we conducted a hierarchical multiple regression analysis in which we sought to use intrinsic self-regulation scores to predict positive affect at Time 2. In the first step of the regression, we entered positive affect at Time 1 and students’ actual report card grades as control variables. Then, in Step 2, we entered scores on both the intrinsic and identified subscales of the Self-Regulation Scale (Ryan & Connell, 1989), and finally in Step 3, we entered the two-way interactions between report card grades and each selfregulation. As one might expect, Time 1 positive affect was a significant predictor of Time 2 positive affect, ␤ ⫽ .54, t(234) ⫽ 10.00, p ⬍ .001, as were actual report card grades, ␤ ⫽ .12, t(234) ⫽ 2.33, p ⬍ .03. Central to our hypothesis, intrinsic selfregulation scores significantly predicted greater positive affect at Time 2, ␤ ⫽ .14, t(234) ⫽ 2.22, p ⬍ .03, but identified regulation scores did not, ␤ ⫽ .04, t(234) ⬍ 1, ns. This result indicated that students with a highly intrinsic orientation toward school reported greater well-being after the receipt of their report card grades. Then, in the final step, we examined interactions between the predictors to address our second and exploratory hypotheses that the relationship between intrinsic self-regulation and positive affect would not be contingent on performance, as assessed by report cards, but that the relationship between identified self-regulation and positive affect would be contingent on performance. To address the first, we examined the interaction between intrinsic self-regulation and report card grades. This did not prove significant, ␤ ⫽ ⫺.10, t(234) ⫽ ⫺1.45, p ⫽ .15, and therefore indicated that the relationship between higher intrinsic self-regulation and greater positive affect is not contingent on performance.

Results Correlations Among Predictor Variables We first performed simple correlations between intrinsic regulation, identified regulation, positive affect at Time 1, positive affect at Time 2, expected grades, and actual report card grades. Intrinsic and identified self-regulation scores were significantly correlated with each other, r(239) ⫽ .56, p ⬍ .001, with Time 1

3 Consistent with the notion that positive and negative affect are distinct and not two ends of the same continuum, positive and negative affect were correlated but not highly so, r(239) ⫽ ⫺.30 at Time 1 and r(239) ⫽ ⫺.40 at Time 2. Neither intrinsic nor identified self-regulation, nor any interactions, were significant in predicting changes in negative affect, although as one might expect, lower report card grades were associated with increases in negative affect, ␤ ⫽ ⫺.24, t(240) ⫽ ⫺4.56, p ⬍ .01.

BURTON, LYDON, D’ALESSANDRO, AND KOESTNER

754

To address our exploratory hypothesis, we examined the interaction between identified self-regulation scores and report card grades, which proved significant, ␤ ⫽ .15, t(234) ⫽ 2.04, p ⫽ .04. The interaction is depicted in Figure 1. Among children who had lower levels of identification, there was little difference in Time 2 positive affect between those with higher report card grades and those with lower grades. However, when only children who had higher levels of identification were examined, a marked difference was observed between those with low and those with high report card grades. Students who were more identified with school but who received low grades on their report cards showed lower levels of Time 2 positive affect than did the others. By contrast, those children who were more identified with school and who received higher grades on their report cards showed greater Time 2 positive affect than did the other groups of children. Overall, intrinsic regulation positively predicted students’ wellbeing after receiving their report cards, independent of identified regulation and of grades received. However, among those who were more identified with school, the extent to which they experienced greater or lesser positive affect was contingent on the grades that they received.

Predicting Report Card Grades

Time 2 Positive Affect (z score)

To address our third primary hypothesis that identified selfregulation would be predictive of performance, we conducted a second hierarchical multiple regression analysis. We entered children’s grade expectations, assessed 1 week before report cards were distributed (Time 1), along with each self-regulation score, followed by the interactions between grade expectations and selfregulation scores. As might be anticipated, expected grades were a strong predictor of actual grades. However, identification also proved to be a significant positive predictor, ␤ ⫽ .24, t(235) ⫽ 3.32, p ⫽ .001. The more that students identified with school, the higher were their report card grades. When controlling for identified self-regulation, there was a nonsignificant trend for intrinsic self-regulation to be negatively associated with grades, ␤ ⫽ –.12, t(235) ⫽ ⫺1.67, p ⫽ .10, although the zero-order correlation with grades was not significant, r(239) ⫽ .06, ns.

0.5

0.25 Low Identified Selfregulation

0

High Identified Selfregulation

-0.25

Summary of Study 1 Findings Results suggested that intrinsic self-regulation predicts increases in indices of students’ psychological well-being, such as positive affect, and that changes in well-being are not contingent on performance, as measured by the grades that students receive. The more that students had an intrinsic academic self-regulation, the greater were the increases in their well-being, regardless of their performance. As hypothesized, identified self-regulation was a significant positive predictor of students’ academic performance such that the more identified students were with their education, the higher their grades. Identification also interacted with students’ grades to predict changes in positive affect. Although intrinsic regulation was associated with psychological well-being in a noncontingent fashion, the relationship between identified selfregulation and students’ well-being was contingent on their performance: The more that students had an identified regulation, the more their well-being was positively associated with their report card grades.

Study 2a To extend the findings of Study 1 on the differing roles of identified and intrinsic self-regulations, we conducted a second study concerning psychological well-being and performance in the context of a specific academic goal, this time with university students. On the basis of the finding that intrinsic self-regulation was positively associated with psychological well-being in Study 1, our objective in Study 2a was to experimentally obtain a similar effect. We sought to induce intrinsic and identified self-regulation orientations and to test whether the inductions of these different regulatory styles would influence students’ well-being and academic performance when writing a difficult examination. Specifically, we predicted that a manipulation involving intrinsic regulation would result in greater psychological well-being immediately following the examination, 10 days after the experimental session. Furthermore, on the basis of results of Study 1, we hypothesized that this effect would not be contingent on how well students thought that they had performed on the exam that day. In addition, we predicted that a manipulation of an identified regulation focus would result in better performance on the midterm examination. We believed that if priming an association between intrinsic motivation and the particular academic goal can influence psychological well-being, then activities related to the course, such as studying, should reactivate the link to intrinsic motivation and thereby affect individuals’ psychological well-being. Therefore, an exploratory hypothesis for Study 2a was that for those primed with intrinsic motivation, the amount of time spent studying would be related to changes in subjective well-being, whereas for those not primed with intrinsic motivation, the amount of time spent studying would be unrelated to changes in well-being.

-0.5 Low

Method

High

Report Card Grades

Figure 1. Study 1: Time 2 positive affect (z score) as predicted by students’ report card grades and identified self-regulation, controlling for Time 1 positive affect.

Participants Participants were 60 (59 female; M for age ⫽ 21 years) undergraduate students enrolled in an upper level undergraduate psychology course at McGill University in Montreal, Que´bec, Canada. At Time 1, 65 individuals

EFFECTS OF INTRINSIC AND IDENTIFIED MOTIVATION participated, with 60 returning for the Time 2 session. There were no differences in Time 1 psychological well-being between those who returned and those who did not, F(1, 63) ⫽ 0.20, MSE ⫽ 1.01, ns. Of the 5 individuals who did not return, 1 was in the control condition, 2 were in the intrinsic condition, and 2 were in the identified condition.

Procedure Recruitment. Approximately 2 weeks prior to a midterm examination, participants were recruited from a psychology class for a study examining academic goals. This class was chosen because it was not a required course, and much of the material is typically reported to be interesting to students, something that would suggest the possibility of students having at least some intrinsic motivation. The experimenters explained that those who participated in this multipart study would be remunerated $10 (U.S.$8.89) for their time. Participants completed a questionnaire 10 days prior to the exam. At the beginning of the session, the experimenters explained that, for ease of investigation, they were interested in an academic goal that was possibly common to many of the students taking part in the study: mastering the course material. All participants endorsed having this goal to some degree.4 Part 1 measures. Because participants were aware of a variety of social psychological phenomena and techniques, the questionnaire included numerous measures to maintain the appearance of it being simply a survey. In addition to a variety of individual differences measures, students were asked questions about their study habits, about their impressions of course difficulty, and about their commitment. Of importance, participants also completed a baseline measure of their current psychological wellbeing, the Satisfaction with Life Scale (Diener, Emmons, Larsen, & Griffin, 1985). In addition, participants completed a measure of perceived locus of causality (Sheldon & Kasser, 1995, 1998) to assess intrinsic and identified motives for the goal of mastering the course material. Self-regulatory manipulation. The last two components of the questionnaire composed the manipulation of regulatory style. Participants were randomly assigned to one of three conditions: control, intrinsic regulation, or identified regulation. For those in the control condition, the questionnaire ended following the completion of the last scale, whereas for those in either of the two experimental conditions, the questionnaire continued for one extra page. The first part of the manipulation consisted of a list of statements with which participants were asked to indicate their agreement by writing the word “Yes” (to denote “Yes, I agree at least somewhat.”) or the word “No” (to denote “No, I disagree completely.”) next to each statement. These response options were constructed to increase the likelihood that participants would endorse the statements on the list (Salancik, 1974). With items included in the identified and intrinsic subscales of Ryan and Connell’s (1989) Self-Regulation Scale, two lists of seven statements were created, one for each condition. For example, in the intrinsic condition, participants were presented with items such as “I find the course material interesting” and “I enjoy the course material,” whereas those in the identified condition endorsed items such as “Mastering the course material is important to me” and “I value being able to learn from the course material.” All participants endorsed the vast majority of the statements, and there were no differences in endorsement between the priming conditions, F(1, 41) ⫽ 1.74, MSE ⫽ 0.04, ns. Following the completion of the statement component, participants were asked to write about their goal of mastering the course material in terms of it being fun, enjoyable, and interesting in the intrinsic condition or in terms of values, identity, and meaning in the identified condition. These words were again chosen on the basis of Ryan and Connell’s (1989) work. Participants in both conditions were able to write about their goals and wrote passages that were approximately equal in length. E-mail manipulation booster. At the end of the experimental session, participants were informed that the second component of the study would take place approximately 10 days later, but that the exact date, time, and

755

location had not yet been set. Under the guise of providing the Time 2 information, participants were asked to record their e-mail addresses for the experimenters if they wished to be contacted. All participants complied. Approximately 4 days prior to the midterm exam, when students were presumably in the process of studying, the experimenters sent an e-mail message to each participant. These messages varied according to condition. In the control condition, the message simply contained the necessary information for attending the next session of the study. However, in the experimental conditions, the message served to reinforce the Time 1 manipulation. In the intrinsic condition, part of the e-mail message read “Thank you for your continued participation in our study on the ways in which students find their academic goal of mastering course material to be interesting and enjoyable,” whereas in the identified condition, this same sentence finished with “the ways in which students identify with their academic goal of mastering course material and find this goal to be important.” Participants returned to the lab for the second session shortly after they finished writing their midterm examination, for which the mean grade was 66%. In this second component, they completed a questionnaire containing measures similar to those assessed in the first session, including their current satisfaction with life, and indicated the grades that they anticipated receiving on the midterm, based on their experience of having just written the test as well as the amount of time they spent studying for the test.

Results Effect of the Manipulation on Psychological Well-Being To address our hypothesis that the induction of an intrinsic self-regulation focus would positively affect psychological wellbeing, we conducted an analysis of covariance, controlling for initial well-being scores. Results indicated a significant main effect of condition on satisfaction with life scores after the midterm test, F(2, 56) ⫽ 5.60, MSE ⫽ 0.37, p ⬍ .001, whereby the well-being of those in the intrinsic condition (M ⫽ 5.50) was significantly higher than that of those in the control condition (M ⫽ 4.90), t(56) ⫽ 3.17, p ⬍ .01, and than that of those in the identified condition (M ⫽ 5.01), t(56) ⫽ 2.55, p ⬍ .01. In addition, a repeated measures analysis of variance revealed a significant Time ⫻ Condition effect, F(2, 57) ⫽ 4.14, MSE ⫽ 0.27, p ⫽ .02, such that there was a significant increase in satisfaction with life from Time 1 to Time 2 for those in the intrinsic condition, t(57) ⫽ 3.12, p ⬍ .01, but not in the control condition, t(57) ⫽ 0.93, ns, or in the identified condition, t(57) ⫽ 0.87, ns. The means from this analysis are presented in Table 1. These results indicated that the intrinsic manipulation led to significant improvements in individuals’ psychological well-being after writing the midterm examination.

Effect of the Manipulation on Midterm Examination Grades Contrary to our hypothesis, an analysis of covariance in which we controlled for previously obtained course grades revealed no significant effects of the manipulations on students’ midterm examination grades, F(2, 50) ⫽ 0.08, MSE ⫽ 1.04, ns. 4 Of the participants, 63% endorsed the goal by indicating “Yes, I have this goal.” The remaining 37% of participants indicated that “Yes, I sort of have this goal.”

BURTON, LYDON, D’ALESSANDRO, AND KOESTNER

Table 1 Study 2a: Mean Satisfaction With Life Scores as a Function of Experimental Condition Point of assessment

a

Condition

Time 1

Time 2

Control Identified Intrinsica

5.17 4.73 5.00

5.01 4.87 5.53

Significant increase from Time 1 to Time 2, p ⬍ .01.

Effect of Perceived Midterm Grades on Psychological Well-Being We conducted further analyses to examine the role of students’ midterm exam grade perceptions in the prediction of subjective well-being after the midterm examination had taken place. On the basis of the results of Study 1 that indicated that the psychological well-being of those who have highly intrinsic regulation is not contingent on their performance, we believed that the well-being of students in the intrinsic induction condition would not be related to their perceived performance on the test. In contrast, we believed that among students not in the intrinsic condition, those who felt that they had performed well on the exam would experience greater subjective well-being than those who felt that they had performed poorly. In a hierarchical multiple regression, Time 1 satisfaction with life scores were entered to control for baseline levels of well-being. Then, students’ perceptions of their midterm exam grades and dummy coded variables reflecting the manipulation were entered. Finally, the interactions between students’ grade perceptions and the manipulation variables were entered. To test for three levels of the experimental manipulation, it was necessary to create two dummy codes, one contrasting the intrinsic condition with the other groups and the second contrasting the identified condition with the other two groups. Results revealed that the intrinsic condition was a highly significant predictor of well-being, ␤ ⫽ .28, t(53) ⫽ 3.08, p ⫽ .003, and that grade perceptions approached significance in prediction of well-being, ␤ ⫽ .25, t(53) ⫽ 1.72, p ⫽ .09, following the midterm. In addition, the Perception of Midterm Grades ⫻ Condition interaction approached significance, ␤ ⫽ ⫺.24, t(53) ⫽ ⫺1.96, p ⫽ .055. This interaction is depicted in Figure 2, whereby the psychological well-being of those students in the intrinsic manipulation condition was not affected by how well they felt that they did on the midterm exam ( pr ⫽ ⫺.15), but the well-being of all other participants was contingent on their perception of having performed well versus having performed poorly ( pr ⫽ .30). Among those not in the intrinsic condition, students experienced greater well-being if they felt that they had performed well on the exam but experienced lower levels of well-being if they felt they had performed poorly. The intrinsic manipulation appeared to serve as a protective factor against these contingencies, with those in the intrinsic condition experiencing greater psychological well-being regardless of their perceptions of achievement. There was no significant main effect or interaction with the dummy code for the identified condition, ts(53) ⬍ 1, ns. Similarly,

when examining the psychological well-being of only those participants in the identified regulation condition, the partial correlation between changes in psychological well-being and midterm exam grade perceptions, controlling for previously obtained course grades, did not reach statistical significance, pr(14) ⫽ .27, p ⫽ .26.

Exploratory Analyses: Days Spent Studying and the Intrinsic Manipulation To investigate a possible mechanism for how the intrinsic manipulation led to greater psychological well-being, we conducted exploratory analyses. In theory, the intrinsic manipulation should have led to a cognitive association between the engagement in course activities, such as studying, and feelings of enjoyment. As a result, the more that students in this condition spent time studying for the midterm examination, the more that they should have experienced greater well-being. Therefore, we investigated differences in the relationship between the amount of time spent studying and changes in psychological well-being from Time 1 to Time 2. In both the control condition, r(17) ⫽ .11, ns, and the identified condition, r(20) ⫽ .05, ns, correlations were not significant. However, in the intrinsic condition there was a significant relationship between changes in satisfaction with life and the amount of time spent studying, r(15) ⫽ .56, p ⫽ .02. The more that participants in the intrinsic condition studied, the greater was the increase in their psychological well-being from Time 1 to Time 2. A comparison of effect sizes revealed that this correlation was significantly greater than those in the identified and control conditions, z ⫽ 1.75, p ⬍ .05, one-tailed. It should be noted that the manipulation did not lead participants in the intrinsic condition (M ⫽ 4.62) to spend more time studying than those in the identified condition (M ⫽ 4.55) or in the control condition (M ⫽ 4.97), F(2, 55) ⬍ 1, ns. In addition, the manipulation did not interact with any measure of individual differences to predict the number of days that students spent studying, Fs(2, 55) ⬍ 1, ns.

Summary of Study 2a Findings Results indicate that the manipulation of intrinsic self-regulation in students’ approach to an academic goal significantly improved

Time 2 Satisfaction with Life (z score)

756

1

0.5 Intrinsic Condition

0

Other Conditions

-0.5

-1 Low

High

Perceived Midterm Exam Grade

Figure 2. Study 2a: Time 2 satisfaction with life (z score) as predicted by condition and students’ perceived midterm examination grades, controlling for Time 1 satisfaction with life.

EFFECTS OF INTRINSIC AND IDENTIFIED MOTIVATION

their subsequent psychological well-being. This intrinsic regulation manipulation appeared to serve as a protective factor against the grade-contingent changes in well-being that were experienced by the other participants in the study. Finally, exploratory analyses suggested a possible mechanism for how students in the intrinsic manipulation condition experienced greater well-being following the midterm exam. In that condition, the more time that students spent studying, the greater was the increase in psychological well-being from Time 1 to Time 2. Presumably, working toward their academic goal with an intrinsic focus allowed them to experience the positive outcomes associated with intrinsic self-regulation.

Study 2b Study 1 revealed that identified self-regulation was associated with better academic performance, as indicated by students’ report card grades. In Study 2b, we sought to expand on this finding in four ways. First, we extended the time frame between the assessment of self-regulation and academic performance to 6 weeks. Second, we assessed students’ academic self-regulatory styles two thirds of the way through the course to ensure that regulation scores were based on a substantial amount of actual experience in the goal pursuit (Fazio & Zanna, 1981). Third, we were able to use previously obtained course grades as a baseline for academic performance. Last, we created a cognitive measure of selfregulation that enabled us to examine the independent, and possibly interactive, effects of an explicit self-report measure and an implicit cognitive measure in predicting academic performance.

Method Participants Of the 60 participants in Study 2a, 53 remained in the sample for Study 2b. Of the participants, 4 did not consent to having their course grades used in the research, and 3 did not have a full set of grades for the semester, leaving a total of 53 participants for analyses involving academic performance.

Procedure Approximately 6 weeks prior to the final examination in the course, participants completed a lexical decision task to ascertain the cognitive accessibility of intrinsic and identified self-regulation words. This task was included to provide an implicit measure of intrinsic and identified regulations. In addition, participants’ scores on the intrinsic and identified items of Sheldon and Kasser’s (1998) measure of perceived locus of causality, collected at the beginning of Study 2a, were used as direct indices of intrinsic and identified regulation. On a 9-point scale, participants rated their agreement with reasons for pursuing the academic goal of mastering course material. On average, students reported high levels of intrinsic (M ⫽ 6.19, Mdn ⫽ 7) and identified (M ⫽ 7.03, Mdn ⫽ 7) self-regulation. Lexical decision task. Using E-Prime software (Psychology Software Tools, n.d.), we presented participants with letter strings on a computer monitor and asked them to indicate whether each string constituted a word or a nonword by pressing particular keys on the computer keyboard. In the initial instructions, participants were reminded of the course name and number. To establish the appropriate context, prior to the presentation of each letter string, the course number 333 was subliminally flashed for 20 ms, followed by a mask of XXX, similar to the methodology used by Mikulincer, Birnbaum, Woddis, and Nachmias (2000). Half of the letter

757

strings were nonwords, which resembled actual words but had letters missing or out of order, and the other half were words. Some words were determined to be of neutral valence (Anderson, 1968) and would later be used to control for individual differences in baseline responding. Words representing the two different types of self-regulation were selected by again borrowing from Ryan and Connell’s (1989) Self-Regulation Scale. Intrinsic items included the words interesting, enjoyable, fun, exciting, and fascinating, whereas identification items included important, meaning, value, identity, and worthwhile. All strings were randomly presented and response latencies to each were measured in milliseconds. Permission to obtain grades. At the conclusion of the session, the experimenters explained to participants that they were also interested in students’ academic performance and how it may be related to variables assessed during the course of the study. Participants were then provided with the option of giving the experimenters consent to obtain their final examination grades. All but 4 individuals consented.

Results Preliminary Analyses To verify that the experimental manipulations in Study 2a did not influence the implicit measures of self-regulation in Study 2b or students’ final examination grades, we conducted analyses of variance. Results indicated that implicit measures (i.e., response latencies) did not differ between the experimental conditions outlined in Study 2a, Fs(2, 50) ⫽ 0.16 – 0.71, MSEs ⫽ 1.05–1.10, ns. Similarly, final examination grades were not affected by the earlier Study 2a manipulations, F(2, 50) ⫽ 0.78, MSE ⫽ 1.01, ns. In addition, explicit measures of self-regulation, assessed prior to the manipulation in Study 2a, were not affected by experimental condition, Fs(2, 62) ⫽ 0.32– 0.74, MSEs ⫽ 1.01–1.02, ns. Thus, the manipulations in Study 2a did not influence implicit or explicit self-regulation scores nor did they influence final examination scores.

Predicting Final Exam Grades: Identified Versus Intrinsic Regulations To create indices of implicit intrinsic and identified regulation, we first performed a logarithmic transformation, as suggested by Fazio (1990), to normalize the positively skewed distribution of reaction time data (Bargh & Chartrand, 2000). Then, to create indices for both of the self-regulation styles in question, we aggregated response latencies to the intrinsic and identified words, respectively. Following this, we removed the variance associated with response latencies to the words of neutral valence from each index to control for individual differences in general speed of responding. Analyses involving the implicit measures of selfregulation reflected these computed values. Lower scores on these indices denote faster response latencies and therefore higher levels of intrinsic and identified motivation. The explicit (self-report) and implicit (response latencies) measures of identification were significantly correlated with each other, r(51) ⫽ ⫺.27, p ⫽ .05, and with this same pattern, the correlation between explicit and implicit measures of intrinsic regulation approached significance, r(51) ⫽ ⫺.25, p ⫽ .07. In addition, explicit measures of identified and intrinsic selfregulations were correlated with each other, r(51) ⫽ .31, p ⫽ .02, as were the implicit measures of identified and intrinsic selfregulations, r(51) ⫽ .29, p ⫽ .03.

758

BURTON, LYDON, D’ALESSANDRO, AND KOESTNER

In addressing our hypothesis that measures of identified selfregulation, both explicit and implicit, would significantly predict academic performance, we conducted two hierarchical multiple regression analyses in which we sought to predict final examination grades. In the first regression, we entered students’ previous course grades in Step 1 to control for general differences in students’ academic achievement. Then, in Step 2, we entered all variables representing intrinsic self-regulation: explicit (selfreport) intrinsic regulation scores, implicit (response latencies) intrinsic regulation scores, and the Explicit ⫻ Implicit Intrinsic Regulation interaction. Finally, in Step 3 we entered all variables representing identified self-regulation scores (explicit identified, implicit identified, and the Explicit ⫻ Implicit Identified Regulation interaction). By examining the R2change from Step 2 to Step 3, we sought to determine if the addition of identified self-regulation significantly contributed to the prediction of students’ final examination grades above and beyond what was accounted for by intrinsic regulation scores. Results indicated that intrinsic regulation, entered in Step 2, did not significantly predict final exam grades, R2change ⫽ .06, Fchange(3, 48) ⫽ 1.88, p ⫽ .15, but that identified regulation, entered in Step 3, did add significantly to the prediction of exam grades, R2change ⫽ .13, Fchange(3, 45) ⫽ 5.04, p ⫽ .004. We then conducted a similar second hierarchical multiple regression, this time reversing the order of entry of intrinsic and identified self-regulation scores. As in the first regression, we entered previous course grades in Step 1. Then, in Step 2, we entered the identified regulation variables (explicit, implicit, and Explicit ⫻ Implicit), and in Step 3, we entered the intrinsic regulation variables (explicit, implicit, and Explicit ⫻ Implicit). Results revealed that identified regulation significantly predicted final exam grades, R2change ⫽ .15, Fchange(3, 48) ⫽ 5.36, p ⫽ .003, but that the addition of intrinsic regulation did not significantly contribute to the prediction of grades, R2change ⫽ .05, Fchange(3, 45) ⫽ 1.77, p ⫽ .17. Together, the results of these two regression analyses indicated that identified self-regulation significantly, and independently of intrinsic self-regulation, predicted students’ academic performance, as measured by their final examination grades, but that intrinsic self-regulation itself was not a significant predictor of performance. The more that students had an identified self-regulation, the higher were their grades on the final examination.

Individual Predictors of Final Examination Grades In addition to examining the role of self-regulation style as a whole in the prediction of final exam grades, we examined the individual contributions of the different self-regulation measures by conducting partial correlations in which we controlled for previously achieved course grades and by conducting a hierarchical multiple regression. The partial correlation between the explicit measure of identified regulation and final exam grades approached significance, pr(50) ⫽ .25, p ⫽ .07, as did that between the explicit measure of intrinsic regulation and final exam grades, pr(50) ⫽ .26, p ⫽ .06. The implicit measure of identification proved to be the most highly related to final exam grades, pr(50) ⫽ ⫺.36, p ⫽ .01, whereby the more cognitively accessible that identified words were, as indicated by shorter response latencies, the higher were students’ grades on the final examination. Though similar, the

partial correlation between response latencies to intrinsic words and final exam grades did not reach significance, pr(50) ⫽ ⫺.23, p ⫽ .11. Then, in a hierarchical multiple regression analysis, we entered previous course grades to again control for individual differences in academic achievement in Step 1. As might be expected, previous course grades were a highly significant predictor of final exam grades, ␤ ⫽ .65, t(51) ⫽ 6.04, p ⬍ .001. In Step 2, we entered the terms for each of the explicit measures and each of the implicit measures. The improvement to the model approached significance, R2change ⫽ .08, Fchange(2, 48) ⫽ 2.41, p ⫽ .10. The sole main effect accounting for unique variance, controlling for the contributions of the other measures, was the implicit measure of identified regulation that also approached significance, ␤ ⫽ ⫺.26, t(48) ⫽ ⫺2.02, p ⬍ .07. The explicit measures of identification and intrinsic self-regulation and the implicit measure of intrinsic self-regulation did not explain any unique variance in final exam grades, ts(48) ⬍ 1.1, ns. Finally, in Step 3, we entered the interaction terms between explicit and implicit measures of identified self-regulation and between explicit and implicit measures of intrinsic self-regulation. This provided a significant contribution to the model, R2change ⫽ .14, Fchange(1, 46) ⫽ 4.89, p ⬍ .02, with the interaction between explicit and implicit measures of identification proving to be a significant predictor of final examination grades, ␤ ⫽ .34, t(46) ⫽ 3.15, p ⬍ .01. The interaction term for intrinsic self-regulation, however, was not significant, t(46) ⬍ 1.

Decomposing the Explicit ⫻ Implicit Identified Regulation Interaction We then conducted an analysis of covariance (controlling for previous course grades) to further decompose the Explicit ⫻ Implicit Identified Regulation interaction and make specific cell comparisons. To this end, we used median splits of the explicit and implicit measures of identified self-regulation. As can be seen in Table 2, this analysis revealed a significant interaction, F(1, 48) ⫽ 3.98, MSE ⫽ 72.41, p ⫽ .05, such that those with low scores on both the explicit and implicit measures of identification had significantly lower final exam grades than did those in the other three groups, ts(48) ⬎ 2.25, ps ⬍ .05. For example, having low selfreported identification did not result in lower examination grades unless one also had low implicit identified regulation scores. Those students who had high levels of either explicit or implicit identification achieved higher grades on the final exam than did those with low scores on both measures of identified regulation, with the

Table 2 Study 2b: Mean Final Examination Grades as a Function of Identified Self-Regulation, Controlling for Previous Course Grades Implicit identification (chronic accessibility) Explicit identification (self-reported)

Low

High

Low High

66.14a (n ⫽ 17) 75.26b (n ⫽ 9)

73.86b (n ⫽ 15) 73.34b (n ⫽ 12)

Note.

Means that do not share subscripts differ at p ⬍ .01.

EFFECTS OF INTRINSIC AND IDENTIFIED MOTIVATION

final exam grades of those with high scores on one or both measures of identified self-regulation not differing from one another, ts ⬍ 1, ns.

759

an interaction between identified regulation and academic performance on psychological well-being, as we did in Study 1.

Strengths of the Research Summary of Study 2b Findings Results highlighted the importance of identified self-regulation in the prediction of academic achievement. Further, the inclusion of an implicit measure of identification, in addition to the explicit (self-report) measures typically used by self-determination theory researchers, provided a significant contribution in the prediction of students’ grades on the course final examination, written more than 1 month following the laboratory session.

Discussion Across studies, we found evidence that intrinsic and identified self-regulations differ in their relative influence on psychological well-being and goal performance. By using a variety of methodologies and statistical controls, we obtained consistent results with children and young adults in academic settings. In Study 1, examining elementary school students longitudinally, we first showed that intrinsic self-regulation positively predicted changes in students’ psychological well-being and that this was independent of their academic performance. Second, we demonstrated that identified regulation positively predicted students’ performance, independent of intrinsic self-regulation, such that those with greater levels of identification had higher report card grades. Finally, we saw evidence that identified self-regulation interacted with academic performance to predict levels of psychological well-being. Unlike the relationship between intrinsic regulation and wellbeing, the association between identified self-regulation and wellbeing was contingent on students’ academic performance. Among those who had high levels of identified self-regulation, well-being was dependent on report card grades, whereas among those who had low levels of identified regulation, there were no such contingencies. Then, in Study 2a, we found additional evidence for the influence of intrinsic self-regulation on psychological well-being by showing that those university students who underwent an experimental intrinsic regulation induction had significantly higher levels of well-being 10 days after writing a midterm examination. Although we failed to demonstrate a contingency between the well-being of students in the identified manipulation condition and their perceived exam grades, we did show that students in the intrinsic condition did not have contingencies between their perceived performance and their psychological well-being and that their active involvement in the pursuit of the goal was related to their well-being. That is, the amount of time that these students spent studying was positively associated with changes in their life satisfaction. Finally, in Study 2b, we obtained further evidence for the importance of identified regulations in the prediction of performance. With the addition of an implicit measure of identified regulation, we saw that final examination grades, obtained over 1 month later, were best predicted by students’ levels of identified self-regulation, and that this was independent of intrinsic regulation. Unfortunately, we did not have a measure of well-being from the time of the final examination, making it impossible to test for

In each study we sought to use statistical or experimental controls to strengthen our empirical tests. For example, in testing the link between intrinsic self-regulation and psychological wellbeing, we controlled for the grades that students received on their report cards, in Study 1, and for the grades that students thought they would receive on the exam they had just written, in Study 2a. In both cases, intrinsic self-regulation predicted psychological well-being, independent of actual or perceived academic performance. This supports the notion that well-being derived from intrinsic pursuits is not contingent on outcome but rather may develop from the positive feelings and satisfaction that are associated with the intrinsically interesting nature of the activity. By using an experimental manipulation, the results of Study 2a demonstrate a causal influence of intrinsic self-regulation that can be sustained over time. We theorized that the manipulation strengthened the association between intrinsic self-regulation and the pursuit of the goal, which, in this case, was mastering the course material. As a result, actions associated with the goal served as cues that reactivated the cognitive link between intrinsic selfregulation and the goal, thereby bolstering psychological wellbeing. We obtained preliminary evidence for this in the finding that the amount of time that students spent studying was strongly associated with changes in levels of psychological well-being for those in the intrinsic condition but not for those in the other conditions. The more time that students in the intrinsic condition spent being actively involved in the pursuit of their goal (i.e., studying), the more that they experienced psychological well-being. The effectiveness of our intrinsic self-regulation manipulation highlights the value of self-determination theory by offering researchers, educators, and mental health practitioners tools that may help individuals to improve their psychological well-being. Such findings provide the possibility of aiding people to constructively deal with situations in which they may find themselves working toward goals that are not consistently pleasant. Reconsidering the goal in terms of intrinsic reasons may help people to reap the psychological benefits associated with pursuing intrinsically motivated goals. In other words, framing the pursuit of a goal in terms of it being interesting, enjoyable, and fun may help to improve people’s general well-being, as it did in our study (see also Sheldon, Kasser, Smith, & Share, 2002; Vansteenkiste, Simons, Lens, Sheldon, & Deci, 2004). Because the manipulation was relatively easy and straightforward to administer, it seems reasonable to expect that, with some initial guidance, people may be able to restructure their own thoughts regarding the reasons for goal pursuit, and therefore positively affect their psychological wellbeing. As cognitive and self-determination theorists might point out, however, such a procedure cannot make a goal become intrinsically motivating. Testament to this was Ratelle et al.’s (2005) unsuccessful attempt to create and condition intrinsic motivation in contrast to their success with the same methodology at conditioning extrinsic motivation. We believe that our procedure can lead individuals to discover and recognize the existing intrin-

760

BURTON, LYDON, D’ALESSANDRO, AND KOESTNER

sically motivating aspects of their goal pursuits but would not successfully create intrinsic motivation ex nihilo. In testing the association between identified self-regulation and performance, we wished to guard against the possibility that previous academic success might lead to higher levels of identified self-regulation and thereby inflate any of the correlations that we would observe. Therefore, in Study 1 we controlled for students’ expectations of their grades, which are likely highly correlated with their previous grades that were unavailable to us, and in Study 2b, we controlled for the grades that students obtained earlier in the term. In both cases, identified self-regulation predicted subsequent academic performance and did so even when controlling for intrinsic self-regulation. It is clear from both Study 1 and Study 2a that the link between intrinsic regulation and psychological well-being is not contingent on performance. The results for identified regulation were more equivocal, with Study 1 showing that psychological well-being was contingent on performance for highly identified individuals. This was not replicated in Study 2a, but this may be due to the ineffectiveness of the identified regulation manipulation in that experiment. Perhaps with an effective manipulation of identification, we would observe the same contingency as that found in Study 1 as well as a greater amount of time spent pursuing the goal. Given that previous research indicated that progress toward autonomous goals (intrinsic and identified combined) leads to greater well-being (Sheldon & Elliot, 1999; Sheldon & Kasser, 1998), researchers may want to further examine the role of contingencies in well-being, with a specific focus on identified selfregulation, and to develop other ways of manipulating the salience of identification. An exciting feature of the results concerning identified selfregulation and performance is our successful assessment of identification at an implicit level. Because self-determination researchers traditionally measure motivation on an explicit level, one might assume that autonomous self-regulations guide behavior only in a highly self-conscious manner. Our results suggest that identified regulations also operate at an implicit level and therefore may have a more powerful and extensive influence on behavior. Theory and research suggest that we are often too tired or mentally taxed to exert conscious, effortful self-regulation (Baumeister, Bratslavsky, Muraven, & Tice, 1998) but that motives operating at an implicit level can guide behavior even under such circumstances (Bargh & Chartrand, 1999). Perhaps internalizing the importance of a goal, as is done when one has an identified self-regulation, leads individuals to construct mental scripts for how to pursue their goals that, subsequently, they are able to follow in a relatively automatic fashion. The incorporation of implicit measures of self-regulation may therefore begin to provide useful insight into how identified regulation operates to affect behavior.

Future Research Although our research demonstrates that performance is best predicted by identified self-regulation scores, we acknowledge that there may be ways in which intrinsic self-regulation enhances academic performance (Cordova & Lepper, 1996). When we consider that, in Study 2a, students in the intrinsic experimental condition maintained their subjective well-being even when they

thought that they had performed poorly that day, it seems possible that intrinsic regulation can positively affect performance. By maintaining their psychological well-being, these students may be more likely to attend the next class and may be quicker to reopen their textbooks. Unlike students who became deflated by their weak performance, these students may be better able to continue with their goal pursuit and improve their subsequent performance. Another direction for future research would be to examine intrinsic and identified self-regulations in other contexts. We chose to recruit participants from classroom settings, an elementary school and a university course, to focus on a single real-life goal common to all participants. Moreover, the goal, an academic one, is a goal that is often used to study motivation (e.g., Grolnick & Ryan, 1987; Vallerand & Bissonnette, 1992). These two samples allowed for the examination of motivational processes at two very different age points and added to the external validity of our findings. However, we acknowledge that confining our research to educational settings presents some limitations. Gagne´ and Deci (2005) posited that when pursuing a goal requires effort and persistence at important but relatively uninteresting tasks, as is often the case with education, identified regulation may be especially predictive of performance. In contrast, when the pursuit of a goal requires less effortful persistence at interesting tasks, as might be more common with self-selected goals, intrinsic regulation may be predictive of performance. Using another context in which to study the differences between intrinsic and identified selfregulations would therefore be valuable. A limited amount of research already has suggested that such distinctions are visible in the nonacademic domains of politics (Koestner, Losier, Vallerand, & Carducci, 1996), close relationships (Lydon et al., 2005), and sports. For example, there is evidence that athletes’ intrinsic selfregulation is associated with their happiness during the season but that their levels of identified self-regulation are associated with how many points they actually score (Paquin, 2005). An additional direction for future research would be to test a mediational model of identification and performance. Theoretically, we would expect identification to be associated with increased effort, persistence, and deliberate practice. This should, in turn, enhance performance over time, especially in contexts that are challenging and difficult. Given that our research makes an empirical distinction between the effects of intrinsic and identified self-regulation, future research might examine other regulations on the self-determination continuum. We believe that a particularly interesting comparison would be between identified and introjected regulations. Although introjection is theorized to be adjacent to identification on the continuum, correlational research suggests that identification is more strongly associated with intrinsic regulation than it is with introjection (Koestner & Losier, 2002). This may result in strong distinctions between these two types of self-regulation and the outcomes with which they are associated. Although highly interesting, experimentally inducing an introjection focus for a real world goal may pose too great a risk because of possible negative effects on well-being and performance. Therefore, researchers may wish to examine introjection experimentally within the context of an artificial goal, limited to the laboratory.

EFFECTS OF INTRINSIC AND IDENTIFIED MOTIVATION

Self-Regulation and the Pursuit of Goals Over Time Whereas our research points out the importance and value of distinguishing between intrinsic and identified motivations, future research might expand on this and examine the development of these motivations over time in the pursuit of specific, meaningful goals. Intrinsic motivation may often act as an initial engine that fuels goal pursuit. However, long-term and significant goals require learning tasks, developing and improving skill sets, and sustaining effort. As Ryan (1995) noted, most meaningful life goals are not always fun and enjoyable, and as Dweck (1999) observed, all significant accomplishments require overcoming obstacles and failures. Therefore, a key challenge in goal pursuit may be the development of identification with the goal. When goal adversity is high, identification should be critical to goal pursuit, whereas when adversity is low, intrinsic motivation may be predictive of goal pursuit. Just as intrinsic and identified motivations are differentially related to various positive outcomes, progress toward these outcomes may be differentially related to motivation via feedback loops. Intrinsic motivation may be influenced by an affect feedback loop in the pursuit of goals, whereas identified motivation may be influenced by goal attainability or progress. We believe that what should be most important for sustaining intrinsic motivation is that the activity continues to meet the individual’s needs for psychological well-being, whereas what should be critical for maintaining identified motivation is that the activity moves one closer to success at achieving goals that are self-defining and express one’s identity. Again, this underscores the potential theoretical value in distinguishing between intrinsic and identified regulations. They both are associated with positive outcomes, but attention to the distinct contributions of each may help researchers to specify how these outcomes are achieved and how the motivations are sustained.

Summary and Conclusion Although there are theoretical distinctions between intrinsic and identified self-regulations, little empirical support exists for these differences. This set of studies provides evidence for such distinctions by differentiating between the outcomes associated with intrinsic and identified self-regulations, by using longitudinal, experimental, and implicit cognitive methodologies. In Study 1, we examined self-regulation and outcomes longitudinally. Then, in Study 2a, we were able to experimentally manipulate intrinsic regulation, and later, in Study 2b, we measured the chronic accessibility of regulations in context. The incorporation of these methods represents a new approach that advances and strengthens existing work on autonomous motivation and points to new possibilities concerning how motivation influences individuals’ psychological states and behavior. Achieving personally significant goals, and being happy doing so, are key to optimizing human potential. Identified motivation represents the extent to which a goal is genuinely meaningful and important to individuals. It functions to sustain energy and effort in goal pursuit when one is faced with challenges, stressors, and even boredom. Intrinsic motivation appears to play a regulating role in these goals, for when there is no inherent interest or enjoyment associated with a particular goal, psychological well-being may

761

suffer. But, because the relationship between intrinsic motivation and well-being is not contingent on performance, possessing intrinsic motivation maintains our well-being even when we are faced with setbacks. Fortunately, the likelihood of encountering setbacks is reduced by identified motivation because of its ability to mobilize energy and induce deliberate practice, ultimately increasing chances of success. In tandem, intrinsic and identified regulations should help people to achieve their goals and, happily, to feel good in the process.

References Anderson, N. (1968). Likeableness ratings of 555 personality-trait words. Journal of Personality and Social Psychology, 9, 272–279. Austin, J., & Vancouver, J. (1996). Goal constructs in psychology: Structure, process, and content. Psychological Bulletin, 120, 338 –375. Bargh, J., & Chartrand, T. (1999). The unbearable automaticity of being. American Psychologist, 54, 462– 479. Bargh, J., & Chartrand, T. (2000). Studying the mind in the middle: A practical guide to priming and automaticity research. In H. Reis & C. Judd (Eds.), Handbook of research methods in social psychology (pp. 253–285). New York: Cambridge University Press. Baumeister, R., Bratslavsky, E., Muraven, M., & Tice, D. (1998). Ego depletion: Is the active self a limited resource? Journal of Personality and Social Psychology, 74, 1252–1265. Black, A., & Deci, E. (2000). The effects of instructors’ autonomy support and students’ autonomous motivation on learning organic chemistry: A self-determination theory perspective. Science Education, 84, 740 –756. Blais, M., Vallerand, R., Brie`re, N., Gagnon, A., & Pelletier, L. (1990). Significance, structure, and gender differences in life domains of college students. Sex Roles, 22, 199 –212. Cordova, D., & Lepper, M. (1996). Intrinsic motivation and the process of learning: Beneficial effects of contextualization, personalization, and choice. Journal of Educational Psychology, 88, 715–730. Deci, E., & Ryan, R. (2000). The “what” and “why” of goal pursuits: Human needs and the self-determination of behavior. Psychological Inquiry, 11, 227–268. Diener, E., Emmons, R., Larsen, R., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 47, 1105–1117. Diener, E., & Seligman, M. (2004). Beyond money: Toward an economy of well-being. Psychological Science in the Public Interest, 5, 1–31. Dweck, C. S. (1999). Self-theories: Their role in motivation, personality, and development. Philadelphia: Taylor and Francis/Psychology Press. Fazio, R. (1990). A practical guide to the use of response latency in social psychological research. In C. Hendriks & M. Clark (Eds.), Research methods in personality and social psychology (pp. 74 –97). Newbury Park, CA: Sage. Fazio, R., & Zanna, M. (1981). Direct experience and attitude-behavior consistency. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 14, pp. 161–202). San Diego, CA: Academic Press. Gagne´, M., & Deci, E. (2005). Self-determination theory and work motivation. Journal of Organizational Behavior, 26, 331–362. Grolnick, W., & Ryan, R. (1987). Autonomy in children’s learning: An experimental and individual difference investigation. Journal of Personality and Social Psychology, 52, 890 – 898. Grolnick, W., Ryan, R., & Deci, E. (1991). Inner resources for school achievement: Motivational mediators of children’s perceptions of their parents. Journal of Educational Psychology, 83, 508 –517. Koestner, R., & Losier, G. (2002). Distinguishing three ways of being internally motivated: A closer look at introjection, identification, and intrinsic motivation. In E. Deci & R. Ryan (Eds.), Handbook of selfdetermination research (pp. 101–121). Rochester, NY: University of Rochester Press. Koestner, R., Losier, G., Vallerand, R., & Carducci, D. (1996). Identified

762

BURTON, LYDON, D’ALESSANDRO, AND KOESTNER

and introjected forms of political internalization: Extending selfdetermination theory. Journal of Personality and Social Psychology, 70, 1025–1036. Laurent, J., Catanzaro, S., Joiner, T., Jr., Rudolph, K., Potter, K., Lambert, S., et al. (1999). A measure of positive and negative affect for children: Scale development and preliminary validation. Psychological Assessment, 11, 326 –338. Levesque, C., & Pelletier, L. (2003). On the investigation of primed and chronic autonomous and heteronomous motivational orientations. Personality and Social Psychology Bulletin, 29, 1570 –1584. Lydon, J., Burton, K., & Menzies-Toman, D. (2005). Commitment calibration with the relationship cognition toolbox. In M. Baldwin (Ed.), Handbook of interpersonal cognition (pp. 126 –152). New York: Guilford. Lydon, J., & Zanna, M. (1990). Commitment in the face of adversity: A value-affirmation approach. Journal of Personality and Social Psychology, 58, 1040 –1047. McClelland, D., Koestner, R., & Weinberger, J. (1989). How do selfattributed and implicit motives differ? Psychological Review, 96, 690 – 702. Mikulincer, M., Birnbaum, G., Woddis, D., & Nachmias, O. (2000). Stress and accessibility of proximity-related thoughts: Exploring the normative and intraindividual components of attachment theory. Journal of Personality and Social Psychology, 78, 509 –523. Miserandino, M. (1996). Children who do well in school: Individual differences in perceived competence and autonomy in above-average children. Journal of Educational Psychology, 88, 203–214. Nix, G., Ryan, R., Manly, J., & Deci, E. (1999). Revitalization through self-regulation: The effects of autonomous and controlled motivation on happiness and vitality. Journal of Experimental Social Psychology, 35, 266 –284. Paquin, C. (2005). The impact of communication on the performance of sports teams. Unpublished master’s thesis, University of Que´bec at Trois-Rivie`res, Trois-Rivie`res, Que´bec, Canada. Psychology Software Tools. (n.d.). E-Prime (Version 1) [Computer software]. Pittsburgh, PA: Author. Ratelle, C., Baldwin, M., & Vallerand, R. (2005). On the cued activation of situational motivation. Journal of Experimental Social Psychology, 41, 482– 487. Reis, H., Sheldon, K., Gable, S., Roscoe, J., & Ryan, R. (2000). Daily well-being: The role of autonomy, competence, and relatedness. Personality and Social Psychology Bulletin, 26, 419 – 435. Ryan, R. (1995). Psychological needs and the facilitation of integrative processes. Journal of Personality, 63, 397– 427. Ryan, R., & Connell, J. (1989). Perceived locus of causality and internal-

ization: Examining reasons for acting in two domains. Journal of Personality and Social Psychology, 57, 749 –761. Ryan, R., & Deci, E. (2003). On assimilating identities into the self: A self-determination theory perspective on internalization and integrity within cultures. In M. Leary & J. Price Tangney (Eds.), Handbook of self and identity (pp. 253–272). New York: Guilford Press. Ryan, R., & Frederick, C. (1997). On energy, personality, and health: Subjective vitality as a dynamic reflection of well-being. Journal of Personality, 65, 529 –565. Salancik, J. R. (1974). Inference of one’s attitude from behavior recalled under linguistically cognitive sets. Journal of Experimental Social Psychology, 10, 415– 427. Sheldon, K., & Elliot, A. (1999). Goal striving, need satisfaction, and longitudinal well-being: The self-concordance model. Journal of Personality and Social Psychology, 76, 482– 497. Sheldon, K., & Kasser, T. (1995). Coherence and congruence: Two aspects of personality integration. Journal of Personality and Social Psychology, 68, 531–543. Sheldon, K., & Kasser, T. (1998). Pursuing personal goals: Skills enable progress, but not all progress is beneficial. Personality and Social Psychology Bulletin, 24, 1319 –1331. Sheldon, K., Kasser, T., Smith, K., & Share, T. (2002). Personal goals and psychological growth: Testing an intervention to enhance goal attainment and personality integration. Journal of Personality, 70, 5–31. Vallerand, R., & Bissonnette, R. (1992). Intrinsic, extrinsic, and amotivational styles as predictors of behavior: A prospective study. Journal of Personality, 60, 599 – 620. Vallerand, R., Fortier, M., & Guay, F. (1997). Self-determination and persistence in a real-life setting: Toward a motivational model of high school dropout. Journal of Personality and Social Psychology, 72, 1161–1176. Vansteenkiste, M., Simons, J., Lens, W., Sheldon, K., & Deci, E. (2004). Motivating learning, performance, and persistence: The synergistic effects of intrinsic goal contents and autonomy-supportive contexts. Journal of Personality and Social Psychology, 87, 246 –260. Watson, D., Clark, L., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Wong, M., & Csikszentmihalyi, M. (1991). Motivation and academic achievement: The effects of personality traits and the quality of experience. Journal of Personality, 59, 539 –574.

Received December 17, 2004 Revision received May 15, 2006 Accepted May 22, 2006 䡲

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 763–779

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.763

A First Large Cohort Study of Personality Trait Stability Over the 40 Years Between Elementary School and Midlife Sarah E. Hampson

Lewis R. Goldberg

University of Surrey and Oregon Research Institute

Oregon Research Institute

This report provides some initial findings from an investigation of the relations between childhood Big Five personality traits assessed by elementary school teachers and similar traits assessed 40 years later by self-reports at midlife (N ⫽ 799). Short-term (1–3 years) test–retest reliabilities were lower (.22–.53) in childhood when personality was developing than they were in adulthood (.70 –.79) when personality stability should be at its peak. Stability coefficients across the 40-year interval between the childhood assessment and the 2 measures of adulthood personality were higher for Extraversion (e.g., .29) and Conscientiousness (e.g., .25) than for Openness (e.g., .16), Agreeableness (e.g., .08), and Neuroticism (e.g., .00). Construct continuity between childhood and adulthood was evaluated by canonical analysis and by structural equation modeling and indicated continuity at both a broad, two-dimensional level and at the level of the Big Five. The findings are discussed in relation to A. Caspi, B. W. Roberts, and R. L. Shiner’s (2005) principles of rank-order personality stability. Keywords: personality stability, Big Five, construct continuity, longitudinal study

M. Digman and formed the basis for some of the pioneering work on the five factor structure of personality (e.g., Digman, 1989). The adult assessments were undertaken over a 4-year period (1998 –2002) after an interval of 35– 43 years (Hampson et al., 2001). Consequently, these data offer an extraordinary opportunity to examine the question of personality stability from childhood to midlife with measures of the five factor model of personality at each time point. We have been unable to find a report of any other longitudinal personality data with a comparable convergence of the key features of our study: the first time of assessment during childhood and the second time in adulthood, spanning a follow-up period of around 40 years between childhood and adulthood, with the same personality factors assessed at both times of measurement. Two goals motivated these analyses: The first is to contribute to the continuing scientific debate about the stability of personality traits over the lifetime. Our contribution provides a unique addition to this debate because of the portion of the life span bridged by the two assessments. A second reason for studying the stability of personality is in the service of another but related purpose. When a child personality predictor of an adult outcome is identified, the pathway from child trait to adult outcome demands investigation. For example, Friedman and his colleagues (e.g., Friedman et al., 1993, 1995) demonstrated that childhood Conscientiousness is predictive of longevity. This provocative finding stimulated us to embark on a longitudinal study of the Hawaii sample, ultimately to attempt replication and, in the shorter term, to test models about the mechanisms and processes involving personality traits that create the pathways from child predictors to adult outcomes (Hampson, Goldberg, Vogt, & Dubanoski, 2006, in press). An evaluation of the stability of personality traits in the Hawaii sample guides the development of these models. There are several ways in which personality stability can be defined and studied (Caspi & Roberts, 1999, 2001). Interindividual

The extent to which child personality is predictive of adult personality has long been a fundamental scientific, philosophical, and even poetic question. Wordsworth’s “The child is father of the man” and Milton’s “The child shows the man, as morning shows the day” are not unambiguous. Neither poet was necessarily suggesting perfect correspondence and complete predictability from youth to maturity. Similarly, the stability of personality characteristics over the life course remains a controversial empirical issue. Proponents of stability debate the findings with advocates of transformation and change (Lewis, 2001; McCrae et al., 2000). In this article, we provide some initial findings from an investigation of the relations between childhood personality traits assessed by elementary school teachers and similar traits assessed by self-reports at midlife. Our analyses are based on data from the Hawaii Personality and Health Cohort, a relatively large and culturally diverse sample that was first assessed between 1959 and 1967 when the participants were children in Hawaii. The childhood assessments were designed and supervised by the late John

Sarah E. Hampson, Department of Psychology, University of Surrey, Guildford, United Kingdom, and Oregon Research Institute; Lewis R. Goldberg, Oregon Research Institute. Support for this project was provided by Grant AG20048 from the National Institute on Aging and by Grant MH55600 from the National Institute of Mental Health. We are indebted to John M. Digman (1923– 1998) for obtaining the childhood personality assessments. Thomas M. Vogt and Joan P. Dubanoski are key members of the current research team investigating the Hawaii Personality and Health cohort, and we thank them for their many contributions to the project. We also thank Brent W. Roberts and Rebecca L. Shiner for their comments on an earlier version of this article. Correspondence concerning this article should be addressed to either Sarah E. Hampson or Lewis R. Goldberg at Oregon Research Institute, 1715 Franklin Boulevard, Eugene, OR 97403-1983. E-mail: [email protected] or [email protected] 763

764

HAMPSON AND GOLDBERG

stability, called differential continuity by Caspi and Roberts (2001), refers to the rank-order stability of members of the same sample over time, and typically it has been evaluated by using correlation coefficients (e.g., Conley, 1984; Roberts & DelVecchio, 2000). Another approach to measuring personality stability has been called absolute continuity (Caspi & Roberts, 2001). The mean levels of traits can be compared at different ages, either with different cohorts in cross-sectional designs (e.g., Conley, 1985; Costa & McCrae, 1988) or with the same cohort at different ages in longitudinal studies (e.g., Roberts & Chapman, 2000). Such studies are used to infer the extent to which personality traits increase or decrease with age. The longitudinal data available for the Hawaii cohort lend themselves to the evaluation of rank-order consistency between childhood and adulthood. An issue for all longitudinal studies that span development from childhood to adulthood is that of construct continuity from one time point to another (e.g., Caspi & Silva, 1995). Scientific advances in personality theory and measurement may result in a lack of comparability between the measures selected at the two times of assessment. Moreover, the very process of personality development from childhood to adulthood may result in construct discontinuity (e.g., Kagan, 1980); at the extreme, the evolution and transformation of personality constructs might be such that equivalent measures are not theoretically possible. Even if the behavioral indicators of the construct at each time point are internally consistent and have high short-term test–retest stability, they need not be identical constructs at the two age points (Asendorpf, 1992b). For example, with infants and toddlers the childhood measures have typically focused on temperament variables, for which there may be no directly comparable adult counterparts, and consequently rank-order consistency cannot be evaluated directly. The potential discontinuity of any given construct has important implications for analyses of personality stability. For example, is it appropriate to attempt to relate measures of the five factor personality trait structure in childhood with measures of the five factor structure at midlife? Recently Measelle et al. (2005) identified Extraversion, Agreeableness, and Conscientiousness in children as young as 5 years of age by using a puppet interview. Other studies suggest that a five factor structure can be derived from teachers’ ratings of children as young as 7 years of age but perhaps not from ratings of children aged 4 to 6 years, for whom only four broader factors may be more appropriate (Mervielde, Buyst, & De Fruyt, 1995). Indeed, Shiner and Caspi (2003) advocated a four factor structure for middle childhood. However, a five factor structure corresponding to the adult Big Five (i.e., the “Little Five”) was identified by John, Caspi, Robins, Moffitt, and Stouthamer-Loeber (1994) in young adolescents. And Digman (1989; Digman & Inouye, 1986; Digman & Takemoto-Chock, 1981) found clear versions of the Big Five factors by using small portions of the very data here under study; later, Digman and Shmelyov (1996) found a clear five factor structure in teacher ratings of Russian school children in Grades 1, 2, and 3. Together, these studies suggest that the five factor model of personality can be used for the assessment of both children and adults to determine rank-order stability. In the present study, the childhood assessments were based on variables that have been shown to conform to the five factor model of personality (Goldberg, 2001). The personality assessments in adulthood were also based on the five factor model. The comparability of constructs at the two times of assessment should min-

imize construct discontinuity. However, there is still a great deal of difference between the behavioral indicators of the Big Five in children as used by teachers in their assessments versus the indicators used by adults at midlife to arrive at their self-descriptions. That is, although the personality constructs assessed at each time point in our study might be viewed as broadly equivalent, the method of assessment cannot be viewed as such. The childhood assessments were based on teachers’ perceptions of the children in their classrooms. We have argued elsewhere that teacher assessments are the ideal method for measuring the personality traits of children too young to provide self-reports (Goldberg, 2001). In brief, teachers are familiar with the child across a variety of classroom and other settings (e.g., recess), and they are in a perfect position to make normative judgments because of their extensive experience with children of the same age. In contrast to these teacher assessments, all of the adult measures analyzed in this article were based on self-reports. Although we intend to expand the adult personality battery to include peer and observer descriptions, at this stage of the project we need not apologize for the use of self-report methodology, given the near ubiquity of such measures in the assessment of adult personality traits. The degree of convergence between the differing perspectives on personality provided by observer ratings versus self-report is itself as hotly debated as is the stability of personality over time (e.g., Funder, 1999; John & Robins, 1993; Kenny, 1994). Whatever the eventual outcome of this debate, the difference in perspective at the two times of our assessments may place a limit on the level of stability that can be obtained. Given that this may be the first large-scale long-term study of rank-order personality stability, we must turn to studies of shorter age ranges to set our initial expectations (e.g., Roberts & DelVecchio, 2000). An examination of the relevant studies of rank-order stability reviewed by Roberts and DelVecchio (2000, Table 1) provides context for the present investigation. Those authors identified 34 studies in which the age at first assessment was between 6 and 12 years, which corresponds to the ages of our child participants at their first assessment. In 24 of these studies, the method of personality assessment was observer ratings; however, there were few studies (not including that of Digman, 1989, which is based on a portion of the data reported here) in which teacher assessments were used. Rubin, Hymel, and Mills (1989) used both parent and teacher ratings of sociability and social withdrawal in young children; social withdrawal observed at kindergarten and at Grade 2 (N ⫽ 52) were significantly correlated (r ⫽ .37, p ⬍ .01), but the correlation (not given) for sociability was not significant. Backteman and Magnusson (1981) used teachers’ ratings of aggressiveness, motor disturbance, timidity, disharmony, distraction, and lack of school motivation in their study of 858 children. They examined stability from age 10 to 13 years and obtained a mean stability coefficient for boys of .52 and for girls of .48. In both of these studies, stability was examined over a brief interval within childhood, which gives an indication of the level of test–retest stability to be expected within childhood for our cohort. In this article, we provide another and somewhat richer examination of the longitudinal stability of teacher assessments. Although there are no comparable longitudinal studies of rankorder stability that bridge the entire interval between middle childhood and middle age, there are studies that evaluate personality trait continuity over intervals within adulthood as long or longer

PERSONALITY STABILITY

than the roughly 40-year time span of our present investigation. Two such studies (Haan, Millsap, & Hartka, 1986; Soldz & Vaillant, 1999) are remarkable for the time span that they encompassed. For 118 participants in the Oakland Growth Study, from adolescence to mid–late adulthood, a period of approximately 50 years, Haan et al. (1986) reported the following rank-order correlations for personality constructs assessed by observer Q-sorts: outgoing versus aloof ⫽ .37, cognitively committed ⫽ .34, selfconfident vs. victimized ⫽ .26, dependable ⫽ .25, assertive versus submissive ⫽ .24, and warm versus hostile ⫽ .14. To the extent that the first two variables are associated with Extraversion and Intellect, respectively, this suggests that those factors may be more stable than the other three of the Big Five. Soldz and Vaillant (1999) examined the stability of the dimensions of the five factor model of personality in their analysis of 163 participants in the Grant Study. These men were rated at the end of college on 25 personality traits by a psychiatrist, and they completed the NEO personality inventory at age 67– 68 years. Scores on the five personality dimensions assessed by the NEO were estimated from the college trait ratings with the help of experts who assessed the relevance of each of the 25 traits to each of the five factors. The stability coefficients across the 30-year interval for the five domains were: Openness to Experience ⫽ .38, Neuroticism ⫽ .20, Extraversion ⫽ .19, Conscientiousness ⫽ .12, and Agreeableness ⫽ .07. The unusually high stability correlation for Openness may reflect its association with intellect/imagination and the possibility that cognitive variables may be more stable than other types of personality traits. More generally, in using previous studies to guide our predictions regarding personality stability in the Hawaii cohort, there are several replicated findings across studies of rank-order personality trait stability that have now become enshrined as general principles (Asendorpf, 1992a; Fraley & Roberts, 2005; Roberts & DelVecchio, 2000). The following six such principles were identified recently by Caspi, Roberts, and Shiner (2005): 1.

As a fundamental rule, the longer the interval between assessments, the lower the level of rank-order stability.

2.

More specifically, rank-order stability increases with age.

3.

Rank-order stability reaches a plateau between the ages of 50 to 70 years.

4.

Rank-order stability does not seem to vary markedly by assessment method.

5.

Rank-order stability does not seem to vary markedly by gender.

6.

Rank-order stability does not seem to vary markedly among the Big Five personality factors.

On the basis of the first principle, higher short-term stability was expected when examining relatively brief test–retest intervals within childhood and within adulthood than when examining longterm stability from childhood to adulthood. Moreover, we predicted that the rank-order stability of traits from childhood to midlife would be lower than that found in previous studies such as Haan et al. (1986) and Soldz and Vaillant (1999). This is because

765

for the Hawaii cohort, personality trait assessments were available when the participants were in the midst of their childhood development, middle childhood (Shiner, 1998), and subsequently when they were in their 4th or 5th decade, when trait stability should be maximized (Principles 2 and 3). On the basis of Principles 5 and 6, we also predicted that rank-order stability would not differ substantially by gender or among the Big Five traits. Finally, given that previous studies had demonstrated that the childhood data conformed to the Big Five trait structure (Digman & Inouye, 1986; Goldberg, 2001), we predicted that we would demonstrate construct continuity between the child and adult versions of these latent traits.

Method Child Participants and Procedures The original child cohort included 2,404 elementary school children comprising six samples obtained by John M. Digman between 1959 and 1967. These samples, which are described in detail in Goldberg (2001), can be divided into two sets on the basis of their relative sizes. Each of the first three samples was reasonably large, and the personality characteristics assessed in these samples were all unipolar in their format (e.g., Gregarious). In contrast, each of the remaining three samples, all drawn from the Laboratory School of the University of Hawaii, was relatively small in size, and the variables were in a bipolar format (e.g., Gregarious vs. Solitary); some of the Lab School children were assessed more than once by different teachers in different elementary school years. The following descriptions of each of the six samples have been adapted from Goldberg (2001). Oahu: Grades 1 and 2 (N ⫽ 885). In 1965, each of 885 children from eight schools on the island of Oahu were assessed by one of 29 teachers on 49 personality attributes. Each attribute consisted of a single word or short phrase, followed by a more extensive definition (e.g., Energetic: Active; full of pep; vigorous; movements are quick, darting). Oahu: Grades 5 and 6 (N ⫽ 834). During the same year, each of 834 older children from the same eight schools were assessed by one of 28 teachers by using the identical set of 49 variables. Kauai: Grade 6 (N ⫽ 502). In 1967, each of 502 sixth-grade children from eight schools on the island of Kauai were assessed by one of 17 teachers on 43 personality attributes. Again, each attribute consisted of a single word or short phrase, followed by a more extensive definition; 39 of these variables had been included in the two Oahu samples. Lab ’59: Grades 1 and 2 (N ⫽ 102). The remaining three partially overlapping samples were all drawn from the University of Hawaii Laboratory School. Four teachers assessed each of the children in their classrooms using the bipolar variables previously used by Cattell and Coan (1957). Lab ’60: Grades 1, 2, and 3 (N ⫽ 149). Of the 102 children assessed in 1959, 93 were reassessed a year later, along with 56 new students, mostly first graders. Two 1st-grade, two 2nd-grade, and two 3rd-grade teachers provided these assessments. The set of 50 bipolar variables used with this sample includes variants of most of those used in the 1959 sample plus 20 new ones devised by Digman. Lab ’63: Grades 5 and 6 (N ⫽ 100). This sample of fifth and sixth graders included 69 of those assessed in 1959 and 73 of those assessed in 1960; of the 100 children, 67 were assessed on all three occasions. Four teachers provided the 1963 child descriptions. Again, Digman modified many of the variables that he had used previously and added a number of new ones, resulting in an enlarged set of 63 bipolar variables. For present purposes, the six child samples were aggregated into two sets: (a) the 2,221 children, originally assessed on a single occasion in Oahu or Kauai schools and (b) those Laboratory School children who were assessed on more than one occasion. The former sample is here used to

766

HAMPSON AND GOLDBERG

assess childhood-to-adulthood personality trait stability, and the latter is used solely to estimate the stability of teacher assessments during childhood.

General Procedures Common to All of the Child Samples The teachers were provided with the names of the students in their classes and with sheets of paper or cardboard, on each of which a particular variable label was printed at the top; the teachers were instructed to rank their students from the highest to the lowest on that attribute. Rectangular boxes were included on each sheet to be used by the teacher for recording the students’ names. All data were collected one attribute at a time, with a fixed nine-step quasi-normal distribution used for each classroom. Teachers were instructed to alternate between the two poles of the distribution in their rankings. The procedure was identical to that of a typical Q-sort, except that individuals, rather than attributes, were ranked (sorted). The resulting data distributions are therefore symmetric and quasi-normal, with very comparable means and standard deviations across both personality variables and classrooms. For the Kauai and the two Oahu samples, each personality attribute was accompanied by a definition of the characteristic— definitions that had been developed beforehand from focus groups of teachers who had been asked to provide typical examples of classroom behaviors relating to that concept.

(SMM), which are a subset of terms from Goldberg’s (1992) 100 unipolar markers of the Big Five factor structure that have been found to be the most univocally associated with the same factor across diverse subject samples; two adjectives tapping self-perceived physical attractiveness; and adult versions of 42 of the 49 childhood variables (6 childhood variables overlapped with 6 SMM variables, and the childhood variable “selfminimizing” was omitted because of its unfamiliarity and replaced with “self-centered,” reverse scored). The childhood variables included the 39 variables that were common to the teacher assessments in Oahu and Kauai. Omitted from the adult questionnaire were all of the defining child behavioral indicators that had been listed under each personality trait term in the teacher assessments. The combined set of 84 trait-descriptive adjectives, listed in alphabetical order, was administered with a 5-point response scale, ranging from 1 (very false) to 5 (very true) as self-descriptors.

Representativeness of the Adult Sample to the Child Cohort

Since July of 1998, our research team has been attempting to locate each of the roughly 2,000 now-adult living members of the childhood cohort and to recruit as many of them as possible into a study of adult health behaviors and outcomes. The details of our location and recruitment procedures are described in Hampson et al. (2001). As the adult participants are recruited for the project, they are periodically mailed a series of questionnaires, and they are invited to attend a half-day session at a medical setting where they complete an extensive battery of physical, medical, personality, and cognitive measures. The present analyses are based on the roughly 400 men and 400 women who completed the first two of our adult questionnaires before the end of 2004. For most of these participants, the interval of time between the teacher assessments during childhood and the completion of the adult questionnaires was approximately 40 years.

To evaluate the extent to which the adult participants might constitute a biased or truncated sample of the original child cohort, we examined the means and standard deviations of their childhood Big Five factor scores, which are provided in Table 1. The values in this table are based on the child factor scores from the total Oahu and Kauai samples (i.e., the Laboratory School samples are excluded), computed separately within the subsamples of boys and girls; thus, if the adult men and women subsamples were completely unbiased, they would have means of zero; and if the adult subsamples were completely unrestricted in range, they would have standard deviations of 1.00. Clearly, there is virtually no range restriction for any of the adult samples on any of the child factors, with the standard deviations ranging from .98 to 1.03. Nor is bias likely to be much of a problem, with the possible exception of Conscientiousness, on which, predictably, the adult samples are about one-tenth of a standard deviation higher than the cohort mean. In evaluating this small effect, one must keep in mind that it is probably at least partially artifactual because we know from previous studies (e.g., Friedman et al., 1995) and from prior analyses of our own cohort (Hampson et al., 2001) that early mortality is associated with low childhood Conscientiousness; thus, at least some of the children who were seen by their teachers as low in Conscientiousness are unlikely to be available for adult recruitment.

Measures of Adult Personality Traits

Analyses

Adult Participants and Procedures

Our first questionnaire was a 16-page survey of demographic variables and health-related behaviors (e.g., smoking, drinking, diet, exercise); included in this questionnaire were the 44 items in the Big Five Inventory (BFI; John & Srivastava, 1999), which is perhaps the best of the various brief sets of five factor markers. John developed each of the five BFI scales to fall roughly halfway between the lexical Big Five factors (Goldberg, 1992) and the five domain scores from the NEO-Personality Inventory— Revised (NEO-PR-R; Costa & McCrae, 1992), and consequently he named his inventory with the lexical expression and used the NEO labels for the scales: Extraversion, Agreeableness, Conscientiousness, Neuroticism, and Openness to Experience. The BFI items were administered with a 5-point response scale, ranging from 1 (very inaccurate) to 5 (very accurate) as self descriptors; the BFI items can be found in Appendix A. For the sake of consistency, we use the BFI labels for the Big Five traits throughout the remainder of this article. Our second questionnaire was administered between 2 and 4 years after the first; the mean length of time between completion of the surveys was 2.8 years. This two-sided single-page questionnaire focused entirely on personality characteristics. One side comprised 46 items: a readministration of the 44 BFI items plus two items in the same format that assessed self-perceived physical attractiveness. The other side comprised 84 items: the 40 personality trait adjectives from Saucier’s (1994) Mini-Markers

We begin our analyses with the three Laboratory School samples, where we have access to test–retest teacher assessments for 1-, 3-, and 4-year intervals. We then turn to the adult participants, for whom we also have short-term (2- to 4-year) test–retest administrations of the BFI. Next we present Big Five factor structures in childhood and adulthood, prior to examining the extent of association between the two sets of personality factors. In addition to rank-order correlations, we evaluate the long-term stability and construct validity of personality traits by using canonical analysis and structural equation modeling. In this article, we focus on findings for the Big Five factors. We have also conducted stability analyses with the middle-level trait clusters and individual items. These, and analyses conducted separately by gender, are available from Sarah E. Hampson and Lewis R. Goldberg.

Results Short-Term Retest Stability of the Childhood Teacher Assessments Table 2 provides the stability correlations from the three Laboratory School samples, for each of the Big Five factors. Stability correlations over 1 year ranged from .52 for Extraversion to .28 for

PERSONALITY STABILITY

767

Table 1 Representativeness of the Original Child Cohort in the Adult Samples: Mean Childhood Factor Scores in Each of the Adult Samples Childhood factor scores (Ms) Male

SDs

Female

Male

Female

Factor

n ⫽ 396

n ⫽ 375

n ⫽ 403

n ⫽ 387

n ⫽ 396

n ⫽ 375

n ⫽ 403

n ⫽ 387

Extraversion Agreeableness Conscientiousness Neuroticism Openness

.02 .05 .09 .07 .05

.03 .06 .10 .06 .05

⫺.04 .06 .13 .00 .08

⫺.04 .07 .12 .00 .09

1.03 0.97 1.04 0.98 0.98

1.03 0.97 1.03 0.97 0.98

1.03 0.98 0.99 1.02 1.00

1.03 0.98 0.98 1.02 1.01

Note. The mean values are deviations (in standard deviation units) for each of the adult samples from the childhood cohort means (0.00), which were computed separately within the total childhood male and female samples; the sample standard deviations can be compared with the total childhood male and female cohort standard deviation of 1.00.

Neuroticism; over 3 years they ranged from .43 for Agreeableness to .22 for Neuroticism; and over 4 years they ranged from .55 for Openness to .36 for Extraversion. One provocative finding from these analyses suggests that the five factors may have different patterns of short-term stability: Agreeableness and Extraversion conformed to the predicted pattern of a linear decrease in stability over the length of the time interval, whereas, paradoxically, for the other three factors stability was higher over 4 years than over 3. The short-term stability of Neuroticism was lower than that of the other four factors.

Short-Term Retest Stability of One of the Adult Self-Report Measures Table 3 provides evidence concerning the stability of the BFI scales over an approximately 3-year period of middle adulthood. The stability correlations were somewhat higher for Extraversion and Openness than for the other three factors. The magnitude of these test–retest correlations was considerably higher than the short-term stability of the childhood assessments.

The Childhood Variables and Big Five Factors

ples, ordered by their factor loadings on five varimax rotated factors, when these factors were derived in the sample (N ⫽ 799) for whom adult personality measures are available. Although the content of the Big Five factors is clearly apparent, the varimax solution is far less simply structured than one finds in analyses of adult personality variables. As evidence of the factorial complexity of these childhood variables, almost three-quarters (29 of the 39) of them have factor loadings of .30 or higher on two or more factors, and some of the secondary loadings are quite substantial in size. Indeed, a number of variables that help to define the Neuroticism factor (e.g., nervous, touchy, complains about others) have somewhat larger loadings on other factors.

Two Strategies for Assessing Big Five Personality Factors in Adulthood Appendix B provides the varimax rotated six-factor solution from an analysis including all 130 variables in the second of the two adult questionnaires. The content of the Big Five factors plus a Physical Attractiveness factor is strikingly clear, with only about one-quarter (36 of 130) of the variables having secondary loadings of .30 or higher (as compared with three-quarters for the childhood

Appendix A provides a listing of the 39 variables that were common to the teacher assessments in the Oahu and Kauai samTable 2 Stability (r) Across Years of the Teacher Assessments for the Big Five Factor Scores in the University of Hawaii Laboratory School Samples Laboratory School samples 1959 vs. 1960a 1960 vs. 1963b 1959 vs. 1963c

Factor Extraversion Agreeableness Conscientiousness Neuroticism Openness M a

N ⫽ 93.

b

N ⫽ 73.

.52 .51 .53 .28 .49 .47 c

N ⫽ 69.

.38 .43 .31 .22 .33 .33

.36 .45 .41 .38 .55 .43

M .42 .46 .42 .29 .46

Table 3 Internal Consistency (Coefficient Alpha) and Test–Retest Stability of Each of the Five BFI Scales BFI scale Extraversion Agreeableness Conscientiousness Neuroticism Openness M

No. of items 8 9 9 8 10

␣1

␣2

Retest r

Corrected r

.84 .78 .80 .82 .79 .81

.84 .80 .81 .85 .83 .83

.79 .70 .70 .71 .79 .74

.94 .87 .87 .85 .98 .90

Note. N ⫽ 799. The two values of coefficient alpha are based on the first (␣1) and the second (␣2) administrations of the Big Five Inventory (BFI) items. The test–retest correlations are corrected for the unreliabilities at each of the two administrations. The average length of the test–retest interval was approximately 1,000 days (2.8 years).

HAMPSON AND GOLDBERG

768

variables). The Big Five factor scores from this six-factor structure were used as one of our two sets of adult personality criteria. One advantage of the Big Five factor scores is that they were derived from a sizeable set of variables; a disadvantage is that all of these variables had been administered on the same occasion. To provide another set of criteria at the adult level, we used factor scores from an analysis of the 44 ⫻ 2 ⫽ 88 BFI items that had been administered on two occasions, roughly 3 years apart. These latter factors have the advantage of correcting for short-term fluctuations in adulthood for each of the Big Five traits. The correlations between the two sets of criterion factors (i.e., Big Five factor scores and BFI factor scores) are as follows: Extraversion ⫽ .91, Agreeableness ⫽ .86, Conscientiousness ⫽ .88, Neuroticism ⫽ .84, and Openness ⫽ .91.

Personality Stability Across the 40-Year Period Table 4 provides the correlations between the corresponding childhood and adult Big Five personality factors for each of the two sets of adult criterion factors. In separate analyses for children assessed in Grades 1 and 2 compared with those assessed in Grades 5 and 6, we found no differences in the patterns of correlations. The 40-year stability correlations for each adult criterion were highest for Extraversion and Conscientiousness, were around zero for Neuroticism, and were intermediate for Openness. In this table, the Big Five domains are ordered by their relative stability from Extraversion (most stable) to Neuroticism (least stable). For both procedures for calculating adult factors, we examined the statistical significance of the differences in stability correlations among all pairs of factors. When ordered by size of the stability correlation (as in Table 4) all differences in correlations between adjacent factors (E vs. C; C vs. O; O vs. A; and A vs. N) are not statistically significant, whereas all other pairs of such differences are significant. Indeed, all pairs of factors that are two steps removed from each other (E vs. O; C vs. A; and O vs. N) differ at least at the .01 level of significance, whereas all other pairs of factors (E vs. A; E vs. N; C vs. N) differ well beyond the

Table 4 Correlations Between the Corresponding Childhood and Adult Personality Factors (r) and Significant Multiple Correlations Factor Extraversion Conscientiousness Openness Agreeableness Neuroticism

r

K

R

F

df

p

.27** .29** .25** .23** .17** .16** .09* .08* .00 .00

3 3

.38 .39

25.22 36.03

5, 756 4, 794

.00 .00

2

.28

15.82

4, 757

.00

Note. The values appearing in the top of each cell are based on the adult factors derived from 130 variables (Big Five factor scores), and the values appearing in the bottom of each cell are based on the adult factors derived from the two administrations of the Big Five Inventory factor scores. K ⫽ Number of childhood factors included in the prediction. * p ⬍ .05. ** p ⬍ .01.

.0001 level of significance. Clearly, then, these differences in factor stability correlations are likely to be highly reliable within this sample. Table 4 also presents the findings from the stepwise multipleregression analyses in which the five childhood factors were used to predict each of the adult criterion. In these analyses, the Gender ⫻ Childhood factor interactions were also tested for the significant predictors. The multiple correlations are reported only when more than one childhood personality factor predicted the adult criterion. Only two of the adult factors were predicted by a combination of childhood factors, and only two gender interactions were significant. The three-variable combination of childhood Extraversion (␤s ⫽ .29 and .31), Disagreeableness (␤s ⫽ .13 and .11), and lack of Conscientiousness (␤s ⫽ .24 and .14) predicted adult Big Five Extraversion and BFI Extraversion, respectively. Lack of Conscientiousness was significant for men but not for women for the Big Five Extraversion criterion. The two-variable combination of childhood Agreeableness (␤ ⫽ .08) and, for men but not for women, Extraversion (␤ ⫽ .10) predicted the adult Big Five Agreeableness factor. Childhood Agreeableness predicted BFI Agreeableness only for men (␤ ⫽ .15).

Canonical Correlations Between the Two Sets of 39 Childhood and Adult Variables What is the maximum possible correlation between the childhood personality variables and their adult counterparts? Perhaps the most straightforward procedure for answering this question is based on optimizing the correlation between sets of variables by using canonical correlation analysis. In our context, each of the 39 childhood variables and each of the 39 adult variables were separately weighted so as to optimize the correlation between those two sets. Each such canonical variate provides an optimal weighting of the variables within each set, after the variance in the preceding canonical variate has been residualized; thus, the intercorrelations among all canonical variates are zero. In our case, only the first two canonical correlations were statistically significant: The first canonical variate had a correlation of .47, which, when corrected for capitalization on chance, corrects to a value of .35; for the second variate, the correlation was .43, which corrects to a value of .30. Those variables with the highest correlations with these two orthogonal factors are presented in Table 5. The largest of the two canonical variates combines variables from the Conscientiousness, Agreeableness, and Neuroticism domains, linking childhood classroom disruptiveness (e.g., restless, careless with others’ property, fidgety, impulsive, and irresponsible vs. mannerly and careful of personal belongings) with adult undependability (e.g., irresponsible and impulsive versus conscientious and neat in appearance). The second canonical variate links childhood Extraversion traits (e.g., outspoken, socially confident, verbally fluent, and assertive versus submissive) with some adult Extravert counterparts (e.g., verbally fluent, gregarious, outspoken, assertive versus fearful).

Latent Modeling of Construct Continuity Up to this point, all of our findings have been based on rather traditional analytic procedures, typically involving only correlations and multiple correlations. We now turn to some analyses

PERSONALITY STABILITY

Table 5 The Variables With the Highest Correlations With Each of the First Two Canonical Variates From the Canonical Analyses (39 Variables) Variable

Child

Adult

First canonical variate Restless Careless of others’ property Fidgety Impulsive Irresponsible Fickle Nervous Spiteful Rude Mannerly Careful of personal belongings Persevering Planful Neat in appearance Conscientious

.63 .57 .56 .48 .45 .46 .44 .34 .31 ⫺.55 ⫺.52 ⫺.47 ⫺.43 ⫺.41 ⫺.39

.33 .36 .25 .47 .54 .17 ⫺.03 .35 .38 ⫺.36 ⫺.27 ⫺.26 ⫺.36 ⫺.41 ⫺.41

Second canonical variate Outspoken Socially confident Verbally fluent Assertive Gregarious Submissive Fearful

.49 .46 .43 .33 .12 ⫺.48 ⫺.27

.36 .14 .46 .28 .37 ⫺.22 ⫺.36

Note. For the first canonical variate, the correlation between the childhood and adult variable sets is .47 (.35 when corrected for capitalization on chance). For the second variate, the correlation is .43 (.30 when corrected).

involving structural equation models to provide additional insight about the long-term stability and continuity of personality-trait constructs. For our analyses of latent construct continuity, we began with the 39 variables assessed by teachers in childhood and the equivalent set of variables assessed by self-reports at midlife (with “self-centered,” reverse scored, substituted for “selfminimizing” in the adult assessment). In exploratory factor analyses of the adult self-ratings, these 39 variables were factored with and without the inclusion of the BFI items. Inspection of the orthogonal varimax five factor solutions in both analyses suggests that we should omit 6 of the 39 variables (mannerly, eccentric, impulsive, happy, seclusive, and lethargic), which were insufficiently strongly or univocally associated with any dimension. Thus, latent construct continuity was examined by using 33 variables as indicators of the adult constructs of Extraversion (assertive, energetic, gregarious, outspoken, socially confident, and verbally fluent vs. submissive), Agreeableness (considerate, adaptable, and self-minimizing/self-centered vs. rude, spiteful, and rigid), Conscientiousness (conscientious, careful, planful, neat in appearance, and persevering vs. careless and irresponsible), Neuroticism (nervous, fidgety, restless, jealous, fickle, fearful, touchy, complaining, and concerned about acceptance), and Openness (imaginative, original, curious, and aesthetically sensitive). Construct continuity was examined separately for each of the Big Five dimensions by using structural equation modeling conducted with Mplus (Version 3), with standard maximum-

769

likelihood estimation (Muthe´n & Muthe´n, 2004). As indices of model fit, we examined the comparative fit index (CFI) and the root-mean-square error of approximation (RMSEA). The same trait indicators were used for both the child and adult latent constructs. To improve the fit of the models, correlated errors were allowed among the indicators within the child and within the adult constructs but not between them. Gender differences were evaluated by conducting multiple-group analyses. However, for none of the Big Five factors did the multiple-group analysis indicate that there were significant gender differences on any parameters, and therefore we report the findings based on the total sample. For each of the Big Five factors, one model was tested in which the parameter estimates for the indicators of each construct were constrained to be equal in childhood and middle age, and another in which the parameters were unconstrained. If freeing the relative weights of the child indicators instead of forcing them to be identical to those for adults produced a better fit, this would suggest some differences in the way the Big Five dimensions were measured for children versus adults. Figures 1–5 show the unconstrained best-fitting structural equation models and provide the standardized path coefficients so that the relative strengths of the variables might be compared. Fit indices are provided in the figure captions. The standardized path coefficients between the child and the adult latent constructs, representing the correlations between them, demonstrated the same pattern of stability observed in our previous analyses, with Extraversion being the most stable of the Big Five factors and Neuroticism the least stable. These analyses also evaluated construct continuity. The indicators (all p ⬍ .05) are ordered in terms of the absolute size of their path estimates for the adult latent construct. Variables with the largest discrepancy between these coefficients for the child versus the adult latent constructs appear to contribute differently to child and adult versions. The variables that showed the largest differences in parameter estimates for the child versus the adult constructs were identified from the best-fitting unconstrained models. The paths for these variables as indicators of the child latent construct were then sequentially freed, and the model fit was compared with the best-fitting model in which all the parameters for the child and adult constructs were constrained to be equal. These analyses suggested that the fit for Extraversion was significantly improved by freeing the variables socially confident and submissive, ␹2Difference(2, N ⫽ 2221) ⫽ 19.27, p ⬍ .001 (see Figure 1); for Conscientiousness, no significant improvement in fit was achieved (see Figure 2); for Openness the fit was significantly improved by freeing aesthetically sensitive, ␹2Difference(1, N ⫽ 2221) ⫽ 4.42, p ⬍ .05 (see Figure 3); for Agreeableness the fit was improved by freeing rigid and adaptable, ␹2Difference(2, N ⫽ 2221) ⫽ 132.95, p ⬍ .001 (see Figure 4); and for Neuroticism the fit was improved by freeing concerned about acceptance and fearful, ␹2Difference(2, N ⫽ 2221) ⫽ 18.92 p ⬍ .02 (see Figure 5). The traits that significantly altered the fit of the models suggest some discontinuities between the child and adult versions of the constructs. For Extraversion, being socially confident was more important for adults (i.e., had a larger path estimate), and not being submissive was more important for children. For Agreeableness, being adaptable and not being rigid were more important for adults than for children. Being fearful was a stronger indicator of adult than of child Neuroticism, whereas being concerned about acceptance was a better indicator of childhood than was adult Neuroti-

HAMPSON AND GOLDBERG

770 Socially confident

Socially confident

Assertive

Assertive .49

Outspoken

.66

.30

.71 .86

Verbally fluent

.60

Outspoken

.60 .56

Child Extraversion

Adult Extraversion

Verbally fluent

.55

.59

.47

.64

.41

- .63

- .33

Gregarious

Gregarious

Energetic

Energetic

Submissive

Submissive Figure 1. The structural equation model for child and adult Extraversion; ␹2(68, N ⫽ 2221) ⫽ 126.86, p ⫽ .000. Comparative fit index ⫽ .991, root-mean-square error of approximation ⫽ .020, 90% confidence interval ⫽ .014 –.025.

bility identified by Caspi et al. (2005), with one exception. These findings contribute to the enduring issue of stability versus change in personality over the life course and have implications for life-course models of the effects of childhood personality on adult outcomes.

cism. Aesthetic sensitivity was the weakest indicator for both childhood and adult Openness, but it was more strongly related to the childhood construct.

Discussion The combination of childhood and midlife personality assessments in the Hawaii Personality and Health Cohort offered a unique opportunity to test the principles of rank-order personality stability across a hitherto unexamined span of 40 years from middle childhood to middle age. As discussed in the following paragraphs, our findings support the principles of personality sta-

Trait Stability Within Childhood and Adulthood Consistent with Principle 2, that rank-order stability increases with age, the 3-year stability correlations within childhood ranged from .36 to .55 (see Table 2). We can compare these correlations with those obtained within adulthood over the roughly comparable

Irresponsible

Irresponsible

Careless

Careless - .83

Careful

- .68

.26

- .87

- .64

.69

Planful Conscientious

.76

Careful

.59

Child Conscientiousness

Adult Conscientiousness

.49

.71

.46

.53

.44

.82

.34

Planful Conscientious

Neat

Neat

Persevering

Persevering Figure 2. The structural equation model for child and adult Conscientiousness; ␹2(68, N ⫽ 2221) ⫽ 97.79, p ⫽ .000. Comparative fit index ⫽ .997, root-mean-square error of approximation ⫽ .014, 90% confidence interval ⫽ .007–.020.

PERSONALITY STABILITY .12

Imaginative Original

Imaginative

.80

.83

.80

.64

.66

Curious

771

Child Openness

Adult Openness

.52

Original

.50

Curious

.29

Aesthetically Sensitive

Aesthetically Sensitive .

Figure 3. The structural equation model for child and adult Openness; ␹2(19, N ⫽ 2221) ⫽ 22.41, p ⫽ .264. Comparative fit index ⫽ .991, root-mean-square error of approximation ⫽ .009, 90% confidence interval ⫽ .000 –.021.

ably higher than stability coefficients for the 40-year interval between childhood and adulthood. And, this conclusion was indeed confirmed. The stability correlations within both childhood and adulthood were higher than the stability correlations spanning the interval from childhood to adulthood: These long-term stability correlations ranged from a low of around zero for Neuroticism to a high of .29 for Extraversion. Given that the childhood assessments were conducted when the personality was developing, and the adult assessments occurred at a developmental plateau when adult personality had stabilized and prior to any changes in later life, to obtain any significant correlations between these two assessments could be seen as remarkable. We anticipated lower stability correlations than those reported in previous studies. For example, reports of correlations within adulthood across decades have ranged from .14 to .37 across 50 years (Haan et al., 1986), and from .07 to .38 across 30 years (Soldz & Vaillant, 1999). However, despite crossing the child–adult boundary in our 40 year follow-up, we found correlations displaying a similar range, from around .00 to .30.

test–retest interval of 2.8 years, which ranged from .70 to .79 (see Table 3). Consistent with Principle 3, that personality stability peaks in the fourth and fifth decades (Roberts & DelVecchio, 2000), the test–retest correlations found here indicate considerably higher test–retest stability at midlife than in childhood. The test– retest correlations in childhood were relatively low for at least two reasons: Different teachers rated the children on each occasion, and this period of middle childhood is when personality is developing and hence is not expected to be highly stable. In contrast, the test–retest correlations in adulthood were based on self-reports during the period of maximum personality stability. However, direct comparisons of these short-term stability correlations between childhood and adulthood should be viewed cautiously, given that the childhood subsamples used for these analyses were from the University of Hawaii Laboratory School, which were smaller and less typical of the elementary school population at the time than our other childhood subsamples, and consequently were not included in our long-term analyses.

Trait Stability Between Childhood and Adulthood Differences Among Traits

According to Principles 1, 2, and 3 of rank-order stability offered by Caspi et al. (2005), stability coefficients over relatively short periods within childhood and adulthood should be consider-

The meta-analysis by Roberts and DelVecchio (2000) showed the highest rank-order stability (population estimates controlling

Rude

Rude

Self-Minimizing Spiteful

.14 - .86

- .67

.61

.57

- .81 - .34

Rigid Considerate Adaptable

Child Agreeableness

Adult Agreeableness

(not) Self-Centered Spiteful

- .57 - .54

.67

.50

.21

.37

Rigid Considerate Adaptable

Figure 4. The structural equation model for child and adult Agreeableness; ␹2(45, N ⫽ 2221) ⫽ 67.78, p ⫽ .016. Comparative fit index ⫽ .996, root-mean-square error of approximation ⫽ .015, 90% confidence interval ⫽ .007–.022.

HAMPSON AND GOLDBERG

772 Nervous

Nervous

Fidgety

Fidgety

Fickle

.68

.73

.61

Jealous

.68

- .02

.62

.64

.75

Fearful Restless

Complaining

.27

Fickle Jealous

.58

Child Neuroticism

Adult Neuroticism

.57

.62

.56

.75

.46

.58

.44

.58

.25

Fearful Restless

Complaining

Touchy

Touchy

Concerned

Concerned Figure 5. The structural equation model for child and adult Neuroticism; ␹2(112, N ⫽ 2221) ⫽ 183.43, p ⫽ .000. Comparative fit index ⫽ .992, root-mean-square error of approximation ⫽ .017, 90% confidence interval ⫽ .012–.021.

for time intervals and age of the samples) for Extraversion and Agreeableness (each .54) and the lowest for Conscientiousness (.51), Neuroticism (.50), and Openness (.51). These values formed the basis for Caspi et al.’s (2005) Principle 6 that stability does not vary by trait. Contrary to that conclusion, however, our study suggests that in the Hawaii cohort, all traits are not equally consistent from childhood to adulthood. Rather, we found several differences among the Big Five factors in their 40-year stabilities. At one extreme is Extraversion, followed closely by Conscientiousness, both of which demonstrate statistically significant links between childhood and adulthood, no matter the type of methodological procedure that is used to analyze those traits. At the other extreme, is Neuroticism, followed closely by Agreeableness, neither of which displays significant longitudinal stabilities over this long time period. In the middle, is the somewhat motley construct of Openness, which tends to provide statistically significant stability correlations, albeit at a much reduced level than that of Extraversion. Although this is but one study to compare with the many used in the meta-analysis, the Hawaii cohort is unique, bringing hitherto unavailable new data to this discussion. None of the studies included in the meta-analysis contributed rank-order stability coefficients spanning childhood and adulthood approaching anywhere near 40 years. We have provided evidence of both the short-term and the long-term relative instability for the construct of Neuroticism. This finding makes good sense, given that the broad dimension of Neuroticism is the most “state-like” of the Big Five traits (Chaplin, John, & Goldberg, 1988) and the domain commonly targeted for

change by therapeutic endeavors. Neuroticism is also the factor that is most private and least visible to outside observers, perhaps even including classroom teachers. Indeed, it is the factor that has been found to be least concordant between descriptions provided by oneself and those of knowledgeable others (e.g., Funder, 1995, 1999). Therefore, the relative instability of Neuroticism found here may have been enhanced by the use of observer reports in childhood and self-reports in adulthood and may be less marked when the same assessment method is used at each time point. At the other extreme is Extraversion versus Introversion, the least evaluatively charged of the Big Five, the most easily observable of them all, and the factor that tends to be most concordant between self-descriptions and those of others, even relative strangers (e.g., Norman & Goldberg, 1966). Both Extraversion and Neuroticism are normally considered the two dimensions of temperament, easily measurable in infancy and toddlerhood. Yet, our data suggest that they differ in both their short-term and their long-term stabilities, and this finding poses a challenge for those developmental theories linking infant temperament with adult personality traits. In between the two extremes are three factors that appear to differ somewhat among themselves in their long-term (but not their short-term) stabilities. Conscientiousness appears relatively stable, whereas Agreeableness (often the largest of the Big Five factors in lexical studies, and a relatively observable trait) does not. To the extent to which Openness entails aspects of cognitive and other school-related abilities, one would expect significant childhood–adulthood stability. On the other hand, to the extent to

PERSONALITY STABILITY

which this construct involves aspects of aesthetic appreciation and related artistic or creative experiences not readily available in childhood, one might expect some discontinuities between these two age periods. Perhaps, then, it is no wonder that the broad construct that combines these various personality characteristics displays some modest childhood-to-adulthood stability but not as strong as that of Extraversion and Conscientiousness.

Limitations to the Interpretation of the Stability Coefficients Although the differences among stability correlations for the Big Five appear to be reliable for the Hawaii cohort, given the lack of comparable studies spanning such a long time interval and bridging the child–adult divide, the generalizability of these findings is yet to be determined. Therefore, it remains a possibility that the differences among the Big Five observed here are unique to this cohort and the relation between these two particular assessment points. Moreover, as noted at the outset of this report, a limitation of this study was that different assessment methods were used in childhood (teacher assessments) and adulthood (selfreports) on different instruments. This may have led to an underestimation of trait stability for some traits (e.g., Neuroticism) and certainly makes comparisons with findings from previous studies in which the same methods were used on both occasions more problematic. Whereas zero stability for Neuroticism is consistent with this methodological explanation, the close-to-zero stability for Agreeableness is less amenable to this account and remains a conundrum. A weakness of any study of trait stability with only two time points is that it does not address the development process, so we cannot make inferences from these single coefficients about differences between the trajectories of trait stability or change over time among the Big Five (Fraley & Roberts, 2005). One striking difference between the childhood and adulthood assessments was the complexity of the Big Five factor structure in childhood compared with in adulthood. Teachers may not have known their students well enough to make fine-grained distinctions among 39 traits, but this seems unlikely given the precautions taken to ensure that teachers were familiar with their students and understood the meaning of the traits to be assessed. Because we know from previous research involving peer descriptions of adults that the structure of peer ratings and self-ratings is virtually identical (e.g., Goldberg, 1990), the complex structure of these teacher assessments could reflect the complexities in their students’ personality traits at this early age, which has implications for construct continuity.

Construct Continuity Whenever one finds instability in a trait between childhood and adulthood, it is worrisome unless the finding reflects nothing more than construct drift over time. It is plausible that teachers’ concepts of children’s personality traits could differ from trait concepts accessed 40 years later via self-reports. We examined construct continuity in the following three ways: (a) by multiple regression analyses, (b) by canonical correlations, and (c) by structural equation modeling. If the childhood version of a trait predicting the

773

corresponding adult version contributed to the prediction of another (i.e., “off-diagonal”) adult trait, this would imply construct discontinuity. In our analyses, most of the stepwise regression analyses that used the five childhood factors to predict each of the adult factors stopped at or before the first step. These analyses provide more support for continuity than for discontinuity at the level of the Big Five factors. When we used the common set of 39 variables in a canonical correlation analysis, we found two substantial orthogonal canonical variates that optimally linked the childhood and adult variables. One of these combined many of the traits in the Conscientiousness, Agreeableness, and Neuroticism domains to form a broad prosocial dimension (reflecting a concern with getting along with others) and the other formed a clear Extraversion dimension (reflecting a concern with getting ahead of others). These canonical variates mirror virtually identically the two higher order dimensions of the Big Five factor structure (see Digman, 1997; Saucier & Goldberg, 2003) proposed by the same investigator, John M. Digman, who much earlier had collected this unique set of teacher assessments. This analytic approach, conducted at the level of the 39 common variables that were allowed to be differentially weighted, implies considerably less construct continuity than the regression approach. Construct continuity was optimized at the level of the two higher order factors rather than at the more specific level of the Big Five. Finally, the structural equation modeling approach to the question of construct continuity explored differences in the measurement of each of the latent constructs for the Big Five childhood versus adult factors. These analyses were conducted separately for each of the Big Five factors because the complexity of the childhood data makes it difficult to get models to converge when cross-loadings between factors are permitted. Some measurement discontinuities were identified in these analyses, demonstrating that improved model fit could be achieved by freeing indicators of the child latent construct to differ from those of the adult latent construct. These differences reflect what may be thought of as variations in more versus less prototypical indicators for the child and the adult versions of the same Big Five trait. Taken together, the various approaches illustrate that the degree of construct continuity observed depends on how it is studied. If indicators of the latent constructs are allowed to freely “crossload,” as in the canonical correlations, then the optimal prediction between child and adult is achieved through latent constructs at a broader level than the Big Five traits. If the Big Five factors are created at each of the two age points and then related through the use of regression analyses, they suggest moderate construct continuity. If the Big Five are assumed, yet their indicators are allowed to vary, as in the structural modeling approach, then subtle differences between the child and adult latent constructs can be observed. Before concluding, we turn briefly to Caspi et al.’s (2005) Principle 5 that rank-order stability does not differ by gender. Our findings confirm this conclusion. In the regression analyses, gender only moderated the prediction of adult personality factors by childhood personality factors in two cases (Extraversion and Agreeableness), and no gender effects were obtained in the structural equation models.

HAMPSON AND GOLDBERG

774 Conclusions

This study has contributed to the continuing scientific debate about the lifetime stability of personality traits by presenting findings across a 40-year interval spanning childhood to middle age. Our findings suggest that for members of the Hawaii cohort there may be differences in stability among the Big Five traits. There is no inconsistency between this finding and the evidence that the Big Five are about equally heritable. Just as a trait could be extremely stable but not heritable (e.g., a trait completely determined by one strong kind of environmental influence), so a trait could be strongly heritable but not at all stable (e.g., the number of hair follicles per square centimeter of scalp in males). It is the genetic developmental program that gets inherited, not the phenotypical expression of a trait at any particular point in time. A further reason for studying trait stability is in service of the larger goal of identifying life span pathways from childhood predictors to adult morbidity and mortality. For example, childhood Conscientiousness has been shown to predict health behaviors that may mediate its effects on longevity (e.g., Friedman et al., 1995; Hampson, Goldberg, Vogt, & Dubanoski, in press). Presumably, if more conscientious children grow up into more conscientious adolescents and adults, then they are more likely to consistently benefit from engaging in health-enhancing behaviors and so increase their chances for a long and healthy life. The moderate stability of Conscientiousness observed here supports such a life course pathway involving this trait. Moreover, the even greater stability of Extraversion spotlights this trait for future studies of its influence on life course pathways to health.

References Asendorpf, J. B. (1992a). Beyond stability: Predicting inter-individual differences in intra-individual change. European Journal of Personality, 6, 103–117. Asendorpf, J. B. (1992b). A Brunswikean approach to trait continuity: Application to shyness. Journal of Personality, 60, 53–77. Backteman, G., & Magnusson, D. (1981). Longitudinal stability of personality characteristics. Journal of Personality, 49, 148 –160. Caspi, A., & Roberts, B. W. (1999). Personality continuity and change across the life course. In L. Pervin & O. P. John (Eds.), Handbook of personality psychology: Theory and research (2nd ed., pp. 300 –326). New York: Guilford Press. Caspi, A., & Roberts, B. W. (2001). Personality development across the life course: The argument for change and continuity. Psychological Inquiry, 12, 49 – 66. Caspi, A., Roberts, B. W., & Shiner, R. L. (2005). Personality development: Stability and change. Annual Review of Psychology, 56, 453– 484. Caspi, A., & Silva, P. A. (1995). Temperamental qualities at age three predict personality traits in young adulthood: Longitudinal evidence from a birth cohort. Child Development, 66, 486 – 498. Cattell, R. B., & Coan, R. A. (1957). Child personality structure as revealed in teachers’ ratings. Journal of Clinical Psychology, 13, 315–327. Chaplin, W. F., John, O. P., & Goldberg, L. R. (1988). Conceptions of states and traits: Dimensional attributes with ideals as prototypes. Journal of Personality and Social Psychology, 54, 541–557. Conley, J. J. (1984). Longitudinal consistency of adult personality: Selfreported characteristics across 45 years. Journal of Personality and Social Psychology, 47, 1325–1333. Conley, J. J. (1985). Longitudinal stability of personality traits: A multitrait–multimethod–multioccasion analysis. Journal of Personality and Social Psychology, 49, 1266 –1282.

Costa, P. T., Jr., & McCrae, R. R. (1988). Personality in adulthood: A 6-year longitudinal study of self-reports and spouse ratings on the NEO Personality Inventory. Journal of Personality and Social Psychology, 54, 853– 863. Costa, P. T., Jr., & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO PI-R) and NEO Five-Factor Inventory (NEO-FFI) professional manual. Odessa, FL: Psychological Assessment Resources. Digman, J. M. (1989). Five robust trait dimensions: Development, stability, and utility. Journal of Personality, 57, 195–214. Digman, J. M. (1997). Higher order factors of the Big Five. Journal of Personality and Social Psychology, 73, 1246 –1256. Digman, J. M., & Inouye, J. (1986). Further specification of the five robust factors of personality. Journal of Personality and Social Psychology, 50, 116 –123. Digman, J. M., & Shmelyov, A. G. (1996). The structure of temperament and personality in Russian children. Journal of Personality and Social Psychology, 71, 341–351. Digman, J. M., & Takemoto-Chock, N. K. (1981). Factors in the natural language of personality: Reanalysis, comparison, and interpretation of six major studies. Multivariate Behavioral Research, 16, 149 –170. Fraley, R. C., & Roberts, B. W. (2005). Patterns of continuity: A dynamic model for conceptualizing the stability of individual differences in psychological constructs across the life course. Psychological Review, 112, 60 –74. Friedman, H. S., Tucker, J. S., Schwartz, J. E., Tomlinson-Keasey, C., Martin, L. R., Wingard, D. L., & Criqui, M. H. (1995). Psychosocial and behavioral predictors of longevity: The aging and death of the “Termites.” American Psychologist, 50, 69 –78. Friedman, H. S., Tucker, J. S., Tomlinson-Keasey, C., Schwartz, J. E., Wingard, D. L., & Criqui, M. H. (1993). Does childhood personality predict longevity? Journal of Personality and Social Psychology, 65, 176 –185. Funder, D. C. (1995). On the accuracy of personality judgment: A realist approach. Psychological Review, 102, 652– 670. Funder, D. C. (1999). Personality judgment: A realistic approach to person perception. San Diego, CA: Academic Press. Goldberg, L. R. (1990). An alternative “Description of personality”: The Big Five factor structure. Journal of Personality and Social Psychology, 59, 1216 –1229. Goldberg, L. R. (1992). The development of markers for the Big Five factor structure. Psychological Assessment, 4, 26 – 42. Goldberg, L. R. (2001). Analyses of Digman’s child-personality traits: Derivation of Big Five factor scores from each of six samples. Journal of Research in Personality, 69, 709 –743. Haan, N., Millsap, R., & Hartka, E. (1986). As time goes by: Change and stability in personality over 50 years. Psychology and Aging, 1, 220 – 232. Hampson, S. E., Dubanoski, J. P., Hamada, W., Marsella, A. J., Matsukawa, J., Suarez, E., & Goldberg, L. R. (2001). Where are they now? Locating former elementary school students after nearly 40 years for a longitudinal study of personality and health. Journal of Research in Personality, 35, 375–387. Hampson, S. E., Goldberg, L. R., Vogt, T. M., & Dubanoski, J. P. (2006). Forty years on: Teachers’ assessments of children’s personality traits predict self-reported health behaviors and outcomes at midlife. Health Psychology, 25, 57– 64. Hampson, S. E., Goldberg, L. R., Vogt, T. M., & Dubanoski, J. P. (in press). Mechanisms by which childhood personality traits influence adult health status: Educational attainment and healthy behaviors. Health Psychology. John, O. P., Caspi, A., Robins, R. W., Moffitt, T. E., & Stouthamer-Loeber, M. (1994). The “Little Five”: Exploring the nomological network of the five-factor model of personality in adolescent boys. Child Development, 65, 160 –178.

PERSONALITY STABILITY John, O. P., & Robins, R. W. (1993). Determinants of interjudge agreement on personality traits: The Big Five domains, observability, evaluativeness, and the unique perspective of the self. Journal of Personality, 61, 521–552. John, O. P., & Srivastava, S. (1999). The Big Five trait taxonomy: History, measurement, and theoretical perspectives. In L. A. Pervin & O. P. John (Eds.), Handbook of personality: Theory and research (2nd ed., pp. 102–138). New York: Guilford. Kagan, J. (1980). Perspectives on continuity. In O. G. Brim, Jr., & J. Kagan (Eds.), Constancy and change in human development (pp. 26 –74). Cambridge, MA: Harvard University Press. Kenny, D. A. (1994). Interpersonal relations: A social relations analysis. New York: Guilford Press. Lewis, M. (2001). Issues in the study of personality development. Psychological Inquiry, 12, 67– 83. McCrae, R. R., Costa, P. T., Jr., Ostendorf, F., Angleitner, A., Hrebickova, M., Avia, M. D., et al. (2000). Nature over nurture: Temperament, personality, and life span development. Journal of Personality and Social Psychology, 78, 173–186. Measelle, J. R., John, O. P., Ablow, J. C., Cowan, P. A., & Cowan, C. P. (2005). Can children provide coherent, stable, and valid self-reports on the Big Five dimensions? A longitudinal study from ages 5 to 7. Journal of Personality and Social Psychology, 89, 90 –106. Mervielde, I., Buyst, V., & De Fruyt, F. (1995). The validity of the Big Five as a model for teachers’ ratings of individual differences among children aged 4 –12 years. Personality and Individual Differences, 18, 525–534. Muthe´n, L. K., & Muthe´n, B. O. (2004). Mplus User’s Guide (3rd ed.). Los Angeles, CA: Authors.

775

Norman, W. T., & Goldberg, L. R. (1966). Raters, ratees, and randomness in personality structure. Journal of Personality and Social Psychology, 4, 681– 691. Roberts, B. W., & Chapman, C. (2000). Change in dispositional well-being and its relation to role quality: A 30-year longitudinal study. Journal of Research in Personality, 34, 26 – 41. Roberts, B. W., & DelVecchio, W. F. (2000). The rank-order consistency of personality traits from childhood to old age: A quantitative review of longitudinal studies. Psychological Bulletin, 126, 3–25. Rubin, K. H., Hymel, S., & Mills, R. S. L. (1989). Sociability and social withdrawal in childhood: Stability and outcomes. Journal of Personality, 57, 237–255. Saucier, G. (1994). Mini-markers: A brief version of Goldberg’s unipolar Big Five markers. Journal of Personality Assessment, 63, 506 –516. Saucier, G., & Goldberg, L. R. (2003). The structure of personality attributes. In M. R. Barrick & A. M. Ryan (Eds.), Personality and work: Reconsidering the role of personality in organizations (pp. 1–29). San Francisco, CA: Jossey-Bass. Shiner, R. L. (1998). How shall we speak of children’s personalities in middle childhood? A preliminary taxonomy. Psychological Bulletin, 124, 308 –332. Shiner, R. L., & Caspi, A. (2003). Personality differences in childhood: Measurement, development, and consequences. Journal of Child Psychology and Psychiatry, 44, 2–32. Soldz, S., & Vaillant, G. E. (1999). The Big Five personality traits and the life course: A 45-year longitudinal study. Journal of Research in Personality, 33, 208 –232.

(Appendixes follow)

HAMPSON AND GOLDBERG

776

Appendix A Five Factors Derived From 39 Childhood Variables From the Sample for Whom Adult Personality Measures Were Available Childhood variables Conscientiousness (C) Careful of personal belongings Planful Persevering Neat in appearance Conscientious Mannerly Irresponsible Careless of others’ property Fidgets Fickle Restless Eccentric Nervous habits Agreeableness (A) Self-minimizing Considerate Submissive Spiteful Rude Jealous Assertive Touchy Complains about others Impulsive Outspoken Openness (O) Imaginative Original Curious Socially Confident Esthetically sensitive Verbally fluent Adaptable Rigid Extraversion (E) Gregarious Energetic Happy Seclusive Lethargic Neuroticism (N) Fearful Concerned about acceptance

C

A

O

E

N

ⴚ.78a ⴚ.71a ⴚ.71a ⴚ.67a ⴚ.59a ⴚ.54a .74a .72a .65a .62a .60a .51a .48a

⫺.11 ⫺.06 ⫺.16 ⫺.02 ⴚ.40 ⴚ.38 .18 .33 .35 .28 .40 .11 .23

.02 .34 .32 ⫺.03 .23 .31 ⫺.22 ⫺.07 .00 ⫺.13 .06 .15 ⫺.03

.02 .04 .02 .13 ⫺.04 .07 ⫺.03 .06 .30 .23 .39 ⴚ.37 .09

⫺.07 ⫺.23 ⫺.17 .07 ⫺.19 .00 .11 .08 .20 .32 .18 .03 .46

⫺.24 ⴚ.45 ⫺.03 .36 .43 .22 .10 .17 .22 .50 .19

ⴚ.63a ⴚ.61a ⴚ.59a .70a .70a .70a .68a .57a .55a .53a .53a

⫺.27 .19 ⴚ.40 .02 .02 ⫺.01 .30 ⫺.10 .09 .12 .45

⫺.24 .10 ⫺.26 .12 .08 .08 .32 ⫺.14 .18 .33 .38

⫺.02 ⫺.01 .19 .16 .15 .35 .01 .47 .40 .18 ⫺.07

⫺.05 ⫺.05 ⫺.06 ⫺.20 ⴚ.34 ⫺.02 ⴚ.30 .28

.02 .06 .18 .22 ⫺.18 .32 ⫺.03 .25

.80a .78a .70a .60a .57a .56a .55a ⴚ.45a

.11 .08 .25 .28 .13 .27 .16 ⫺.14

.05 ⫺.12 ⫺.07 ⫺.23 .18 ⫺.07 ⴚ.44 .37

.12 .12 ⫺.18 .01 .21

.30 .27 ⫺.24 ⫺.14 ⫺.16

.20 .30 .38 ⫺.28 ⴚ.40

.72a .64a .56a ⴚ.71a ⴚ.49a

.01 ⫺.05 ⫺.28 .06 .06

.09 .15

⫺.05 .31

⫺.24 .01

⫺.16 ⫺.03

.72a .67a

Note. N ⫽ 799. These 39 variables were common to the Kauai and the two Oahu samples. Only short-hand labels are used in this table; for the complete variable descriptions, see Goldberg (2001). Correlations of .30 or more are listed in boldface type. a Highest loading for each variable.

PERSONALITY STABILITY

777

Appendix B Six Varimax Factors Derived From 130 Personality Items in the Adult Sample Item Neuroticism (N) Tense [C] Nervous [C] Fretful [M] Moody [M] Jealous [C, M] Can be tense [B] Temperamental [M] Worries a lot [B] Gets nervous easily [B] Envious [M] Can be moody [B] Restless [C] Fidgety [C] Complaining [C] Spiteful [C] Is depressed, blue [B] Suspicious [C] Harsh [M] Fearful [C] Touchy [C, M] Tends to find fault with others [B] Rigid [C] Fickle [C] Starts quarrels with others [B] Self-centered [C] Lethargic [C] Concerned about acceptance [C] Submissive [C] Impulsive [C] Is relaxed, handles stress well [B] Is emotionally stable, not easily upset [B] Relaxed [M] Emotionally stable [C] Remains calm in tense situations [B] Unenvious [M] Adaptable [C] Openness (O) Imaginative [C, M] Is original, comes up with new ideas [B] Is inventive [B] Has an active imagination [B] Creative [M] Likes to reflect, play with ideas [B] Is curious about many different things [B] Original [C] Philosophical [M] Is ingenious, a deep thinker [B] Values artistic, aesthetic experiences [B] Curious [C] Intellectual [M] Is sophisticated in art, music, or lit. [B] Deep [M] Knowledgeable [C] Perceptive [C] Persevering [C] Complex [M] Verbally fluent [C] Is full of energy [B]

N

O

C

E

A

PA

.76a .70a .68a .66a .65a .64a .64a .63a .63a .62a .62a .59a .58a .58a .58a .57a .57a .56a .55a .53a .52a .51a .50a .45a .41a .39a .37a .28a .28a ⴚ.59a ⴚ.57a ⴚ.47a ⴚ.45a ⴚ.45a ⴚ.33a ⴚ.30a

.00 ⫺.04 ⫺.08 .02 ⫺.05 .03 .09 ⫺.08 ⫺.17 ⫺.12 .05 .12 .07 ⫺.11 ⫺.05 ⫺.01 .06 .04 ⫺.12 .06 .03 .03 ⫺.02 .02 .16 ⫺.12 ⫺.06 ⫺.12 .27 .30 .23 .15 .11 .33 .06 .30

⫺.07 ⫺.14 ⫺.15 ⫺.08 ⫺.16 ⫺.03 ⫺.02 ⫺.14 ⫺.17 ⫺.10 ⫺.10 ⫺.13 ⫺.18 ⫺.03 ⫺.02 ⫺.20 .04 .02 ⫺.18 ⫺.05 .03 .06 ⫺.16 ⫺.15 ⫺.14 ⫺.29 ⫺.03 ⫺.11 ⫺.24 .16 .17 .09 .31 .27 .02 .16

⫺.01 ⫺.12 ⫺.04 ⫺.09 .02 ⫺.06 .09 ⫺.10 ⫺.17 .02 ⫺.14 .14 .03 .09 ⫺.04 ⫺.22 ⫺.06 .04 ⫺.11 .04 .08 ⫺.08 ⫺.05 .11 .02 ⫺.22 ⫺.03 ⫺.24 .15 .04 .02 ⫺.10 .08 ⫺.06 ⫺.10 .07

⫺.11 .08 .02 ⫺.21 ⫺.09 ⫺.12 ⫺.15 .08 .07 ⫺.03 ⫺.21 ⫺.04 ⫺.02 ⫺.20 ⫺.29 ⫺.14 ⫺.09 ⴚ.38 .04 ⫺.04 ⴚ.31 ⫺.28 ⫺.11 ⴚ.35 ⴚ.36 ⫺.10 .18 .23 .03 .10 .25 .26 .21 .18 .01 .29

⫺.07 ⫺.05 ⫺.06 .09 .02 ⫺.08 .06 ⫺.11 ⫺.12 .03 .03 ⫺.11 ⫺.01 ⫺.04 .03 ⫺.02 ⫺.10 ⫺.00 ⫺.07 .04 ⫺.08 ⫺.12 .08 .03 ⫺.04 .04 ⫺.10 ⫺.09 .09 .03 .00 .07 ⫺.01 .05 ⫺.16 ⫺.06

.05 .05 .05 ⫺.01 .04 .16 .09 .12 .11 .22 .06 .05 .20 .01 .16 .23 .29 .41 .09 .19 .28

.09 .17 .08 .12 .08 .07 .12 .10 .03 ⫺.03 .08 .13 .11 .08 .02 .17 .13 .10 ⫺.05 .34 .33

.15 .06 ⫺.02 .03 .11 .14 .17 .11 .06 ⫺.04 .20 .13 .04 .02 .14 .14 .13 .15 ⫺.10 .01 .12

.15 .17 .16 .08 .13 .00 ⫺.04 .23 ⫺.05 ⫺.01 ⫺.02 ⫺.11 .07 .07 .00 .14 ⫺.01 ⫺.15 ⫺.12 ⫺.01 .17

⫺.07 ⫺.05 ⫺.10 .07 .00 ⫺.11 .01 ⫺.03 .03 .12 ⫺.06 .07 ⫺.10 ⫺.06 .14 ⫺.12 ⫺.06 ⫺.07 .27 ⫺.03 ⫺.17

(Appendixes continue)

.71a .70a .68a .68a .67a .65a .64a .62a .62a .61a .59a .58a .56a .55a .54a .48a .48a .46a .45a .38a .34a

HAMPSON AND GOLDBERG

778

Appendix B (continued) Item Self-reliant [C] Eccentric [C] Aesthetically sensitive [C] Uncreative [M] Prefers work that is routine [B] Unintellectual [M] Has few artistic interests [B] Conscientiousness (C) Organized [M] Efficient [M] Does things efficiently [B] Makes plans and follows through with them [B] Planful [C] Does a thorough job [B] Perseveres until the task is finished [B] Careful of my belongings [C] Systematic [M] Practical [M] Conscientious [C] Neat in appearance [C] Is a reliable worker [B] Sensible [C] Energetic [C, M] Tends to be disorganized [B] Disorganized [M] Inefficient [M] Sloppy [M] Irresponsible [C] Careless [M] Tends to be lazy [B] Can be somewhat careless [B] Is easily distracted [B] Extraversion (E) Talkative [C, M] Is talkative [B] Is outgoing, sociable [B] Outspoken [C] Extraverted [M] Has an assertive personality [B] Socially confident [C] Assertive [C] Generates a lot of enthusiasm [B] Bold [M] Gregarious [C] Tends to be quiet [B] Quiet [M] Shy [M] Is sometimes shy, inhibited [B] Is reserved [B] Bashful [M] Withdrawn [M] Seclusive [C] Agreeableness (A) Is considerate and kind to almost everyone [B] Warm [M] Is helpful and unselfish with others [B] Kind [M] Sympathetic [M] Considerate [C] Cooperative [M] Thoughtful [C] Likes to cooperate with others [B] Is generally trusting [B] Has a forgiving nature [B] Obedient [C] Mannerly [C] Happy [C]

N

O

C

E

A

PA

⫺.11 .32 .20 .16 .22 .17 .15

.34 .32a .32a ⫺.55a ⫺.43a ⴚ.43a ⴚ.42a

.31 ⫺.14 .05 ⫺.08 .09 ⫺.20 .02

.02 ⫺.07 .02 ⫺.10 ⫺.15 ⫺.12 .00

.06 ⫺.16 .21 ⫺.14 .16 ⫺.12 ⫺.05

⫺.11 .06 ⫺.02 ⴚ.32 .11 ⫺.18 ⫺.03

⫺.07 ⫺.08 ⫺.08 ⫺.16 .00 ⫺.04 ⫺.10 .03 ⫺.02 ⫺.10 .01 ⫺.01 ⫺.06 ⫺.21 ⫺.14 .20 .21 .24 .22 .31 .30 .26 .30 .35

.12 .10 .18 .19 .24 .17 .21 .01 .32 .12 .16 .07 .17 .23 .30 .00 .01 ⫺.04 .01 ⫺.02 .00 ⫺.11 .05 .04

.75a .71a .67a .59a .58a .56a .55a .55a .52a .51a .48a .42a .39a .38a .30a ⴚ.69a ⴚ.68a ⴚ.63a ⴚ.58a ⴚ.56a ⴚ.56a ⴚ.49a ⴚ.49a ⴚ.40a

.06 .09 .01 .13 .03 .09 .01 .05 ⫺.01 ⫺.09 .05 .21 .09 .02 .30 ⫺.07 ⫺.02 ⫺.10 ⫺.12 .03 .01 ⫺.17 .06 ⫺.02

.05 .12 .19 .18 .10 .05 .11 .17 ⫺.03 .22 .18 .33 .16 .33 .15 .00 .04 ⫺.11 ⫺.07 ⫺.21 ⫺.08 ⫺.06 ⫺.07 .05

.14 .03 .11 ⫺.07 ⫺.07 .05 ⫺.09 .08 ⫺.09 ⫺.16 ⫺.22 .27 .03 ⫺.03 .13 ⫺.21 ⫺.14 ⫺.07 ⫺.27 ⫺.02 ⫺.11 ⫺.14 ⫺.19 ⫺.10

.15 .15 ⫺.11 .18 .07 .12 ⫺.23 .06 ⫺.08 .02 .03 .00 ⫺.05 .13 .23 .10 .18 .43 .42

.20 .20 .22 .32 .23 .36 .29 .35 .44 .37 .22 .01 .02 .03 .01 .05 ⫺.02 ⫺.01 .10

.01 .01 .09 .10 .07 .13 .22 .17 .09 .14 ⫺.01 .03 .08 ⫺.10 ⫺.07 .08 ⫺.06 ⫺.17 ⫺.09

.74a .72a .69a .61a .58a .57a .53a .52a .49a .48a .44a ⴚ.81a ⴚ.78a ⴚ.72a ⴚ.70a ⴚ.65a ⴚ.64a ⴚ.47a ⴚ.46a

.13 .09 .30 ⫺.17 .11 ⫺.05 .15 ⫺.09 .26 ⫺.13 .19 ⫺.03 .04 .10 .07 ⫺.07 .08 ⫺.22 ⫺.23

⫺.02 ⫺.02 .08 ⫺.05 ⫺.00 .09 .17 .03 .09 .08 .20 .00 .04 ⫺.08 ⫺.09 ⫺.09 ⫺.09 ⫺.15 ⫺.05

⫺.18 ⫺.11 ⫺.13 ⫺.10 .05 ⫺.11 ⫺.16 ⫺.03 ⫺.19 ⫺.18 ⫺.28 ⫺.02 .03 ⴚ.34

.09 .18 .13 .16 .12 .18 .16 .25 .13 .13 .14 ⫺.02 .21 .17

.15 .11 .17 .16 .11 .26 .21 .22 .18 .03 ⫺.06 .27 .33 .18

⫺.04 .18 .01 .04 .09 .04 .00 .08 .02 ⫺.01 .04 ⫺.18 .08 .21

a

.72a .69a .68a .67a .62a .61a .60a .59a .58a .50a .47a .40a .39a .35a

.06 .10 .06 .17 ⫺.08 .04 .01 .17 ⫺.08 .05 .06 .03 .11 .09

PERSONALITY STABILITY

779

Appendix B (continued) Item Rude [C, M] Is sometimes rude to others [B] Unsympathetic [M] Can be cold and aloof [B] Cold [M] Physical Attractiveness (PA) Is physically attractive Good-looking Is not good looking Unattractive

N

O

C

E

A

.44 .38 .21 .39 .41

⫺.02 .05 ⫺.17 .15 .08

⫺.04 ⫺.02 .14 .15

.24 .28 ⫺.18 ⫺.17

PA

⫺.09 ⫺.05 ⫺.15 .04 ⫺.04

.06 .04 ⫺.11 ⫺.25 ⫺.14

⫺.51 ⫺.50a ⴚ.50a ⴚ.46a ⴚ.46a

⫺.05 ⫺.06 ⫺.05 ⫺.06 ⫺.07

.17 .15 ⫺.17 ⫺.15

.17 .16 ⫺.19 ⫺.15

.13 .17 ⫺.15 ⫺.16

.71a .71a ⴚ.72a ⴚ.71a

a

Note. N ⫽ 762. Loadings of .30 or more are listed in boldface type. [B] ⫽ Big Five Inventory; [M] ⫽ Mini-Markers; [C] ⫽ childhood variables. a Highest factor loading for each item.

Received December 5, 2005 Revision received April 18, 2006 Accepted April 26, 2006 䡲

Instructions to Authors For Instructions to Authors, please visit www.apa.org/journals/psp and click on the “Instructions to Authors” link in the Journal Info box on the right.

PERSONALITY STABILITY

779

Appendix B (continued) Item Rude [C, M] Is sometimes rude to others [B] Unsympathetic [M] Can be cold and aloof [B] Cold [M] Physical Attractiveness (PA) Is physically attractive Good-looking Is not good looking Unattractive

N

O

C

E

A

.44 .38 .21 .39 .41

⫺.02 .05 ⫺.17 .15 .08

⫺.04 ⫺.02 .14 .15

.24 .28 ⫺.18 ⫺.17

PA

⫺.09 ⫺.05 ⫺.15 .04 ⫺.04

.06 .04 ⫺.11 ⫺.25 ⫺.14

⫺.51 ⫺.50a ⴚ.50a ⴚ.46a ⴚ.46a

⫺.05 ⫺.06 ⫺.05 ⫺.06 ⫺.07

.17 .15 ⫺.17 ⫺.15

.17 .16 ⫺.19 ⫺.15

.13 .17 ⫺.15 ⫺.16

.71a .71a ⴚ.72a ⴚ.71a

a

Note. N ⫽ 762. Loadings of .30 or more are listed in boldface type. [B] ⫽ Big Five Inventory; [M] ⫽ Mini-Markers; [C] ⫽ childhood variables. a Highest factor loading for each item.

Received December 5, 2005 Revision received April 18, 2006 Accepted April 26, 2006 䡲

Instructions to Authors For Instructions to Authors, please visit www.apa.org/journals/psp and click on the “Instructions to Authors” link in the Journal Info box on the right.

Journal of Personality and Social Psychology 2006, Vol. 91, No. 4, 780 –795

Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.780

Relating Emotional Abilities to Social Functioning: A Comparison of Self-Report and Performance Measures of Emotional Intelligence Marc A. Brackett and Susan E. Rivers

Sara Shiffman

Yale University

Skidmore College

Nicole Lerner and Peter Salovey Yale University Three studies used J. D. Mayer and P. Salovey’s (1997) theory of emotional intelligence (EI) as a framework to examine the role of emotional abilities (assessed with both self-report and performance measures) in social functioning. Self-ratings were assessed in ways that mapped onto the Mayer– Salovey–Caruso Emotional Intelligence Test (MSCEIT), a validated performance measure of EI. In Study 1, self-ratings and MSCEIT scores were not strongly correlated. In Study 2, men’s MSCEIT scores, but not self-ratings, correlated with perceived social competence after personality measures were held constant. In Study 3, only the MSCEIT predicted real-time social competence, again, just for men. Implications for analyzing how emotional abilities contribute to social behavior are discussed, as is the importance of incorporating gender into theoretical frameworks and study designs. Keywords: emotion, emotional intelligence, MSCEIT, gender, social competence

(Gross, 1998). Intelligent processing and effective management of emotional information are necessary to navigate the social world (Keltner & Kring, 1998). Emotional intelligence (EI) theory, which explicates the cognitive and emotional mechanisms that process emotional information, provides a unified framework to study the role of emotional abilities in social functioning (Mayer & Salovey, 1997; Salovey & Mayer, 1990). Mayer and Salovey’s model of EI identifies four interrelated emotional abilities, including the perception, use, understanding, and management of emotion. The purpose of the research described here is to examine the relationship between EI and social functioning. Demonstrating that EI is related to social functioning would support the emerging literature on the importance of emotional abilities for building better quality relationships. Different approaches to measuring EI can influence the validity of the construct, however. Thus, we now present an overview of the theory of EI and the two primary approaches to measuring EI: performance-based tests and self-report inventories.

Emotions contain information about a person’s relationship with the environment and can be triggered when the person– environment relationship changes (Lazarus, 1991). During social interactions, verbal and nonverbal emotional expressions convey information about one’s own and others’ thoughts, intentions, and behaviors (Buck, 1984; Ekman, 1973; Keltner & Haidt, 2001). Emotional abilities, including the ability to perceive, use, understand, and manage emotion, contribute to optimal social functioning (Denham et al., 2003; Eisenberg, Fabes, Guthrie, & Reiser, 2000; Feldman, Philippot, & Custrini, 1991; Nowicki & Duke, 1994; Savage, 2002). For example, accurately perceiving a person’s emotions (type and intensity) facilitates the prediction and understanding of that person’s subsequent actions (Elfenbein, Marsh, & Ambady, 2002). Understanding the significance of emotional states regarding the person– environment relationship guides attention, decision making, and behavioral responses (Damasio, 1994). Managing emotions effectively also is critical to optimal social functioning as this skill enables one to express socially appropriate emotions and behave in socially acceptable ways

EI: Theory and Measurement Two areas of psychological research informed the conceptualization of EI. The first pertains to how emotions and thinking interact (e.g., Bower, 1981; Clark & Fiske, 1982; Isen, Shalker, Clark, & Karp, 1978; Zajonc, 1980). Whereas intelligence and emotion often were considered in opposition (De Sousa, 1987), accumulating research in the 1980s documented how cognition and affect were integrated processes; affect influences many aspects of cognitive functioning, including memory, attention, and decision making (e.g., Damasio, 1994; Forgas & Moylan, 1987; Mayer & Bremer, 1985; Salovey & Birnbaum, 1989; Singer & Salovey, 1988). Accordingly, the theory of EI postulates that the information value of emotions can make thinking more intelligent.

Marc A. Brackett, Susan E. Rivers, Nicole Lerner, and Peter Salovey, Department of Psychology, Yale University; Sara Shiffman, Department of Psychology, Skidmore College. Research reported in this article was funded in part by National Cancer Institute Grant R01-CA68427. The third study was supported in part by the Student Opportunity Funds, Skidmore College. We are grateful for the help provided by Jack Mayer, Rebecca Warner, Paulo Lopes, Zorana Ivcevic, Nicole Katulak, Amy Latimer, Tanja Wranik, Elizabeth Mobayed, Aaron Kamholtz, and many research assistants who helped us with the data collection and coding. Correspondence concerning this article should be addressed to Marc A. Brackett, Department of Psychology, Yale University, P.O. Box 208205, New Haven, CT 06520-8205. E-mail: [email protected] 780

MEASURING EMOTIONAL INTELLIGENCE

EI theory also was developed as the concept of intelligence was broadening to include an array of mental abilities, including social, practical, and personal intelligence, rather than merely a monolithic g (e.g., Cantor & Kihlstrom, 1987; Gardner, 1983; Sternberg, 1985). Specific intelligences often are distinguished according to the kinds of information on which they operate (J. B. Carroll, 1993; Wechsler, 1997). EI operates on “hot” cognitions or information processing that involves matters of personal and emotional importance to individuals and their relationships (Abelson, 1963; Zajonc, 1980; see also Mayer & Mitchell, 1998). EI is distinguishable from other mental skills, such as verbal–propositional intelligence, which operates primarily on “cold” cognitive processes. EI also is conceptually and empirically distinct from temperament and personality traits, such as neuroticism (see Mayer, Salovey, & Caruso, 2004; Salovey & Mayer, 1990). Whereas neuroticism involves individual differences in thresholds of emotional reaction, latency, intensity, and recovery time (Rothbart, 1989) as well as the ease with which emotions are activated or aroused (Eisenberg et al., 1995), EI involves the accurate processing of emotionrelevant information (e.g., facial expressions) and the ability to use emotions in reasoning in order to solve problems. By way of example, an individual may be predisposed to a certain level of emotional reactivity and intensity, but emotion management skills determine how the person’s emotions are dealt with once activated. The four emotional abilities constituting the EI model are arranged such that the more basic psychological processes (i.e., perceiving emotions) are at the foundation, and more advanced processes (i.e., reflective regulation of emotion) are at the top of the model and are thought, to some extent, to be dependent on the lower level abilities. Within each dimension, there is a developmental progression of skills from the more basic to the more sophisticated (see Mayer & Salovey, 1997). Abilities within each dimension also are expected to develop with experience and age. The theory specifies that the four abilities contribute to the higher order construct of EI (Mayer & Salovey, 1997), which has been supported empirically (Mayer, Salovey, Caruso, & Sitarenios, 2003). Below is a brief description of the four EI abilities; more detailed information is available elsewhere (Brackett & Salovey, 2004; Mayer & Salovey, 1997; Rivers, Brackett, Salovey, & Mayer, in press). Perceiving emotion pertains to the ability to identify emotions in oneself and others, as well as in other stimuli, including voices, stories, music, and works of art (e.g., Ekman & Friesen, 1975; Nowicki & Mitchell, 1998; Scherer, Banse, & Wallbott, 2001). Using emotion involves the ability to harness feelings that assist in certain cognitive enterprises, such as reasoning, problem solving, decision making, and interpersonal communication. Emotions can create diverse mental sets that prove more or less tuned to various kinds of reasoning tasks (e.g., Isen, 1987; Palfai & Salovey, 1993; Schwarz, 1990; Schwarz & Clore, 1996). Understanding emotion involves language and propositional thought that reflect the capacity to analyze emotions. This skill includes an understanding of the emotional lexicon; the manner in which emotions combine, progress, transition from one to the other; and the outcomes of emotional experiences (e.g., Frijda, 1988; Lane, Quinlan, Schwartz, Walker, & Zeitlin, 1990). Managing emotion pertains to the ability to reduce, enhance, or modify an emotional response in oneself and others, as well as the ability to experience a range of emotions while also making decisions about the appropriateness or

781

usefulness of the emotion in a given situation (e.g., Eisenberg et al., 2000; Gross, 1998). Currently, there are two distinct types of EI theories and measurement tools. Mayer, Salovey, and Caruso (2000) distinguished the ability model described above from mixed models. Ability models conceptualize EI as a set of mental skills that can be assessed with performance tests. The first comprehensive performance test of EI was the Multifactor Emotional Intelligence Scale (Mayer, Caruso, & Salovey, 1999), which led to a briefer test, the Mayer–Salovey–Caruso Emotional Intelligence Test (MSCEIT, Version 2.0; Mayer, Salovey, & Caruso, 2002a). As a performance test, the MSCEIT assesses the ability to manage emotions, for example, with vignettes describing particular emotional problems, asking participants to rate a number of possible actions on a scale ranging from very ineffective to very effective. Responses are evaluated through a comparison of responses made by either experts or a normative sample. Mixed models, in contrast, are based primarily on popular depictions of EI (Goleman, 1995, 1998) and include three classes of constructs: perceived emotional (and other) abilities, competencies, and personality traits. For instance, Bar-On (1997) included the perceived ability to handle relationships and traits such as optimism in his model of EI. Proponents of this mixed approach, sometimes called the personality or trait approach, generally use self-report inventories to measure EI (Bar-On, 1997; Boyatzis, Goleman, & Rhee, 2000; Petrides & Furnham, 2003; Schutte et al., 1998). Two of the most widely used self-report inventories, the Emotion Quotient Inventory (Bar-On, 1997) and the Self-Report EI Test (Schutte et al., 1998), are strongly associated with indices of well-being, neuroticism, and depression (rs ⫽ 兩.50 to .70兩; Bar-On, 1997, 2000; Brackett & Mayer, 2003; Dawda & Hart, 2000; Newsome, Day, & Catano, 2000; Parker, Taylor, & Bagby, 2001). The associations of the MSCEIT to the Emotion Quotient Inventory and the Self-Report EI Test are rather low (rs ⱕ .22; Brackett & Mayer, 2003; David, 2005), indicating that self-report measures based on popularizations of EI and performance measures based on Mayer and Salovey’s (1997) EI theory yield different information about the same person. Evidence is accumulating that the MSCEIT has a factor structure congruent with the theory on which it is based; it also is reliable, distinct from established measures of personality, and not especially susceptible to response distortion (Barchard, 2001; Brackett & Mayer, 2003; S. A. Carroll & Day, 2004; Lopes, Salovey, & Straus, 2003; Lumley, Gustavson, Patridge, & Labouvie-Vief, 2005; Mayer et al., 2003). The MSCEIT also is incrementally valid in the prediction of better quality relationships among romantic partners (Brackett, Warner, & Bosco, 2005) and friends (Lopes et al., 2003, 2004); lower levels of drug and alcohol consumption and deviant behaviors among men (Brackett, Mayer, & Warner, 2004); important workplace outcomes, including stress management and leadership potential (Janovics & Christiansen, 2002; Lopes, Grewal, Kadis, Gall, & Salovey, in press); and lower levels of anxiety and depression (David, 2005; see Mayer et al., 2004, for a review). The goal of the studies presented here was to use theoretically derived self-report and performance measures of EI to examine the role of emotional abilities in the social behaviors expected to influence the quality of relationships. This relationship has yet to be examined because (a) reliable and valid performance measures

782

BRACKETT, RIVERS, SHIFFMAN, LERNER, AND SALOVEY

of EI have emerged only recently, and (b) there has been a lack of content-valid self-report EI measures. If valid, a theoretically derived self-report measure would make it possible to examine whether (a) EI can be detected with a self-report measure and (b) the relationship between self-report and performance measures of EI operate in a similar fashion to cognitive intelligence (e.g., Dunning, Johnson, Ehrlinger, & Kruger, 2003). Past research on EI has used self-report measures that have little to do with formal definitions of emotion and intelligence; consequently, these measures fail to conceptually and empirically map onto EI theory. Thus, we measured self-rated EI in ways that mapped onto a theoretically derived performance test of EI, the MSCEIT.

Overview of the Studies Three studies used Mayer and Salovey’s (1997) theory of EI as a framework to examine the role of emotional abilities in social functioning. As with any psychological construct, knowledge about EI is limited by our ability to operationally define it and validly measure it. In Study 1, we examined the relationship between self-rated and performance measures of EI. We then examined whether the EI measures were incrementally valid in the prediction of social behaviors, including perceived social competence (Study 2) and observable behaviors in a social encounter (Study 3). Because gender differences exist on many emotional abilities, we conducted all analyses separately for men and women. Women, for example, tend to outperform men on a variety of performance measures of emotional abilities (L. R. Brody & Hall, 1993, 2000), including the MSCEIT (Brackett & Mayer, 2003), perhaps because, as Fivush and colleagues have shown, parents tend to talk about emotions more with their daughters than with their sons (e.g., Adams, Kuebli, Boyle, & Fivush, 1995; Fivush, 1991, 1998; Fivush, Brotman, Buckner, & Goodman, 2000). There also is evidence for the presence of gender differences in the relationship between emotional abilities and relevant outcomes. Eisenberg et al. (1995) reported that emotion regulation was related to social functioning for boys but not for girls, and Brackett et al. (2004) reported that MSCEIT scores predicted social deviance (drug and alcohol use, aggressive acts) for men but not for women. In contrast, Custrini and Feldman (1989) showed that the ability to decode and encode emotions contributed to social competence for girls but not boys. There are few theoretical explanations for these differences offered in the literature. Shields (2002) suggested that behaviors often are interpreted on the basis of the gender of the actor. Indeed, Bacon and Ashmore (1985) found that parents categorize children’s social behaviors differently depending on the gender of the child. Thus, social functioning may be defined differently for men and women and for boys and girls. To account for this, in Studies 2 and 3, we selected assessments of social functioning that were expected to be less susceptible to gender differences in both occurrence and interpretation. For example, in Study 2, we selected responses to positive and negative events in close friend relationships as outcomes because men and women typically respond similarly when attempting to achieve comparable goals (Gable, Reis, Impett, & Asher, 2004; Rusbult, 1993).

Study 1 The purpose of Study 1 was to examine the relationship between self-rated and performance measures of EI by using instruments that tapped the same theoretical dimensions of EI. Because there are no self-report measures that map onto Mayer and Salovey’s (1997) model of EI and the MSCEIT, we developed the Self-Rated Emotional Intelligence Scale (SREIS). As a second assessment of self-rated EI, we also asked participants to estimate their performance on the MSCEIT. Previous work has demonstrated that performance and selfreport measures of mental abilities, such as verbal–propositional intelligence, are related only modestly (e.g., Paulhus, Lysy, & Yik, 1998). Because daily life provides little explicit feedback in the domain of human emotions, we predicted that the association between EI measures is weaker than the association between verbal ability measures, which we also collected. College students, for instance, may have some indication of their overall verbal ability because they receive feedback, such as their SAT scores and school performance; however, until the recent advent of social and emotional learning programs in schools and the workplace, few institutions have devoted time to developing, assessing, and thus providing feedback on emotional skills (Ciarrochi, Forgas, & Mayer, 2006). The first two hypotheses guiding Study 1 were as follows: 1.

Self-rated and performance tests of EI are weakly related.

2.

Participants are more accurate at estimating their verbal intelligence than their EI.

As noted above, gender differences in emotional abilities are reported in the literature; therefore, we expected to see gender differences in EI. Hence, the final hypothesis was as follows: 3.

MSCEIT scores are higher for women than men.

Method Participants Two hundred ninety-one undergraduates (65% female) at a state university participated for partial completion of a course requirement. Participants were primarily White, single, and heterosexual and ranged in age from 17 years to 29 years (M ⫽ 18.9, SD ⫽ 1.26). The majority of participants (72%) were in their 1st year at the university.

EI Measures MSCEIT. The MSCEIT, Version 2.0 (Mayer et al., 2002a), is a performance measure of EI that assesses how well people solve emotion-laden problems across four domains, including the perception, use, understanding, and management of emotions. The test contains 141 items that are divided among eight tasks (two for each of the four theoretical domains). The MSCEIT measures Perceiving Emotions by asking respondents to identify the emotions expressed in photographs of people’s faces (Faces) as well as the feelings suggested by artistic designs and landscapes (Pictures). Use of Emotion to Facilitate Thought is measured by two tasks that assess the ability to (a) describe emotional sensations in a cross-modality matching task involving nonfeeling vocabulary (Sensations) and (b) identify the feelings that might facilitate or interfere with the successful performance of various cognitive and behavioral tasks (Facilitation). Understanding Emo-

MEASURING EMOTIONAL INTELLIGENCE tions is measured by two tests that pertain to a person’s ability to analyze blended or complex emotions (Blends) and to understand how emotional reactions change over time or how they follow one another (Changes). Finally, Managing Emotions is measured by two tasks pertaining to the ability to manage one’s own emotions (Emotion Management) and the emotions of others (Social Management). The MSCEIT is scored with both consensus and expert scoring methods, which tend to converge (r ⬎ .90; Mayer et al., 2003). In consensus scoring, respondents are given credit for correct answers to the extent that their answers match those provided by the normative sample (over 5,000 heterogeneous individuals). Expert scoring relies on emotions experts (researchers) to indicate what they believe are the correct answers. Similar to consensus scoring, respondents receive credit for correct answers to the extent that they match those of the experts. The test publisher provides five scores, one for each domain, as well as a total EI score. Because consensus and expert scores were highly correlated, r(290) ⫽ .94, and there were no significant differences in correlations between consensus and expert scores and the criteria in the studies reported here, we arbitrarily used consensus scores. Also, in all three studies, we report analyses using the total MSCEIT score because of our focus on EI as an overall construct and not the individual abilities that comprise EI.1 Two criteria confirmed the decision to use the total score: (a) Confirmatory factor analysis of the MSCEIT supports the one-factor model (Mayer et al., 2003); and (b) in the present study, the part–whole correlations between the four dimension scores and the total MSCEIT score were high and statistically significant, rs(287) ⫽ .65 to .78. The split-half reliability coefficient for the total score was .94. For more information on the psychometric properties of the MSCEIT, see Mayer, Salovey, and Caruso (2002b; Mayer et al., 2003). SREIS. The SREIS was developed to map onto the emotional abilities measured by the MSCEIT. To develop the SREIS, we first examined and amended items from relevant scales, such as the Trait Meta-Mood Scale (Salovey, Mayer, Goldman, Turvey, & Palfai, 1995), and the self-report measure of EI by Schutte et al. (1998). For example, we used the item “By looking at their facial expressions, I recognize the emotions people are experiencing” from the Trait Meta-Mood Scale because it mapped onto the perception of emotion domain on the MSCEIT. We wrote additional items to cover all four EI domains adequately. Before administering the SREIS, 10 graduate students familiar with Mayer and Salovey’s (1997) model of EI rated the content validity of each item. Items for which there was less than 75% agreement were dropped, yielding a total of 34 items. The final scale included 9 items for Perceiving Emotions (e.g., “I can tell how people are feeling by listening to the tone of their voice”), 8 items for Using Emotions (e.g., “I can access my emotions/feelings in order to help me improve my problem solving abilities”), 8 items for Understanding Emotions (e.g., “It’s hard for me to describe my feelings” [reverse scored]), and 9 items for Managing Emotions (e.g., “I have difficulty managing my emotions” [reverse scored]). Participants rated each item on a response scale ranging from 1 (disagree strongly) to 5 (agree strongly). A preliminary factor analysis (principal axis with oblique rotation) of the 34-item scale suggested that our hypothesized four-factor solution was optimal. However, 6 items had factor loadings on unintended factors, and 3 items had loadings below 兩.30兩. These items were dropped, and the remaining 25 items were factor analyzed again. Items with factor loadings above 兩.35兩 on the pattern matrix were retained. Six items comprised each of the scales for the Perceiving, Using, and Managing Emotions domains, and 4 items comprised the Understanding Emotions domain. Because our primary interest was on the EI construct overall, we computed a total EI score by averaging across the scales. The part–whole correlations between the four dimension scores and the total SREIS score were high and statistically significant, rs(287) ⫽ .57 to .78, and the full scale was reliable (␣ ⫽ .84).

783

Estimated performance tasks. Participants estimated their performance on the MSCEIT both prior to and after completing the test. They estimated how well they would perform (or did perform) relative to all other students, all other male students, and all other female students at their university. For example, for the Understanding Emotions domain, participants responded to the question “I think I would perform [did perform] better than ____% of all other male students at the university on a test that measured my understanding of emotion concepts and the complexity of emotion” on an 11-point scale spaced in intervals of 10 (0%–100%). The estimates within each EI domain across the three comparisons correlated highly for both the pre- and postestimates (rs ⫽ .66 to .89, ps ⬍ .001). Thus, we created two total scores (pre- and postestimates of EI performance) by averaging responses across domains and comparison groups separately for both the pre- and postestimates. Both of these scales were highly reliable (␣s ⬎ .82).

Verbal Intelligence Measures Verbal SAT. Verbal SAT scores were used as a proxy measure of verbal intelligence. We obtained consent to access the participants’ (n ⫽ 228) scores from the university registrar. Self-rated verbal intelligence. A 10-item self-report scale (␣ ⫽ .83) measured self-rated verbal intelligence (Paulhus et al., 1998). Participants responded to each item (e.g., “I have a good vocabulary”) on a response scale ranging from 1 (very inaccurate) to 5 (very accurate). Estimated performance task. As with EI, participants compared themselves with their peers on their verbal intelligence. Participants indicated the extent to which their verbal SAT scores were better than all college students (in general), all students at their university, all other female students at their university, and all other male students at their university. For example, they responded to the question “I think my verbal SAT scores are better than ____% of all other male students at this university” on an 11-point scale spaced in intervals of 10 (0%–100%). Performance ratings for all four comparisons were correlated highly (rs ⫽ .81 to .92, ps ⬍ .001); thus, a total score was computed, which was highly reliable (␣ ⫽ .96).

Procedure All data were collected in one 75-min session. Participants completed the self-administered measures in groups of 25 to 50 in the following order: the self-rated EI and verbal intelligence measures, MSCEIT, post-MSCEIT estimate of performance, and demographics. Verbal SAT scores were collected from the college registrar after data collection.

Results Descriptive Statistics and Gender Differences Data were screened carefully for outliers and missing values. Four participants with extremely low MSCEIT total scores (⬍ 59) were dropped from all analyses. Descriptive statistics on the four EI measures are reported in Table 1. MSCEIT scores were comparable to other samples in the literature (Brackett et al., 2004; Mayer et al., 2002b). Mean scores on the SREIS were significantly above the midpoint on the 5-point scale, indicating that participants agreed that they possessed higher than average EI, t(286) ⫽ 18.79, p ⬍ .001. Mean scores on the estimated performance tasks also were significantly above the midpoint: preestimate, t(285) ⫽ 14.72; postestimate, t(285) ⫽ 11.88; ps ⬍ .001. Participants, on 1

Domain-level findings are available from Marc A. Brackett.

BRACKETT, RIVERS, SHIFFMAN, LERNER, AND SALOVEY

784

Table 1 Correlations Between EI Measures for Men (Above Diagonal) and Women (Below Diagonal) and Descriptive Statistics (Study 1) Measure

MSCEIT

SREIS

Pre-MSCEIT estimate

Post-MSCEIT estimate

MSCEIT SREISa Pre-MSCEIT estimateb Post-MSCEIT estimateb Sample mean (SD) Women’s mean (SD) Men’s mean (SD) Gender differences: t tests (␩2)

— .19** .12 .03 90.66 (12.57) 93.64 (11.80) 85.09 (12.10) 5.80*** (.105)

.27** — .42*** .37*** 3.46 (0.41) 3.44 (0.42) 3.49 (0.39) 1.03 (.004)

.23* .53*** — .67*** 62.23 (14.05) 60.90 (14.10) 64.72 (13.68) 2.21* (.017)

.24* .31*** .67*** — 60.01 (13.97) 58.00 (14.20) 63.64 (12.86) 3.26*** (.037)

Note. Total sample: 275 ⬍ N ⬍ 286; men: 98 ⬍ n ⬍ 100; women: 177 ⬍ n ⬍ 186. EI ⫽ emotional intelligence; MSCEIT ⫽ Mayer–Salovey–Caruso Emotional Intelligence Test; SREIS ⫽ Self-Rated Emotional Intelligence Scale. a Responses ranged from 1 (disagree strongly) to 5 (agree strongly). b Responses ranged from 0 (I would perform better than 0% of others) to 100 (I would perform better than 100% of others), in 10-point increments. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.

average, predicted that they would (and did) perform better than about 60% of their peers. Indeed, nearly 80% of the participants believed that they would (or did) perform above the 50th percentile on the MSCEIT. To examine gender differences on the EI measures, we conducted a 2 (gender) ⫻ 4 (test type: MSCEIT, SREIS, and pre- and postestimates) repeated measures multivariate analysis of variance (MANOVA) after standardizing the test scores. There were significant differences in men’s and women’s scores on the four EI measures, F(3, 816) ⫽ 23.27, p ⬍ .001, as shown in Table 1. Follow-up analyses showed that the gender differences were significant for three of the four EI measures. Consistent with our hypotheses and previous research (Brackett et al., 2004; Mayer et al., 1999), our results indicated that women scored significantly higher than men on the MSCEIT. On the two estimated performance tasks (both before and after taking the MSCEIT), men’s scores were significantly higher than women’s scores. There were no significant gender differences in SREIS responses. With respect to verbal ability, SAT scores ranged from 340 to 750 (M ⫽ 541.53, SD ⫽ 70.60). Scores for men (M ⫽ 547.12, SD ⫽ 73.72) and women (M ⫽ 538.90, SD ⫽ 69.15) were not significantly different, t(226) ⬍ 1. Participants were slightly above the midpoint on the self-reported verbal intelligence scale (M ⫽ 3.19, SD ⫽ 0.68). There were no significant gender differences in self-ratings of verbal ability (women, M ⫽ 3.24, SD ⫽ 0.67; men, M ⫽ 3.09, SD ⫽ 0.69), t(284) ⫽ 1.78, p ⬎ .05. Means on the estimated performance task indicated that participants believed their verbal intelligence to be slightly above average (M ⫽ 55.84, SD ⫽ 18.07) and significantly greater than 50% of their peers, t(284) ⫽ 5.46, p ⬍ .001. Consistent with previous research (Bailey & Mettetal, 1977; Bennett, 1996), our results showed that men (M ⫽ 62.38, SD ⫽ 18.42) had significantly higher self-rated verbal intelligence than women (M ⫽ 52.31, SD ⫽ 16.90), t(283) ⫽ 4.65, p ⬍ .001.

Relationship Between Performance Tests and Self-Ratings EI. Consistent with our hypotheses, our results showed that EI self-ratings were not related strongly to performance on the MSCEIT. Although the correlation between the MSCEIT and the SREIS was significant, the relationship was not strong, r(287) ⫽

.19, p ⬍ .01. MSCEIT scores were not related to scores on the estimated performance tasks, r(286) ⫽ .10 and r(275) ⫽ .03, respectively; ps ⬎ .05. The correlations between the SREIS and the two estimated performance measures were significant, rs(275) ⫽ .35 to .46, ps ⬍ .001. There were no significant gender differences in the strength of these correlations (Fischer’s zs ⬍ 1.96), as shown in Table 1. Because individuals who score higher in EI (MSCEIT) may have a more accurate perception of their emotional abilities, we assigned participants to the bottom, second, third, or top quartile on the basis of their MSCEIT test performance (following Dunning et al., 2003) and conducted a 4 (EI scale: MSCEIT, SREIS, and pre- and postestimate) ⫻ 4 (quartile) repeated measures MANOVA. There was a significant Quartile ⫻ Measure interaction, F(9, 810) ⫽ 36.34, p ⬍ .001. Follow-up analyses showed that individuals with higher EI scores were not more accurate in their self-ratings; individuals in the bottom two quartiles overestimated their performance on the MSCEIT on all three self-ratings, and those in the upper two quartiles underestimated their MSCEIT performance. Verbal intelligence. As predicted, participants were, in general, more accurate in estimating their verbal intelligence than their EI. The two self-report indices of verbal intelligence were related significantly to verbal SAT scores: for self-rated verbal intelligence, r(228) ⫽ .43, p ⬍ .001, and for the estimated performance measure, r(226) ⫽ .53, p ⬍ .001. There were no significant gender differences in the strength of these correlations. These correlations were somewhat higher than those found in other studies, which generally yield correlations in the .30 –.35 range (Paulhus et al., 1998).

Discussion Self-rated and performance measures of EI were not strongly related, suggesting that perceptions of one’s EI may not be an accurate indicator of EI and that these measures are most likely tapping into different mental processes. These findings can be interpreted in the context of more extensive studies showing that people are notoriously bad at both self-reporting their mental abilities and estimating their own performance on ability tests

MEASURING EMOTIONAL INTELLIGENCE

(Bailey & Mettetal, 1977; Dunning et al., 2003; Mabe & West, 1982; Paulhus et al., 1998; Reilly & Mulhern, 1995). There are a number of possible explanations for the lack of correspondence between the EI measures. First, self-reports are prone to social desirability response biases (Paulhus, 1991). Second, an individual’s level of EI may influence the self-ratings. That is, similar to individuals with lower intelligence, individuals with lower EI may lack the metacognitive skills to report on their EI. Highly emotionally intelligent people, on the other hand, may be inaccurate because they overestimate the EI of others (Dunning et al., 2003). It also is possible that chronic self-views of EI interfere with the ability to estimate EI accurately (Ehrlinger & Dunning, 2003). The strong correlation between the two self-rating tasks (SREIS scores and estimates of MSCEIT performance) provides some support for this possibility. Finally, individuals may not have preconceived notions about their EI. This may explain why the relationship between self-rated EI and performance EI was weaker than the relationship between self-rated and performance measures of verbal intelligence. Compared with EI, verbal–propositional intelligence is more highly institutionalized, thus providing individuals with more opportunities for feedback in this domain. It is important to mention, however, that we compared participants’ self-reported verbal intelligence with their verbal SAT scores, which they had taken previously; we did not administer a verbal ability test during the testing session. With regard to gender, we replicated previous studies indicating that women perform better than men on the MSCEIT (e.g., Brackett & Mayer, 2003). This finding also supports other research on emotional abilities showing the women are more skilled in the emotions domain than are men. For example, there is evidence that women use a more varied emotions vocabulary (e.g., Adams et al., 1995; Fivush et al., 2000) and are better than men at reading nonverbal behaviors, including facial expressions of emotions (e.g., Hall, 1978, 1984; McClure, 2000; Rosenthal, Hall, DiMatteo, Rogers, & Archer, 1979). On the EI self-ratings, gender differences emerged for the estimated performance tasks but not for the SREIS. Men’s estimated MSCEIT performance was significantly higher than women’s. This effect held even after participants took the MSCEIT and even though women outperformed men on the MSCEIT. This result is not necessarily surprising as women tend to underestimate their abilities in other achievement settings, whereas men tend to overestimate theirs (Lenney, 1977; T. Roberts, 1991). This is especially so when performance criteria are unclear (Lenney, 1977), as may well be the case with emotional abilities. In sum, this study showed that people are particularly poor at both providing self-reports and estimating their performance on ability measures of EI, indicating that self-rated EI may not be a good proxy assessment of ability EI. Although self-rated and performance measures of EI are relatively distinct, each may contribute to understanding of the role of emotional abilities in social functioning. This question is explored in the next two studies.

785

global relationship quality among friends (Lopes et al., 2004) and romantic partners (Brackett et al., 2005). In this study, we moved beyond global relationship quality assessments and examined whether self-rated and performance measures of EI were related to perceived social competence with friends. Specifically, we examined the relationship between EI and the strategies that individuals reported they used in response to positive and negative emotions in relationships. These two contexts were selected for three reasons: (a) Each requires the use of emotional abilities; (b) previous research on responses to positive events and conflict showed no gender differences (Gable et al., 2004, and Rusbult, 1993, respectively); and (c) responses to positive events and conflict can be classified as effective or ineffective with regard to relationship well-being (Gable et al., 2004; Rusbult, Verrete, Whitney, Slovic, & Lipkus, 1991). Because the validity of ability and self-report measures of EI to predict important outcomes above and beyond well-studied measures of personality, well-being, and intelligence has been questioned (e.g., N. Brody, 2004; Hedlund & Sternberg, 2000; McCrae, 2000), we examined the relationship between the EI measures and social competence outcomes by controlling for these variables. Unlike self-report measures based on popularizations of EI, we did not expect that the SREIS would correlate too highly with these other indices because it is based on EI theory. Nevertheless, as a self-report measure, the SREIS may share some semantic content and method variance with existing measures, such as Neuroticism.2 Thus, we made the following hypothesis: 1. The MSCEIT is mostly uncorrelated with measures of personality, well-being, and verbal intelligence, whereas the SREIS is moderately correlated with measures of personality and wellbeing, but not verbal intelligence. There is a general consensus that performance tests (as opposed to self-report scales) are the gold standard in intelligence research because they measure the actual capacity to perform well at mental tasks, not just one’s self-efficacy about certain skills (J. B. Carroll, 1993). Because it is likely that a person’s actual knowledge and reasoning ability about emotions, in contrast to perceived ability, contribute to effective social functioning, we tested a second hypothesis: 2. The MSCEIT, but not the SREIS, is associated with perceived interpersonal strategies after personality, well-being, and verbal intelligence are held constant. More specifically, the MSCEIT is correlated positively with constructive responses and negatively with destructive responses to both relationship problems and positive events.

Method Participants Three hundred fifty-five undergraduates (61% female) at a private research university participated in this study for partial completion of a course requirement. The group of participants was 58% White, 19% Asian, 9% African American, 6% Hispanic, and 7% other. Participants were primarily single and heterosexual and ranged in age from 18 years to 34

Study 2 Emotional abilities help individuals to form and maintain functional interpersonal relationships (Keltner & Haidt, 2001). For example, emotional abilities are associated with perceptions of

2 Because the results were comparable across the two self-rating assessments (SREIS, performance estimates), only the SREIS was used in Studies 2 and 3.

786

BRACKETT, RIVERS, SHIFFMAN, LERNER, AND SALOVEY

years (M ⫽ 20.13, SD ⫽ 2.90). The majority of participants (72%) were in their 1st year at the university.

Measures of EI MSCEIT. An online version of the MSCEIT, described in detail in Study 1, was used. Prior research has suggested that booklet and online forms of the MSCEIT produce indistinguishable scores (Mayer et al., 2003). Split-half reliability of the total score was .91. SREIS. A revised 19-item measure of the SREIS was used (Brackett, 2004; see Appendix). The revised scale corrected for ambiguous statements contained in the original SREIS. For instance, we changed the statement “It’s hard for me to describe my feelings” from the Understanding Emotions domain to “I could easily write a lot of synonyms for words like happiness or sadness” in order to represent the contents of the MSCEIT more accurately. A confirmatory factor analysis of the revised SREIS supported both one- and four-factor solutions. Thus, there is converging evidence that the four basic dimensions of EI can be detected with both self-report and performance tests, which both load on one hierarchical factor of EI. Participants indicated the extent to which the 19 randomly ordered statements on the SREIS accurately described them by using a 5-point Likert scale, ranging from 1 (very inaccurate) to 5 (very accurate). Cronbach’s alpha for the total scale was .77.

Personality Measures Participants used a response format ranging from 1 (strongly disagree) to 5 (strongly agree) for each personality measure. Revised NEO Personality Inventory. Personality traits were assessed with the 240-item Revised NEO Personality Inventory (Costa & McCrae, 1992), which measures five global dimensions of personality: Neuroticism, Extraversion, Openness to Experience, Agreeableness, and Conscientiousness (all ␣s ⱖ .88). Empathy. Empathy was assessed with the Mehrabian and Epstein (1972) 19-item scale (␣ ⫽ .75). Psychological well-being. Psychological well-being was assessed with Ryff’s (1989) theoretically based self-report inventory, which contains 36 items to measure six dimensions: self-acceptance, environmental mastery, purpose in life, positive relations with others, personal growth, and autonomy. The reliability of the full scale was high (␣ ⫽ .90). Subjective well-being. Subjective well-being was measured with the 5-item Satisfaction With Life Scale (Diener, Emmons, & Larsen, 1985; Pavot & Diener, 1993). The reliability of this scale was high (␣ ⫽ .84).

Social Competence Outcomes Perceived interpersonal strategies to positive events. Responses to positive events happening to another person were measured with a modified version of the Perceived Responses to Capitalization Attempts Scale (Gable et al., 2004). This scale assesses the extent to which respondents engage in various behaviors in reaction to a friend sharing a positive event (e.g., friend receives a good grade or has a great conversation with a potential love interest). Because the measure was developed originally for couples, the term partner was replaced with the combined term roommate/ suitemate/close friend, and the events were altered slightly to be relevant to school. Participants rated each item with the stem “When my roommate/ suitemate/close friend tells me about something good that has happened to him/her.” The scale contained 12 items classified into four categories: active constructive (e.g., “I usually react to this person’s good fortune enthusiastically”), active destructive (e.g., “I often find a problem with it”), passive constructive (e.g., “I say very little, but I am happy for this person”), and passive destructive (e.g., “I don’t pay much attention”). Cronbach’s alphas ranged from .50 to .72.

Perceived interpersonal strategies to negative events. Responses to negative events occurring in a relationship with a roommate/suitemate/ close friend were assessed with a 16-item scale adapted from Rusbult, Johnson, and Morrow’s (1986) Accommodation Scale. This scale assesses the extent to which individuals engage in various strategies during a conflict in a close relationship. The items were divided into four categories: active constructive (e.g., “When this person and I have problems, I discuss things with him/her”), passive constructive (e.g., “When this person and I are angry with each other, I give things some time to cool off on their own rather than taking action”), active destructive (e.g., “When this person and I have a disagreement, I end up screaming at him/her”), and passive destructive (e.g., “When I am annoyed at this person, I avoid spending time with him/her”). Cronbach’s alphas ranged from .64 to .85.

Procedure Participants completed all the measures during two 1-hr sessions, except for the MSCEIT, which was completed online before the other scales.

Results Descriptives Performance and self-rated EI measures. Mean scores on the MSCEIT were comparable to Study 1 (M ⫽ 97.55, SD ⫽ 10.63). Consistent with Study 1, women (M ⫽ 99.05, SD ⫽ 10.31) had significantly higher MSCEIT scores than did men (M ⫽ 95.00, SD ⫽ 10.72), t(346) ⫽ 3.49, p ⬍ .01, ␩2 ⫽ .034. As in Study 1, mean scores on the SREIS were significantly higher than the midpoint, t(349) ⫽ 27.17, p ⬍ .001, indicating that participants had inflated self-ratings. In contrast to Study 1, women’s selfratings (M ⫽ 3.75, SD ⫽ 0.41) were significantly higher than men’s (M ⫽ 3.59, SD ⫽ 0.38), t(346) ⫽ 3.67, p ⬍ .001, ␩2 ⫽ .038. Finally, as predicted, MSCEIT and SREIS scores were unrelated, r(327) ⫽ .07, p ⬎ .05, confirming that self-report and performance measures likely are tapping into different psychological processes. Social competence. A MANOVA showed that men and women reported significantly different responses on each of the subscales, F(8, 353) ⫽ 8.85, p ⬍ .001. As shown in Table 2, women were more likely than men to report using active constructive responses to positive events, but men were more likely than women to report using passive destructive, passive constructive, and active destructive responses, Fs(1, 360) ⬎ 9.0, ps ⬍ .01. In response to negative relationship events, women were more likely than men to use active constructive, passive destructive, and passive constructive responses, Fs(1, 360) ⬎ 4.0, ps ⬍ .05. Men and women did not differ in their use of active destructive responses to negative events, F(1, 360) ⬍ 1.0.

Associations Among Measures of EI, Personality, WellBeing, and Verbal Intelligence Significant correlations between MSCEIT scores and the personality variables ranged from 兩.11兩 to 兩.24兩, whereas for the SREIS they ranged from 兩.11兩 to 兩.54兩 (see Table 3). Among the Big Five, the differences between the MSCEIT-personality and SREISpersonality correlations were statistically significant for Extraversion, Openness, and Agreeableness (Fischer’s zs ⬎ 1.96). There were no significant differences in the sizes of the correlations for men and women (Fischer’s zs ⬍ 1.96). To gain a more comprehensive perspective on these associations, we performed two sep-

a

Note. EI ⫽ emotional intelligence; MSCEIT ⫽ Mayer–Salovey–Caruso Emotional Intelligence Test; SREIS ⫽ Self-Rated Emotional Intelligence Scale. Big Five, psychological well-being, empathy, life satisfaction, and Verbal SAT scores held constant. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.

.01 ⫺.07 ⫺.07 ⫺.01 .06 ⫺.15 ⫺.04 ⫺.11 ⫺.08 ⫺.13 .05 .00 .25** ⫺.15 ⫺.17* ⫺.28** 3.38 (0.83) 1.60 (0.57) 3.91 (0.60) 2.91 (0.82) 3.21 (0.76) 1.64 (0.60) 3.65 (0.67) 2.63 (0.77)

.15 ⫺.23** ⫺.03 ⫺.30***

⫺.08 ⫺.17* .07 ⫺.05

.20** ⫺.09 ⫺.05 ⫺.13

.09 ⫺.22* ⫺.02 ⫺.27**

.11 ⫺.03 ⫺.10 ⫺.06 .12 ⫺.04 .00 ⫺.21* ⫺.14 ⫺.08 ⫺.04 ⫺.03 .29*** ⫺.18* ⫺.29*** ⫺.33*** 3.75 (0.64) 1.50 (0.53) 2.12 (0.78) 1.44 (0.56)

Positive events Active constructive Active destructive Passive constructive Passive destructive Negative events Active constructive Active destructive Passive constructive Passive destructive

3.46 (0.64) 1.75 (0.64) 2.39 (0.80) 1.67 (0.64)

.16 ⫺.34*** ⫺.22* ⫺.38***

⫺.06 ⫺.14* ⫺.10 ⫺.10

.29*** ⫺.15* ⫺.35*** ⫺.21**

.10 ⫺.33*** ⫺.23* ⫺.33***

Women (n ⫽ 180) Men (n ⫽ 99) Women (n ⫽ 180) Men (n ⫽ 98) Women (n ⫽ 210–214) Men (n ⫽ 129–131) Women (n ⫽ 213–217) Men (n ⫽ 127–129) Women (n ⫽ 219) Men (n ⫽ 143) Perceived response

SREIS M (SD)

MSCEIT

Zero-order

Table 2 Zero-Order and Partial Correlations Between EI Scales and Perceived Responses to Interpersonal Events (Study 2)

MSCEIT

Partiala

SREIS

MEASURING EMOTIONAL INTELLIGENCE

787

arate multiple regression analyses by using all of the personality and verbal intelligence scores as predictor variables and the two EI tests as outcome measures for the full sample. The multiple correlation for the standard regression with the MSCEIT as the outcome was .30 (adjusted R2 ⫽ .06). Consistent with previous work, our results showed that Agreeableness (␤ ⫽ .18) and psychological well-being (␤ ⫽ .24) were related to total MSCEIT scores ( ps ⬍ .01; Brackett & Mayer, 2003; Lopes et al., 2003). The multiple R for the regression with the SREIS as the outcome variable was .62 (adjusted R2 ⫽ .36). In this analysis, Extraversion (␤ ⫽ .21), Openness to Experience (␤ ⫽ .32), and psychological well-being (␤ ⫽ .40) were significant predictors ( ps ⬍ .001) of total SREIS scores. Thus, compared with the MSCEIT, there was more overlap between the SREIS and the personality variables.

Predictive and Incremental Validity of EI Measures Table 2 presents the zero-order and partial correlations (controlling for personality, psychological well-being, empathy, life satisfaction, and verbal SAT) between the two EI measures and responses to positive and negative events. Because our central question pertained to the incremental validity of the EI measures, we focused on the partial correlations as opposed to the zero-order correlations. MSCEIT scores correlated significantly with perceived interpersonal strategies, but only for men. For men, MSCEIT scores were correlated negatively with the tendency to use both active and passive destructive responses and passive constructive responses to positive interpersonal events, prs(98) ⫽ ⫺.23 to ⫺.33, ps ⬍ .05. For women, the MSCEIT was associated (negatively) with just one response to positive events: active destructive strategies, r(216) ⫽ ⫺.14, p ⬍ .05, which dropped to nonsignificance after personality and verbal intelligence were held constant. The strength of the associations between the MSCEIT and both active and passive destructive responses to positive events was significantly different for men and women (Fischer’s zs ⫽ 2.07 and 2.46, respectively). A similar pattern of findings emerged for perceived responses to negative events. For men, MSCEIT scores were associated negatively with both active and passive destructive strategies, prs(98) ⫽ ⫺.22 and ⫺.27, respectively; ps ⬍ .05. For women, the MSCEIT correlated negatively with just one outcome: active destructive strategies, r(214) ⫽ ⫺.17, p ⬍ .05, which also dropped to nonsignificance after personality and verbal intelligence were held constant. The strength of the associations between the MSCEIT and passive destructive responses to negative events was significantly different for men and women (Fischer’s z ⫽ 2.18). After controlling for personality and intelligence measures, we found that the SREIS correlated significantly with just one of the scales for men: passive destructive strategies in response to positive events, pr(98) ⫽ ⫺.21, p ⬍ .05. There were no significant gender differences in the strength of the partial correlations of the SREIS with these outcomes. Finally, for men, the correlation between the MSCEIT and active destructive responses to positive events was significantly stronger than the correlation between the SREIS and this response (Fischer’s z ⫽ 2.09).

Discussion First, we demonstrated that the MSCEIT and the SREIS are not significantly correlated, replicating Study 1. Then, we supported our

BRACKETT, RIVERS, SHIFFMAN, LERNER, AND SALOVEY

788

Table 3 Relationship Between EI Measures and Personality Variables (Study 2) MSCEIT

SREIS

Measure

All (316 ⬍ N ⬍ 347)

Men (114 ⬍ n ⬍ 129)

Women (202 ⬍ n ⬍ 218)

All (311 ⬍ N ⬍ 336)

Men (115 ⬍ n ⬍ 131)

Women (196 ⬍ n ⬍ 215)

Big Five traits Neuroticism Extraversion Openness to Experience Agreeableness Conscientiousness Empathy Psychological well-being Life satisfaction Verbal SAT

⫺.13* .03 .04 .19*** .15** .13* .19*** .12* ⫺.05

⫺.25** .23** ⫺.09 .15 .04 .01 .25** .11 ⫺.18

⫺.12 ⫺.12 .05 .17* .20** .07 .16* .13 .02

⫺.21*** .43*** .41*** .04 .19*** .31*** .46*** .22*** .04

⫺.22* .44*** .39*** .05 .11 .32*** .48*** .15 .11

⫺.25*** .42*** .37*** ⫺.06 .22** .24** .45*** .25** ⫺.03

Note. EI ⫽ emotional intelligence; MSCEIT ⫽ Mayer–Salovey–Caruso Emotional Intelligence Test; SREIS ⫽ Self-Rated Emotional Intelligence Scale. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.

hypothesis that the MSCEIT did not overlap significantly with existing personality measures, such as the Big Five, which also corroborates earlier findings (Mayer et al., 2004). Also, as predicted, the SREIS was more highly correlated with measures of personality than the MSCEIT. The greater overlap was expected because of the inherent shared method variance between the measures and the similar semantic content between items on the SREIS (e.g., Management of Emotions Scale) and the personality scales (e.g., Neuroticism). Our main goal was to examine the incremental validity of both measures in predicting perceived social competence with friends. Consistent with our hypotheses, our results indicated that the MSCEIT, but not the SREIS, was incrementally valid. However, this was the case for men only. Men with lower MSCEIT scores were more likely to use both passive and active destructive strategies in response to both relationship conflict and others’ reports of positive events. It is unclear why the MSCEIT predicted social competence outcomes only for men. However, this finding is not unique, as others have reported relationships between emotional abilities and social competence for only one gender (e.g., Brackett et al., 2004; Custrini & Feldman, 1989; Eisenberg et al., 1995). Shields (2002) and Maccoby (1998) suggested that emotions play a different role in the social interactions of men and women (and boys and girls) to the extent that the genders occupy different emotional worlds. Accordingly, emotional skills may operate differently in the social worlds of each gender. In this study, we used self-report scales of social competence that have not shown gender differences in previous research (Gable et al., 2004; Rusbult, 1993); however, there were significant gender differences on these scales in our sample. Thus, our operationalization of social competence may in fact be different for men and women. In Study 3, we attempted to better capture social competence by measuring real-time social behaviors.

Study 3 The aim of Study 3 was to assess whether the MSCEIT and SREIS predicted observable behaviors in a social encounter, namely, interacting with an ostensible stranger in a gettingacquainted meeting. We expected that scores on the MSCEIT, but not the SREIS, would predict social success. Because definitions

of social success are contingent on the expectations, norms, and roles of a situation, and typically vary with the goals, beliefs, and motives of a given observer, valid operationalizations of social success often are obscured by its abstract, multifarious nature (Topping, Bremner, & Holmes, 2000). With this in mind, we used an assessment strategy designed to capture multiple facets of social success. These assessments included both key player evaluations (confederate and participant) and naive observer ratings. In Study 2, the relationship between EI and social competence was significant for men only perhaps because there were gender differences on the social competence outcome. To minimize the presence of gender differences in Study 3, we carefully structured interaction so that participants approached it with the same specified goal. Gender differences in behavior are less likely to occur when the context is held constant (Christensen & Heavey, 1990).

Method Participants Fifty students (28 women and 22 men) from a liberal arts college participated in exchange for course credit or monetary payment ($8). The group of participants was 74% White, 10% Asian, 4% African American, and 4% Hispanic. The average age was 19.3 years (SD ⫽ 0.91).

Procedure The study consisted of two sessions. In the first session, participants completed the MSCEIT and the SREIS. One week later, participants returned individually to the laboratory under the auspices of performing a task with another participant. After first completing a personality assessment (described below), participants were told by the experimenter that they would be doing a group task that was “designed to measure how well two people work together.” The experimenter informed participants that “groups are most successful in the task when the partners know a little bit about one another.” This description was intended to provide participants with a clear goal in the interaction (i.e., to become personally acquainted with the confederate). Finally, the experimenter instructed the participant to sit with his or her partner (actually a confederate) in the waiting room while the group task was being set up. When participants first entered the waiting room, a same-sex confederate behaved according to a script requiring the confederate to greet the par-

MEASURING EMOTIONAL INTELLIGENCE ticipant and then wait for the participant to initiate conversation. If the participant did not initiate conversation within 3 min, the confederate began a conversation. Confederates were instructed to respond the way they normally would during a social interaction but to let the participants lead the conversation. The extensively trained confederates were blind to the purpose and hypotheses of the study. Two hidden video camcorders recorded the interactions. After 6 min, the experimenter entered the waiting room and asked participants to complete an interaction evaluation measure (described below) in a room separate from the confederate. Finally, the experimenter explained the true purpose of the experiment, the need for deception, and the use of hidden video cameras. Participants were given the opportunity to have their tape recordings erased. Only 1 participant made this request.

Materials EI measures. EI was assessed with the MSCEIT and the SREIS. Participants completed the online version of the MSCEIT, discussed in detail in Study 1. Split-half reliability of the total score was .89. The revised 19-item SREIS, described in Study 2, was used to assess perceived EI. Cronbach’s alpha for the full scale was .66. Personality assessment. Personality was assessed with the 54-item Big Five Inventory (Johnson, Donahue, & Kentle, 1991), which measures Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. Cronbach’s alphas for the five subscales ranged from .77 to .88. Participant self-evaluations of social competence. Participants evaluated their behavior during the waiting room interaction with the confederate by indicating their agreement with six items: (a) “I am satisfied with my actions,” (b) “My actions were appropriate for the context of the situation,” (c) “I made it easy for the other participant to talk to me,” (d) “I was genuinely involved in the conversation,” (e) “My actions were an accurate representation of how I normally behave,” and (f) “I would change something about the way I behaved” (reverse scored). They scored their responses on a 7-point Likert-type scale ranging from 1 (completely disagree) to 7 (completely agree). These six items were reliable (␣ ⫽ .82) and were averaged to form an overall self-evaluation score, with higher scores indicating more positive self-evaluations. Confederate evaluations of social competence. Because the central objective of the waiting room interaction was for the participants to get to know the confederate, confederates first responded to the item “This participant seemed interested in me” by using a 7-point scale ranging from 1 (completely disagree) to 7 (completely agree). Confederates then reported the extent to which they agreed with 11 items evaluating the social engagement, friendliness, likeability, confidence, and competence of each participant by using the same 7-point scale. The 11 items were highly reliable (␣ ⫽ .93) and were averaged together to form a confederate rating of social competence.

Video Analysis Four judges (two men and two women) who were blind to the purpose, hypotheses, and EI of the participants reviewed the interactions. The judges independently rated each participant on four dimensions: (a) social engagement, (b) comfort level, (c) ability to work well with others, and (d) overall social competence. Judges rated level of social engagement by taking into account the amount of personal information discussed, efforts to prolong or continue the conversation, and the amount of times participants made eye contact with the confederate on a 5-point Likert-type scale ranging from 1 (not at all engaged) to 5 (extremely engaged). Participants who responded in full sentences, asked questions, and kept their gaze focused on the confederate received the highest possible scores. For the comfort rating, judges rated body movements, speech patterns, hand gestures, and overall posture on a 5-point Likert scale ranging from 1 (not at all comfortable) to 5 (extremely comfortable). Participants who did not fidget, spoke in an even rhythm, and

789

smiled frequently received the highest possible scores. The social engagement and comfort ratings were made at 90-s intervals yielding four separate judgments for each interaction and each judge. Because judges’ ratings were made on interval scales, interrater reliability was assessed with Cronbach’s alpha (Harris, 2001). Alphas ranged from .71 to .96 for the four social engagement ratings across the raters and from .63 to .82 for the four comfort ratings across the raters. Reliability across the four time and judge ratings for social engagement and comfort was very high (␣s ⫽ .97 and .89, respectively). Scores were averaged across time and raters to form composite scores for social engagement and comfort. Overall competence and team player ratings were made at the conclusion of each interaction. Judges evaluated overall competence by responding to the following: “The goal of this interaction was for participants to get to know one another. How successful was this participant in accomplishing that goal?” Responses were scored on a 5-point Likert-type scale ranging from 1 (not at all successful) to 5 (extremely successful). Judges evaluated the extent to which each participant was a team player by responding to the question “How confident are you that this participant would work well collaborating with others; is s/he a ‘team player’?” Responses were scored on a 5-point Likert-type scale ranging from 1 (not at all confident) to 5 (extremely confident). Interrater reliability was high for overall competence and team player ratings (␣s ⫽ .96 and .92, respectively). Scores were averaged across judges to form a total score for each judgment. We also measured the duration of conversation for each interaction by measuring the total conversation length and dividing it by the total duration of the interaction. Conversation duration was defined as any moment when either the participant or the confederate was speaking.

Results Descriptives Consistent with Studies 1 and 2, our results indicated that on the MSCEIT, women (M ⫽ 104.99, SD ⫽ 9.63) scored significantly higher than men (M ⫽ 95.31, SD ⫽ 11.46), t(49) ⫽ 10.54, p ⬍ .01, ␩2 ⫽ .18. On the SREIS, men (M ⫽ 3.65, SD ⫽ 0.29) and women (M ⫽ 3.55, SD ⫽ 0.47) did not differ significantly, t(49) ⬍ 1.00. As predicted, MSCEIT and SREIS scores were unrelated, r(50) ⫽ .03. Confederate and participant evaluations were comparable for both men and women. Gender differences emerged for three of the four dimensions, whereby judges rated women as more socially engaged, more competent, and better able to work well with others, ts(49) ⫽ 2.56 to 2.99, ps ⬍ .01. Women (M ⫽ 69.00, SD ⫽ 29.55) talked significantly more during the interaction than did men (M ⫽ 47.15, SD ⫽ 41.32), t(44) ⫽ 2.09, p ⬍ .05. Overall there were strong, positive correlations between the confederates’ and judges’ ratings. Participants who were liked more by the confederates and who were rated as more interested by the confederates also were rated by the judges as more socially engaged, more competent, and better able to work with others, rs(44) ⫽ .47 to .60. Participants who were rated as more socially competent by the confederate also reported being satisfied by their performance in the interaction, r(50) ⫽ .37. Participants who were satisfied with their performance in the interaction also were rated by the judges as more socially engaged and competent and better able to work well with others, rs(44) ⫽ .46 to .50. The only lack of correspondence was with the comfort ratings; judges’ and confederate ratings of comfort were uncorrelated, r(44) ⫽ ⫺.06.

Predictive and Incremental Validity of EI Measures Table 4 displays the zero-order correlations between both EI measures and the social competence variables. Because of the

BRACKETT, RIVERS, SHIFFMAN, LERNER, AND SALOVEY

790

Table 4 Zero-Order Correlations of the MSCEIT and the SREIS With the Social Behaviors (Study 3) MSCEIT Ratings

Men

Confederate and participant ratings Confederate overall rating Confederate interest rating Participant self-evaluation Judges’ ratings Social engagement Comfort Team player Overall social competence

SREIS

Women

Men

Women

.20 .48* .13

.31 .27 ⫺.14

⫺.41 ⫺.41 .01

.35 .19 .16

.47† ⫺.16 .53* .51*

⫺.02 ⫺.19 ⫺.16 ⫺.03

⫺.34 .08 ⫺.44 ⫺.36

.28 .03 .19 .27

Note. N ⫽ 44 to 50 (women ⫽ 26 –28; men ⫽ 18 –22); Ns vary due to technical difficulties with the video recordings. MSCEIT ⫽ Mayer– Salovey–Caruso Emotional Intelligence Test; SREIS ⫽ Self-Rated Emotional Intelligence Scale. † p ⬍ .10. * p ⬍ .05.

small sample size for each gender, we examined scatterplots for all of the correlations; there was no evidence that individual outliers were driving any of the effects. Scores on the MSCEIT were associated with performance in the interaction, but only for men. Specifically, men with higher EI were more likely than men with lower EI to be rated as (a) showing greater interest in the confederate, (b) more socially engaged, (c) more socially competent, and (d) being a team player. These findings remained significant after statistically controlling for the Big Five (which were unrelated to the outcomes) in multiple regression analyses. Comparing correlations of the MSCEIT with these variables for men and women showed that there was a significant difference for being a team player (Fischer’s z ⫽ 2.37). As predicted, the SREIS did not correspond significantly with any of the social competence measures, indicating that beliefs about one’s emotional abilities were not significantly correlated with social success. For men, there were significant differences in the strength of the correlations between the MSCEIT and the SREIS, including confederates’ overall and interest ratings, and the judges’ ratings of social engagement, whether a participant was a team player, and overall social competence (Fischer’s zs ⬎ 1.96).

Discussion By evaluating observable behaviors in a social interaction, we provided additional evidence supporting the incremental validity of performance measures of EI. The MSCEIT, but not the SREIS, was associated with social competence, but only for men. When interacting with the confederate, men with higher MSCEIT scores were judged as more socially engaged and socially competent by both the confederate and naive judges. Finally, although not statistically significant, there were a number of counterintuitive negative correlations between the SREIS and the positive social outcomes for men (rs ⬎ ⫺.34). This study extends the findings of Study 2 by showing that performance on a task-based measure of EI predicts social competence for men who are not only in already formed social relationships but are establishing new social relationships. The reasons

why the MSCEIT is associated with social competence for men only are unknown; a more thorough discussion of this issue will be presented in the General Discussion.

General Discussion The present investigation yielded two primary findings: (a) Self-ratings of EI, as assessed by the SREIS, and performance measures of EI, as assessed by the MSCEIT, were not strongly correlated; and (b) after statistically controlling for personality, the MSCEIT was associated with interpersonal competence for men, whereas the SREIS was generally unrelated to social competence for both genders. There are a number of possible explanations for why self-report and performance measures of EI are less correlated than parallel measures of verbal intelligence. In Western culture, people receive little explicit feedback about their emotional abilities in comparison to other mental skills. For instance, various institutions promote certain mental abilities: Schools build knowledge, meditative retreats train consciousness, and guilds reinforce musical talent. People who attend these societies receive feedback on their performance. Few institutions are devoted explicitly to developing emotional capacities, however. Discrepancies between self-ratings and performance measures of EI may diminish as education systems incorporate social and emotional learning programs (Brackett & Katulak, in press).

Does EI Contribute to Social Competence? Because intelligence measures are judged in part according to their ability to predict theoretically related behavioral outcomes (cf. American Psychological Association Public Affairs Office, 1996; Funder, 2001), we were interested in whether the EI measures were associated with the ability to form and maintain functional relationships. For men, performance on ability but not self-report measures of EI was related to the quality of social interaction. It is important to note that these findings extend previous work by demonstrating that the MSCEIT is associated both with specific interpersonal strategies that individuals report using and with real-time interpersonal competence. However, the findings are limited in that the MSCEIT was related to outcomes only for men. Specifically, in comparison to men with lower EI, men with higher EI (a) reported using less destructive strategies in both positive and negative emotional situations with a friend (Study 2) and (b) were judged by the confederate and naive judges to be more socially competent in a laboratory-based social interaction with someone they did not know (Study 3). Although these findings need to be replicated, they are consistent with the results of a number of studies suggesting that EI may be an important variable for understanding social adaptation (Lopes et al., 2003), particularly among men (Brackett et al., 2004). It is unclear why the MSCEIT was unrelated to social competence for women in the studies reported here. We suggest four explanations for the gender differences in our results. First, because women are generally higher in EI than men, it is possible that the gender differences in correlations are due to a threshold effect. There may be a minimum level of EI that is needed to function effectively in social situations, and the proportion of men who fall below this threshold may be higher than the

MEASURING EMOTIONAL INTELLIGENCE

proportion of women. Because women have higher MSCEIT scores than men, women (as a group) may have attained that threshold. Differences in scores for women, then, would not explain variance in social competence. Researchers would need to test these hypotheses in a sample with a large number of lowscoring women to see whether the effects are due to EI or gender. In the present studies, the number of women with low EI was too small to test such questions. Second, it may be the case that the MSCEIT is not tapping into EI for women in the same way as it does for men. Emotional abilities may operate or manifest differently for men and women. In the United States, women are expected to be more adept emotionally than men (e.g., L. R. Brody & Hall, 2000); thus, their abilities may be quite different. Third, emotions operate within social norms, and the norms governing appropriate gendered behavior for men and women are different. In everyday experience, expressing emotions that violate social norms and display rules can lead to social consequences (Frijda & Mesquita, 1994; Saarni, 1999); thus, learning to regulate these emotions is adaptive (Goffman, 1959; Hochschild, 1983). Gender norms may influence how emotional abilities operate in men and women. The MSCEIT may be biased in that it better assesses the emotional abilities of men (and thus better predicts relevant social outcomes for men), but it may not capture the abilities of women adequately (and thus is not related to social outcomes for women). Finally, we may have selected gendered conceptualizations of social competence. Despite our efforts to use social competence measures that were comparably applicable to men and women, there were significant gender differences on each measure. As a result, our social competence outcomes may have been more valid for men than for women.

What Are the Limitations of the EI Measures? In the studies reported here, we used only one self-report and one performance test of EI, but at present these are the only available instruments of EI that map onto Mayer and Salovey’s (1997) theoretical model. Thus, the lack of association between the two measures may not generalize beyond these tests. For example, on the SREIS, we asked participants about their emotional abilities (e.g., “Do you have a good emotions vocabulary?”), not whether they believed in their ability to use these skills. A well-designed self-efficacy measure of EI may correlate more strongly with a performance measure (Bandura, 1977, 1997). It is likely that the overlap between the SREIS and many of the personality measures was due to both shared method variance and similar semantic overlap among the items on the scales, although we were unable to test these hypotheses in the present studies. For example, items on the Emotion Management subscale of the SREIS resemble items on both the Neuroticism and Well-Being subscales. Moreover, the reliability of the SREIS was not ideal in all studies (.84, .77, and .66 for Studies 1, 2, and 3, respectively). A more comprehensive self-report measure may yield slightly different findings than those presented here. Although studies have shown that the MSCEIT predicts a wide range of criteria in multiple life domains (see Mayer et al., 2004), it is possible that the MSCEIT is limited in its ability to measure Mayer and Salovey’s (1997) four-domain model of EI. For exam-

791

ple, the MSCEIT does not assess real-time emotion regulation or the ability to express emotion effectively. There also is concern that consensus scoring on the MSCEIT reflects conformity to social norms rather than skill (R. D. Roberts, Zeidner, & Matthews, 2002). However, in the domain of emotions, skill and conformity are not disentangled easily because emotional skills necessarily reflect attunement to social norms and expectations (Lopes et al., 2004). In addition, there is high agreement between expert and consensus scores, which indicates that emotions experts generally view consensual responses as correct (Lopes et al., 2004). MSCEIT scores also are not correlated significantly with social desirability (David, 2005; Lopes et al., 2003).

Future Directions The present research does not address three important issues. First, we know little about the processes through which EI operates in interpersonal relationships. The conceptualization of EI may be more complex in social situations. For example, it may be necessary to assess the EI of both friends when studying relationship quality because there may be an additive effect of EI in dyads (Brackett et al., 2005). Additionally, it would be important to test whether EI scores predict social success during other interactions, including situations that involve cooperation or unequal distributions of power. Moreover, there are no published experiments with mood inductions to assess whether EI skills are instrumental to achieving social success when in a negative emotional state, for instance. Such studies also will allow researchers to identify the processes through which EI operates in interpersonal relationships. Second, we know little about why men and women differ in their performance on the MSCEIT and why the MSCEIT predicts social competence for men but not for women. Any of the explanations described above are possible. The results of this research, in combination with previous investigations, provide sufficient evidence that gender is an important variable to examine in both theories of emotion and empirical investigations. Finally, in this study, we focused on EI as a coherent unified construct and conducted all analyses by using EI total scores. We did not examine the contribution of the individual abilities comprising EI (i.e., the perception, use, understanding, and management of emotion) to social functioning. It is possible that each emotional ability explains unique variance in various aspects of social functioning.

Conclusion Although research on EI is in its incipient stages, these studies suggest that measuring EI with performance tests such as the MSCEIT, as opposed to self-report inventories, makes it possible to analyze the degree to which emotional abilities contribute to social functioning. There is much to be learned about EI. Indeed, performance tests such as the MSCEIT likely will be updated as we learn more about the construct. As an analogy, our knowledge of how intelligence is measured and what it predicts is the product of almost a full century of research.

References Abelson, R. P. (1963). Computer simulation of “hot cognitions.” In S. Tomkins & S. Messick (Eds.), Computer simulation of personality (pp. 277–298). New York: Wiley.

792

BRACKETT, RIVERS, SHIFFMAN, LERNER, AND SALOVEY

Adams, S., Kuebli, J., Boyle, P. A., & Fivush, R. (1995). Gender differences in parent– child conversations about past emotions: A longitudinal investigation. Sex Roles, 33, 309 –323. American Psychological Association Public Affairs Office. (1996). Intelligence: Knowns and unknowns. American Psychologist, 51, 77–101. Bacon, M. K., & Ashmore, R. D. (1985). How mothers and fathers categorize descriptions of social behavior attributed to daughters and sons. Social Cognition, 3, 193–217. Bailey, R. C., & Mettetal, G. W. (1977). Sex differences in the congruency of perceived intelligence. Journal of Genetic Psychology, 131, 29 –36. Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191–215. Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman. Barchard, K. A. (2001). Emotional and social intelligence: Examining its place in the nomological network. Unpublished doctoral dissertation, University of British Columbia, Vancouver, British Columbia, Canada. Bar-On, R. (1997). Bar-On Emotional Quotient Inventory: Technical manual. Toronto, Ontario, Canada: Multi-Health Systems. Bar-On, R. (2000). Emotional and social intelligence: Insights from the Emotional Quotient Inventory. In R. Bar-On & J. D. A. Parker (Eds.), The handbook of emotional intelligence (pp. 363–388). San Francisco: Jossey-Bass. Bennett, M. (1996). Men’s and women’s self-estimates of intelligence. Journal of Social Psychology, 136, 411– 412. Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129 –148. Boyatzis, R. E., Goleman, D., & Rhee, K. (2000). Clustering competence in emotional intelligence: Insights from the Emotional Competence Inventory (ECI). In R. Bar-On & D. A. Parker (Eds.), Handbook of emotional intelligence (pp. 343–362). San Francisco: Jossey-Bass. Brackett, M. A. (2004). [The Self-Rated Emotional Intelligence Scale]. Unpublished raw data. Brackett, M. A., & Katulak, N. (in press). The emotionally intelligent classroom: Skill-based training for teachers and students. In J. Ciarrochi & J. D. Mayer (Eds.), Improving emotional intelligence: A practitioner’s guide. New York: Psychology Press/Taylor & Francis. Brackett, M. A., & Mayer, J. D. (2003). Convergent, discriminant, and incremental validity of competing measures of emotional intelligence. Personality and Social Psychology Bulletin, 29, 1147–1158. Brackett, M. A., Mayer, J. D., & Warner, R. M. (2004). Emotional intelligence and its relation to everyday behaviour. Personality and Individual Differences, 36, 1387–1402. Brackett, M. A., & Salovey, P. (2004). Measuring emotional intelligence with the Mayer–Salovey–Caruso Emotional Intelligence Test (MSCEIT). In G. Geher (Ed.), Measuring emotional intelligence: Common ground and controversy (pp. 179 –194). Happauge, NY: Nova Science Publishers. Brackett, M. A., Warner, R. M., & Bosco, J. (2005). Emotional intelligence and relationship quality among couples. Personal Relationships, 12, 197–212. Brody, L. R., & Hall, J. A. (1993). Gender and emotion. In M. Lewis & J. M. Haviland (Eds.), Handbook of emotions (pp. 447– 460). New York: Guilford Press. Brody, L. R., & Hall, J. A. (2000). Gender, emotion, and expression. In M. Lewis & J. M. Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 338 –349). New York: Guilford Press. Brody, N. (2004). Emotional intelligence: Science and myth. Intelligence, 32, 109 –111. Buck. R. (1984). The communication of emotion. New York: Guilford Press. Cantor, N., & Kihlstrom, J. F. (1987). Personality and social intelligence. Englewood Cliffs, NJ: Prentice-Hall.

Carroll, J. B. (1993). Human cognitive abilities: A survey of factor-analytic studies. New York: Cambridge University Press. Carroll, S. A., & Day, A. L. (2004, August). Faking emotional intelligence: Comparing response distortion on ability and mixed-model measures. Paper presented at the Academy of Management, New Orleans, LA. Christensen, A., & Heavey, C. L. (1990). Gender and social structure in the demand/withdraw pattern of marital conflict. Journal of Personality and Social Psychology, 59, 73– 81. Ciarrochi, J., Forgas, J., & Mayer, J. (Eds.). (2006). Emotional intelligence in everyday life: A scientific inquiry (2nd ed.). Philadelphia: Psychology Press/Taylor & Francis. Clark, M. S., & Fiske, S. T. (Eds.). (1982). Affect and cognition: The 17th Annual Carnegie Symposium on Cognition. Hillsdale, NJ: Erlbaum. Costa, P. T., Jr., & McCrae, R. R. (1992). NEO–PI–R professional manual—Revised NEO Personality Inventory (NEO–PI–R) and NEO FiveFactor Inventory (NEO–FFI). Odessa, FL: Psychological Assessment Resources. Custrini, R., & Feldman, R. S. (1989). Children’s social competence and non-verbal encoding and decoding of emotion. Journal of Child Clinical Psychology, 18, 336 –342. Damasio, A. (1994). Descartes’ error: Emotion, reason, and the human brain. New York: Avon Books. David, S. A. (2005). Emotional intelligence: Developmental antecedents, psychological and social outcomes. Unpublished doctoral dissertation, University of Melbourne, Melbourne, Victoria, Australia. Dawda, D., & Hart, S. D. (2000). Assessing emotional intelligence: Reliability and validity of the Bar-On Emotional Quotient Inventory (EQ-i) in university students. Personality and Individual Differences, 28, 797– 812. Denham, S. A., Blair, K. A., DeMulder, E., Levitas, J., Sawyer, K., Auerbach-Major, S., & Queenan, P. (2003). Preschool emotional competence: Pathway to social competence. Child Development, 74, 238 – 256. De Sousa, R. (1987). The rationality of emotion. Cambridge, MA: MIT Press. Diener, E., Emmons, R. A., & Larsen, R. J. (1985). The Satisfaction With Life Scale. Journal of Personality Assessment, 49, 71–75. Dunning, D., Johnson, K., Ehrlinger, J., & Kruger, J. (2003). Why people fail to recognize their own incompetence. Current Directions in Psychological Science, 12, 83– 87. Ehrlinger, J., & Dunning, D. (2003). How chronic self-views influence (and potentially mislead) estimates of performance. Journal of Personality and Social Psychology, 84, 5–17. Eisenberg, N., Fabes, R. A., Guthrie, I. K., & Reiser, M. (2000). Dispositional emotionality and regulation: Their role in predicting quality of social functioning. Journal of Personality and Social Psychology, 78, 136 –157. Eisenberg, N., Fabes, R. A., Murphy, B., Maszk, P., Smith, M., & Karbon, M. (1995). The role of emotionality and regulation in children’s social functioning: A longitudinal study. Child Development, 66, 1360 –1384. Ekman, P. (1973). Darwin and facial expression: A century of research in review. New York: Academic Press. Ekman, P., & Friesen, W. V. (1975). Unmasking the face: A guide to recognizing emotions from facial clues. Englewood Cliffs, NJ: PrenticeHall. Elfenbein, H. A., Marsh, A. A., & Ambady, N. (2002). Emotional intelligence and the recognition of emotion from facial expressions. In L. F. Barrett & P. Salovey (Eds.), The wisdom in feeling: Psychological processes in emotional intelligence (pp. 37–59). New York: Guilford Press. Feldman, R. S., Philippot, P., & Custrini, R. J. (1991). Social competence and nonverbal behavior. In R. S. Feldman & B. Rime (Eds.), Fundamentals of nonverbal behavior (pp. 329 –350). New York: Cambridge University Press.

MEASURING EMOTIONAL INTELLIGENCE Fivush, R. (1991). Gender and emotion in mother– child conversations about the past. Journal of Narrative and Life History, 1, 325–341. Fivush, R. (1998). Methodological challenges in the study of emotional socialization. Psychological Inquiry, 9, 281–283. Fivush, R., Brotman, M. A., Buckner, J. P., & Goodman, S. H. (2000). Gender differences in parent– child emotion narratives. Sex Roles, 42, 233–253. Forgas, J. P., & Moylan, S. (1987). After the movies: Transient mood and social judgments. Personality and Social Psychology Bulletin, 13, 467– 477. Frijda, N. H. (1988). The laws of emotion. American Psychologist, 43, 349 –358. Frijda, N. H., & Mesquita, B. (1994). The social roles and functions of emotions. In S. Kitayama & H. R. Markus (Eds.), Emotion and culture: Empirical studies of mutual influence (pp. 51– 87). Washington, DC: American Psychological Association. Funder, D. C. (2001). Personality. Annual Review of Psychology, 52, 197–221. Gable, S. L., Reis, H. T., Impett, E., & Asher, E. R. (2004). What do you do when things go right? The intrapersonal and interpersonal effects of sharing positive events. Journal of Personality and Social Psychology, 87, 228 –245. Gardner, H. (1983). Frames of mind: The theory of multiple intelligences. New York: Basic Books. Goffman, E. (1959). The presentation of self in everyday life. New York: Doubleday Anchor. Goleman, D. (1995). Emotional intelligence. New York: Bantam. Goleman, D. (1998). Working with emotional intelligence. New York: Bantam. Gross, J. J. (1998). The emerging field of emotion regulation: An integrative review. Review of General Psychology, 2, 271–299. Hall, J. A. (1978). Gender effects in decoding nonverbal cues. Psychological Bulletin, 85, 845– 857. Hall, J. A. (1984). Nonverbal sex differences: Communication accuracy and expressive style. Baltimore: Johns Hopkins University Press. Harris, R. J. (2001). A primer of multivariate statistics (3rd ed.). Mahwah, NJ: Erlbaum. Hedlund, J., & Sternberg, R. J. (2000). Too many intelligences? Integrating social, emotional, and practical intelligence. In R. Bar-On & J. D. A. Parker (Eds.), The handbook of emotional intelligence (pp. 136 –167). San Francisco: Jossey-Bass. Hochschild, A. (1983). The managed heart: Commercialization of human feeling. Berkeley: University of California Press. Isen, A. M. (1987). Positive affect, cognitive processes, and social behavior. Advances in Experimental Social Psychology, 20, 203–253. Isen, A. M., Shalker, T. E., Clark, M., & Karp, L. (1978). Affect, accessibility of material in memory, and behavior: A cognitive loop? Journal of Personality and Social Psychology, 36, 1–12. Janovics, J., & Christiansen, N. D. (2002). Emotional intelligence in the workplace. Paper presented at the 16th Annual Conference of the Society of Industrial and Organizational Psychology, San Diego, CA. Johnson, O. P., Donahue, E. M., & Kentle, R. L. (1991). The “Big Five” inventory—Versions 4a and 54 (Tech. Rep.). Berkeley, CA: Institute of Personality Assessment and Research. Keltner, D., & Haidt, J. (2001). Social functions of emotions. In T. J. Mayne & G. A. Bonanno (Eds.), Emotions: Current issues and future directions. Emotions and social behavior (pp. 192–213). New York: Guilford Press. Keltner, D., & Kring, A. M. (1998). Emotion, social function, and psychopathology. Review of General Psychology, 2, 320 –342. Lane, R. D., Quinlan, D. M., Schwartz, G. E., Walker, P. A., & Zeitlin, S. B. (1990). The Levels of Emotional Awareness Scale: A cognitive– developmental measure of emotion. Journal of Personality Assessment, 55, 124 –134.

793

Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford University Press. Lenney, E. (1977). Women’s self-confidence in achievement settings. Psychological Bulletin, 84, 1–13. Lopes, P. N., Brackett, M. A., Nezlek, J. B., Schutz, A., Sellin, I., & Salovey, P. (2004). Emotional intelligence and social interaction. Personality and Social Psychology Bulletin, 30, 1018 –1034. Lopes, P. N., Grewal, D., Kadis, J., Gall, M., & Salovey, P. (in press). Evidence that emotional intelligence is related to job performance and affect and attitudes at work. Psicothema. Lopes, P. N., Salovey, P., & Straus, R. (2003). Emotional intelligence, personality, and the perceived quality of social relationships. Personality and Individual Differences, 3, 641– 659. Lumley, M. A., Gustavson, B. J., Patridge, R. T., & Labouvie-Vief, G. (2005). Assessing alexithymia and related emotional ability constructs using multiple methods: Interrelationships among measures. Emotion, 5, 329 –342. Mabe, P. A., III, & West, S. G. (1982). Validity of self-evaluation of ability: A review and meta-analysis. Journal of Applied Psychology, 67, 280 –296. Maccoby, E. E. (1998). The two sexes: Growing up apart, coming together. Cambridge, MA: Harvard University Press. Mayer, J. D., & Bremer, D. (1985). Assessing mood and affect-sensitive tasks. Journal of Personality Assessment, 49, 95–99. Mayer, J. D., Caruso, D., & Salovey, P. (1999). Emotional intelligence meets traditional standards for an intelligence. Intelligence, 27, 267–298. Mayer, J. D., & Mitchell, D. C. (1998). Intelligence as a subsystem of personality: From Spearman’s g to contemporary models of hotprocessing. In W. Tomic & J. Kingma (Eds.), Advances in cognition and educational practice. Vol. 5: Conceptual issues in research on intelligence (pp. 43–75). Greenwich, CT: JAI Press. Mayer, J. D., & Salovey, P. (1997). What is emotional intelligence? In P. Salovey & D. J. Sluyter (Eds.), Emotional development and emotional intelligence: Educational implications (pp. 4 –30). New York: Basic Books. Mayer, J. D., Salovey, P., & Caruso, D. (2000). Models of emotional intelligence. In R. J. Sternberg (Ed.), Handbook of human intelligence (pp. 396 – 420). New York: Cambridge University Press. Mayer, J. D., Salovey, P., & Caruso, D. (2002a). The Mayer–Salovey– Caruso Emotional Intelligence Test (MSCEIT), Version 2.0. Toronto, Ontario, Canada: Multi-Health Systems. Mayer, J. D., Salovey, P., & Caruso, D. (2002b). MSCEIT technical manual. Toronto, Ontario, Canada: Multi-Health Systems. Mayer, J. D., Salovey, P., & Caruso, D. (2004). Emotional intelligence: Theory, findings, and implications. Psychological Inquiry, 15, 197–215. Mayer, J. D., Salovey, P., Caruso, D., & Sitarenios, G. (2003). Measuring emotional intelligence with the MSCEIT V2.0. Emotion, 3, 97–105. McClure, E. B. (2000). A meta-analytic review of sex differences in facial expression processing and their development in infants, children, and adolescents. Psychological Bulletin, 126, 424 – 453. McCrae, R. R. (2000). Emotional intelligence from the perspective of the five-factor model of personality. In R. Bar-On & J. D. A. Parker (Eds.), The handbook of emotional intelligence (pp. 263–276). San Francisco: Jossey-Bass. Mehrabian, A., & Epstein, N. (1972). A measure of emotional empathy. Journal of Personality, 40, 525–543. Newsome, S., Day, A. L., & Catano, V. M. (2000). Assessing the predictive validity of emotional intelligence. Personality and Individual Differences, 29, 1005–1016. Nowicki, S., Jr., & Duke, M. P. (1994). Individual difference in nonverbal communication of affect: The Diagnostic Analysis of Nonverbal Accuracy Scale. Journal of Nonverbal Behavior, 18, 9 –35. Nowicki, S., & Mitchell, J. (1998). Accuracy in identifying affect in child

794

BRACKETT, RIVERS, SHIFFMAN, LERNER, AND SALOVEY

and adult faces and voices and social competence in preschool children. Genetic, Social, and General Psychology Monographs, 124, 39 –59. Palfai, T. P., & Salovey, P. (1993). The influence of depressed and elated mood on deductive and inductive reasoning. Imagination, Cognition, and Personality, 13, 57–71. Parker, J. D. A., Taylor, G. J., & Bagby, R. M. (2001). The relationship between alexithymia and emotional intelligence. Personality and Individual Differences, 30, 107–115. Paulhus, D. L. (1991). Measurement and control of response bias. In J. P. Robinson, P. R. Shaver, & L. S. Wrightsman (Eds.), Measures of personality and social psychological attitudes (pp. 17–59). New York: Academic Press. Paulhus, D. L., Lysy, D. C., & Yik, M. S. M. (1998). Self-report measures of intelligence: Are they useful as proxy IQ tests? Journal of Personality Psychology, 66, 525–554. Pavot, W., & Diener, E. (1993). Review of the Satisfaction With Life Scale. Psychological Assessment, 5, 164 –172. Petrides, K. V., & Furnham, A. (2003). Trait emotional intelligence: Behavioral validation in two studies of emotion recognition and reactivity to mood induction. European Journal of Personality, 17, 39 –57. Reilly, J., & Mulhern, G. (1995). Gender differences in self-estimated IQ: The need for care in interpreting group data. Personality and Individual Differences, 18, 189 –192. Rivers, S. E., Brackett, M. A., Salovey, P., & Mayer, J. D. (in press). Measuring emotional intelligence using ability-based assessment. In G. Matthews, M. Zeidner, & R. D. Roberts (Eds.), The science of emotional intelligence. New York: Oxford University Press. Roberts, R. D., Zeidner, M., & Matthews, G. (2002). Does emotional intelligence meet traditional standards for an intelligence? Some new data and conclusions. Emotion, 1, 196 –231. Roberts, T. (1991). Gender and the influence of evaluations on selfassessments in achievement settings. Psychological Bulletin, 109, 297– 308. Rosenthal, R., Hall, J. A., DiMatteo, M. R., Rogers, P. L., & Archer, D. (1979). Sensitivity to nonverbal communication: The PONS test. Baltimore: Johns Hopkins University Press. Rothbart, M. K. (1989). Biological processes in temperament. In G. A. Kohnstamm & J. E. Bates (Eds.), Temperament in childhood (pp. 77– 110). Chichester, England: Wiley. Rusbult, C. E. (1993). Understanding responses to dissatisfaction in close relationships: The exit, voice, loyalty, and neglect model. In S. Worchel & J. A. Simpson (Eds.), Conflict between people and groups: Causes, processes, and resolutions (pp. 30 –59). Chicago: Nelson-Hall. Rusbult, C. E., Johnson, D. J., & Morrow, G. D. (1986). Impact of couple patterns of problem solving on distress and nondistress in dating relationships. Journal of Personality and Social Psychology, 50, 744 –753. Rusbult, C. E., Verrete, J., Whitney, G. A., Slovic, L. F., & Lipkus, I. (1991). Accommodation processes in close relationships: Theory and

preliminary empirical evidence. Journal of Personality and Social Psychology, 60, 53–78. Ryff, C. D. (1989). Happiness is everything, or is it? Explorations on the meaning of psychological well-being. Journal of Personality and Social Psychology, 57, 1069 –1081. Saarni, C. (1999). Developing emotional competence. New York: Guilford Press. Salovey, P., & Birnbaum, D. (1989). Influence of mood on health-relevant cognitions. Journal of Personality and Social Psychology, 57, 539 –551. Salovey, P., & Mayer, J. D. (1990). Emotional intelligence. Imagination, Cognition, and Personality, 9, 185–211. Salovey, P., Mayer, J. D., Goldman, S. L., Turvey, C., & Palfai, T. P. (1995). Emotional attention, clarity, and repair: Exploring emotional intelligence using the Trait Meta-Mood Scale. In J. W. Pennebaker (Ed.), Emotion, disclosure, and health (pp. 125–154). Washington, DC: American Psychological Association. Savage, C. R. (2002). The role of emotion in strategic behavior. In L. F. Barrett & P. Salovey (Eds.), The wisdom in feeling (pp. 211–236). New York: Guilford Press. Scherer, K. R., Banse, R., & Wallbott, H. G. (2001). Emotion inferences from vocal expression correlate across languages and cultures. Journal of Cross-Cultural Psychology, 32, 76 –92. Schutte, N. S., Malouff, J. M., Hall, L. E., Haggerty, D. J., Cooper, J. T., Golden, C. J., & Dornheim, L. L. (1998). Development and validation of a measure of emotional intelligence. Personality and Individual Differences, 25, 167–177. Schwarz, N. (1990). Feelings as information: Informational and motivational functions of affective states. In E. T. Higgins & E. M. Sorrentino (Eds.), Handbook of motivation and cognition (Vol. 2, pp. 527–561). New York: Guilford Press. Schwarz, N., & Clore, G. L. (1996). Feelings and phenomenal experiences. In E. T. Higgins & A. W. Kruglanski (Eds.), Social psychology: Handbook of basic principles (pp. 433– 465). New York: Guilford Press. Shields, S. A. (2002). Speaking from the heart: Gender and the social meaning of emotion. Cambridge, England: Cambridge University Press. Singer, J. A., & Salovey, P. (1988). Mood and memory: Evaluating the network theory of affect. Clinical Psychology Review, 8, 211–251. Sternberg, R. J. (1985). The triarchic mind: A new theory of human intelligence. New York: Penguin. Topping, K., Bremner, W., & Holmes, E. (2000). Social competence: The social construction of the concept. In R. Barron & J. D. A. Parker (Eds.), Handbook of emotional intelligence (pp. 28 –39). San Francisco: JosseyBass. Wechsler, D. (1997). WAIS–III: Wechsler Adult Intelligence Scale (3rd ed.). San Antonio, TX: Psychological Corporation. Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psychologist, 35, 151–175.

MEASURING EMOTIONAL INTELLIGENCE

795

Appendix Self-Rated Emotional Intelligence Scale The following set of items pertains to your insight into emotions. Please use the rating scale below to describe how accurately each statement describes you. Describe yourself as you generally are now, not as you wish to be in the future. Describe yourself as you honestly see yourself, in relation to other people you know of the same sex as you are, and roughly your same age. Please read each statement carefully, and then write the letter that corresponds to how inaccurately or accurately each statement describes you. Very inaccurate 1

Moderately inaccurate 2

Neither nor 3

Moderately accurate 4

Very accurate 5

Number

Domain

Item wording

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

P U R M1 M2 P U R M1 M2 P U R M1 M2 P R M1 M2

By looking at people’s facial expressions, I recognize the emotions they are experiencing. I am a rational person and I rarely, if ever, consult my feelings to make a decision (r). I have a rich vocabulary to describe my emotions. I have problems dealing with my feelings of anger (r). When someone I know is in a bad mood, I can help the person calm down and feel better quickly. I am aware of the nonverbal messages other people send. When making decisions, I listen to my feelings to see if the decision feels right. I could easily write a lot of synonyms for emotion words like happiness or sadness. I can handle stressful situations without getting too nervous. I know the strategies to make or improve other people’s moods. I can tell when a person is lying to me by looking at his or her facial expression. I am a rational person and don’t like to rely on my feelings to make decisions. I have the vocabulary to describe how most emotions progress from simple to complex feelings. I am able to handle most upsetting problems. I am not very good at helping others to feel better when they are feeling down or angry (r). My quick impressions of what people are feeling are usually wrong (r). My “feelings” vocabulary is probably better than most other persons’ “feelings” vocabularies. I know how to keep calm in difficult or stressful situations. I am the type of person to whom others go when they need help with a difficult situation.

Note. P ⫽ Perceiving Emotion; U ⫽ Use of Emotion; (r) ⫽ reverse scored; R ⫽ Understanding Emotion; M1 ⫽ Managing Emotion (self); M2 ⫽ Social Management.

Received March 11, 2005 Revision received May 16, 2006 Accepted May 18, 2006 䡲

APPRAISING AND MANAGING RISK SECOND EDITION

Vernon L. Quinsey, Grant T. Harris, Marnie E. Rice, and Catherine A. Cormier PART OF THE APA LAW AND PUBLIC POLICY: PSYCHOLOGY AND THE SOCIAL SCIENCES SERIES – SERIES EDITOR: BRUCE D. SALES Predicting future violence among criminal offenders is notoriously difficult. In the first edition of this popular book, authors Quinsey, Harris, Rice, and Cormier argued that community risk management can be improved by using actuarial assessment and by combining what is known about the prediction of violence, the study of clinical decision making, and the literature on treatment outcome and program evaluation. They reported on their long-term research at the Oak Ridge Division of the Penetanguishene Mental Health Care Centre, in Ontario, Canada— CONTENTS: probably the most thoroughly studied maximum security psychiatric Introduction ■ Acknowledgments ■ Part I. Historical and Methodological Context facility in the world—and chronicled the development of their assess■ Chapter 1. Historical Perspectives ■ Chapter 2. Previous Research on ment instrument, the Violence Risk Appraisal Guide (VRAG). Prediction ■ Chapter 3. Methods and Measurement ■ Chapter 4. Clinical Since that first edition appeared, the field of violence risk Judgment ■ Part II. A New Generation of Follow-Up Studies ■ Chapter 5. assessment has exploded. The literature base is much larger and Mentally Disordered Offenders ■ Chapter 6. Fire Setters ■ Chapter 7. richer, as is the relevant commentary. In this new edition, the authors Sex Offenders ■ Part III. Development of Violence Prediction Instruments update their review, focusing on the actuarial instruments they ■ Chapter 8. Actuarial Prediction of Violence ■ Chapter 9. Criticisms of Actuarial Risk Assessment ■ Part IV. Altering the Risk of Violence and developed and described earlier and on the measures they have Conclusions ■ Chapter 10. Treatment and Management ■ Chapter 11. continued to develop. In their lively style, they review the Conclusions ■ Appendixes ■ References ■ Author Index ■ Subject Index commentary on risk appraisal, addressing 20 of the most common ■ About the Authors arguments against actuarial risk appraisal. They clarify how to score items of the Violence Risk Appraisal Guide (VRAG) and Sex Offender Risk Appraisal Guide (SORAG) based on extensive inquiries from professionals who use these instruments in the field. Lastly, they provide a more detailed description of the development of the SORAG. This book is a must-read for all legal and psychological professionals and policy makers who participate in decisions on whether, when, and under what circumstances violent offenders should be released, or how they should be managed. 2006. 456 pages. Hardcover.

List: $79.95 • APA Member/Affiliate: $49.95 • ISBN 1-59147-343-8 • Item # 4316068

ALSO AVAILABLE VIOLENCE AND GENDER REEXAMINED Richard B. Felson 2002 • 273 pages • Hardcover • List: $49.95 • APA Member/ Affiliate: $39.95 • ISBN 1-55798-895-1 • Item # 431675A

PREVENTING VIOLENCE Research and Evidence-Based Intervention Strategies Edited by John R. Lutzker 2006 • 320 pages • Hardcover • List: $69.95 • APA Member/Affiliate: $49.95 ISBN 1-59147-342-X • Item # 4316067

PREVENTING YOUTH VIOLENCE IN A MULTICULTURAL SOCIETY Edited by Nancy G. Guerra and Emilie Phillips Smith 2006 • 304 pages • Hardcover • List: $69.95 • APA Member/Affiliate: $49.95 ISBN 1-59147-327-6 • Item # 4316064

APA Books Ordering Information

800-374-2721

www.apa.org/books In Washington, DC, call: 202-336-5510 TDD/TTY: 202-336-6123 • Fax: 202-336-5502 In Europe, Africa, or the Middle East, call: 44-207-240-0856

AMERICAN PSYCHOLOGICAL ASSOCIATION AD0400

BEST SELLERS from the American Psychological Association

Handbook of Multimethod Measurement in Psychology

Emotion-Focused Therapy for Depression

MMPI-2: A Practitioner’s Guide

Edited by Michael Eid and Ed Diener

Leslie S. Greenberg, and Jeanne C. Watson

2006. 464 pages. Hardcover. List: $99.95 APA/Member Afﬁliate: $64.95 ISBN 1-59147-318-7 • Item # 4316063

2006. 344 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-280-6 • Item # 4317081

2006. 688 pages. Hardcover. List: $89.95 APA Member/Afﬁliate: $64.95 ISBN 1-59147-287-3 • Item # 4317086

Adlerian Therapy

Bullying Prevention

Theory and Practice

Creating a Positive School Climate and Developing Social Competence

Jon Carlson, Richard E. Watts, and Michael Maniacci

Pamela Orpinas and Arthur M. Horne

2006. 304 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $44.95 ISBN 1-59147-285-7 • Item # 4317084

2006. 293 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-282-2 • Item # 4317082

Constructive Divorce

Road Rage

Procedural Justice and Sociolegal Reform

Assessment and Treatment of the Angry, Aggressive Driver

Penelope Eileen Bryan

Tara E. Galovski, Loretta S. Malta, and Edward B. Blanchard

2006. 344 pages. Hardcover. List: $69.95 APA Member/Afﬁliate: $54.95 ISBN 1-59147-279-2 • Item # 4316056

2006. 256 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $44.95 ISBN 1-59147-305-5 • Item # 4317088

Your Practicum in Psychology A Guide for Maximizing Knowledge and Competence

The Ethical Practice of Psychology in Organizations

Edited by Janet R. Matthews and C. Eugene Walker

Second Edition

2006. 320 pages. Paperback. List: $49.95 APA Member/Afﬁliate: $39.95 ISBN 1-59147-328-4 • Item # 4313006

2006. 330 pages. Hardcover. List: $39.95 APA Member/Afﬁliate:$34.95 ISBN 1-59147-353-5 • Item # 4312006

Essentials of Clinical Hypnosis

Practical Ethics for Psychologists

An Evidence-Based Approach

A Positive Approach

Steven Jay Lynn and Irving Kirsch

Samuel J. Knapp and Leon Vandecreek

List: $69.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-344-6 • Item # 4317095

2006. 304 pages. Paperback. List: $49.95 APA Member/Afﬁliate: $39.95 ISBN 1-59147-326-8 • Item # 4312004

Edited by Rodney L. Lowman

Women and Victimization Contributing Factors, Interventions, and Implications

Ethical Practice in Small Communities

TK Logan, Robert Walker, Carol E. Jordan and Carl G. Leukefeld

Challenges and Rewards for Psychologists

2006. 408 pages. Hardcover. List: $79.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-316-0 • Item # 4316059

Janet A. Schank and Thomas M. Skovholt

Sex and Love in Intimate Relationships

2006. 184 pages. Paperback. List: $49.95 APA Member/Afﬁliate: $39.95 ISBN 1-59147-346-2 • Item # 4312005

Research and Evidence-Based Intervention Strategies

2006. 304 pages. Hardcover. List: $49.95 APA Member/Afﬁliate: $39.95 ISBN 1-59147-286-5 • Item # 4317085

Edited by John R. Lutzker

Examining Identity and Development in Lesbian, Gay, and Bisexual People

Edited by Allen M. Omoto and Howard S. Kurtzman 2006. 336 pages. Hardcover. List: $69.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-232-6 • Item # 4316052

Psychological Interventions in Childhood Chronic Illness Dennis Drotar 2006. 404 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-330-6 • Item # 4318027

Psychological Treatment of Chronic Illness A Biopsychosocial Therapy Approach

Len Sperry 2006. 311 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-354-3 • Item # 4317096

Chronic Depression Interpersonal Sources, Therapeutic Solutions

Jeremy W. Pettit and Thomas E. Joiner 2006. 232 pages. Hardcover. List: $59.95 APA Member/Afﬁliate: $44.95 ISBN 1-59147-306-3 • Item # 4317089

Therapeutic Alliances in Couple and Family Therapy An Empirically-Informed Guide to Practice

Myrna L. Friedlander, Valentín Escudero, and Laurie Heatherington 2006. 352 pages. Hardcover. List: $59.95 APA Member/Afﬁliate $49.95 ISBN 1-59147-331-4 • Item # 4317093

Couple Power Therapy Building Commitment, Cooperation, Communication, and Community in Relationships

Peter L. Sheras and Phyllis R. Koch-Sheras

Imagery in Psychotherapy

2006. 264 pages. Hardcover. List: $49.95 APA Member/Afﬁliate: $39.95 ISBN 1-59147-235-0 • Item # 4317074

Jerome L. Singer

Preventing Violence

Robert W. Firestone, Lisa A. Firestone, and Joyce Catlett

Sexual Orientation and Mental Health

Edited by James N. Butcher

2006. 320 pages. Hardcover. List: $69.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-342-X • Item # 4316067

Preventing Youth Violence in a Multicultural Society Edited by Nancy G. Guerra and Emilie Phillips Smith 2006. 304 pages. Hardcover. List: $69.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-327-6 • Item # 4316064

Reforming Punishment

2006. 280 pages. Hardcover. List: $69.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-333-0 • Item # 4316066

Psychological Limits to the Pain of Imprisonment

Emerging Adults in America

Craig Haney

Edited by Jeffrey Jensen Arnett and Jennifer Lynn Tanner

2006. 320 pages. Hardcover. List: $79.95 APA Member/Afﬁliate: $49.95 ISBN 1-59147-317-9 • Item # 4316061

Identity and Story Creating Self in Narrative

Edited by Dan P. McAdams, Ruthellen Josselson, and Amia Lieblich 2006. 357 pages. Hardcover. List: $59.59 APA Member/Afﬁliate: $49.95 ISBN 1-59147-356-X • Item # 4316071

Coming of Age in the 21st Century

2006. 376 pages. Hardcover. List: $79.95 APA Members/Afﬁliates: $49.95 ISBN 1-59147-329-2 • Item # 4317092

In the Room With Men A Casebook of Therapeutic Change

Edited by Matt Englar-Carlson and Mark A. Stevens 2006. 376 pages. Hardcover. List: $59.95 APA Member/Afﬁliate $49.95 ISBN 1-59147-332-2 • Item # 4317094 AD0457

To O r d e r : 8 0 0 - 374 -2721 • w w w. a p a .or g / b o o ks

Reforming Punishment Psychological Limits to the Pain of Imprisonment CRAIG HANEY Today the United States imprisons more people than any other nation. Its prisons are overcrowded, contain unprecedented numbers of mentally ill and nonviolent prisoners, and grossly over represent minorities. Too many prison systems still do too little to provide meaningful programming and other forms of effective rehabilitation. Yet adverse conditions can cause prisoners to adapt to the pains of imprisonment in ways that are problematic while they are imprisoned and dysfunctional after they are released. In this critical and incisive but theoretically and empirically CONTENTS: Chapter based analysis, Haney examines the key psychological issues 1. Introduction. Humane Values, at the heart of these problems. He uses modern psychologiPsychology, and The cal theory not only to challenge current prison practices but Pains of Imprisonment • Part also to point to ways psychologists, policymakers, and others One. Psychology and can help create a more effective and humane justice system. The Prison Form – Chapter 2. Psychology and the History of 2006. 320 pages. Hardcover. Imprisonment. Justice and Human Nature • Chapter 3. Spreading Pain.

ISBN 1-59147-317-9 • Item # 4316061 List: $79.95 • APA Member/Affiliate: $49.95 Part of the Law and Public Policy: Psychology and the Social Sciences series.

SERIES EDITOR: BRUCE D. SALES

The Punitive State and the State of the Prisons • Chapter 4. Attribution Error as Modern Prison Policy. Race, Drugs, and the Rage to Punish • Part Two. Prison As Criminogenic Context – Chapter 5. Context Matters. Modern Psychological Theory and Rational Prison Policy • Chapter 6. The Social Context of Prison. Surviving the Modern Fortress Warehouse • Chapter 7. The Situational Pathologies of Prison. Maintaining Control Through Force and Intimidation • Chapter 8. Prisoners in Extremis. The Lives of the Mentally Ill and Developmentally Disabled in Prison • Chapter 9. Limiting Unnecessary Prison Pain. Psychological Theory and Humane Prison Policy • Chapter 10. Concluding Thoughts on the Pains of imprisonment

ALSO AVAILABLE MAKING GOOD How Ex-Convicts Reform and Reguild Their Lives SHADD MARUNA, FORWARD BY HANS TOCH

2001. 211 pages. Hardcover List: $39.95 APA Member/Affiliate: $34.95 ISBN 1-55798-731-9 Item # 431645A PREVENTING VIOLENCE Research and Evidence-Based Intervention Strategies EDITED BY JOHN R. LUTZKER

2005. 320 pages. Hardcover. List: $69.95 APA Member/Affiliate: $49.95 ISBN 1-59147-342-X Item # 4316067 SUBSTANCE ABUSE TREATMENT FOR CRIMINAL OFFENDERS An Evidence-Based Guide for Practitioners DAVID W. SPRINGER, C. AARON MCNEECE AND ELIZABETH MAYFIELD ARNOLD

2003. 252 pages. Hardcover. List: $39.95 APA Member/Affiliate: $34.95 ISBN 1-55798-990-7 Item # 4317006

APA Books Ordering Information

800-374-2721 www.apa.org/books In Washington, DC, call: 202-336-5510 TDD/TTY: 202-336-6123 • Fax: 202-336-5502 In Europe, Africa, or the Middle East, call: 44-207-240-0856

AD0405

UNDERSTANDING DEPRESSION IN WOMEN APPLYING EMPIRICAL RESEARCH TO PRACTICE AND POLICY CAROLYN M. MAZURE AND GWENDOLYN PURYEAR KEITA

T

his book is directed toward understanding why we see sex differences in depression, how depression affects women, and how best to treat and prevent depression in women. Women are more likely to suffer from depression than are men, and depression is the leading cause of disability for women CONTENTS: throughout the world. The editors Acknowledgements ■ Foreward: The Importance of Studying Women and survey the findings of experts in Depression ■ Introduction: Understanding Depression in depression and explore the latest findings Women ■ Chapter 1. The Etiology of Gender Differences in on treatment, prevention, and service Depression ■ Chapter 2. Treatment and Prevention of delivery. Drawing on the work of more Depression in Women ■ Chapter 3. Targeting Populations for than 40 top experts in the field, this Prevention and Treatment of Depression in the Context of Reproductive Events ■ Chapter 4. Improving Services and book will influence the work of Outreach for Women With Depression researchers, practitioners and policy makers for years to come. The book is divided into four major chapters, which focus on sex differences in the etiology of depression, treatment of depression in women, treatment of special populations of women, and effective detection and prevention of depression, respectively. 2006. 200 pages. Hardcover.

List: $49.95 • APA Member/Affiliate: $39.95 • ISBN 1-59147-406-X • Item # 4316076

ALSO AVAILABLE DEPRESSION with Michael D. Yapko Series: Specific Treatments for Specific Populations VHS – List: $99.95 • APA Member/Affiliate: $69.95 ISBN 1-59147-257-1 • Item # 4310363 DVD – List: $99.95 • APA Member/Affiliate: $69.95 ISBN 1-59147-258-X • Item # 4310717

CHRONIC DEPRESSION Interpersonal Sources, Therapeutic Solutions Jeremy W. Petitt and Thomas E. Joiner 2006 • 213 pages • Hardcover • List: $59.95 • APA Member/Affiliate: $44.95 ISBN 1-59147-306-3 • Item # 4317089

EXPERIENCES OF DEPRESSION Theoretical, Clinical and Research Perspectives Sidney J. Blatt 2004 • 359 pages • Hardcover • List: $49.95 • APA Member/Affiliate: $39.95 ISBN 1-59147-095-1 • Item # 4317039

APA Books Ordering Information

800-374-2721

www.apa.org/books In Washington, DC, call: 202-336-5510 TDD/TTY: 202-336-6123 • Fax: 202-336-5502 In Europe, Africa, or the Middle East, call: 44-207-240-0856

AMERICAN PSYCHOLOGICAL ASSOCIATION AD0437

Disorders of the Self A PERSONALITY-GUIDED APPROACH Marshall L. Silverstein

I

n this thought-provoking book, Marshall L. Silverstein applies a self psychological viewpoint, as formulated and broadened by Kohut, to understanding the personality disorders designated on Axis II of the DSM-IV. He recasts the disorders as disorders of the self, grouping them into one of three patterns: those that center on (1) combating devitalization, (2) forestalling fragmentation, or (3) seeking alternative pathways to a cohesive self. For each group, he outlines the descriptive psychopathology and main theoretical viewpoints and then presents a self psychological reformulation of how the behavior and symptom patterns CONTENTS: Acknowledgments represent deficits in self-cohesion. ■ Introduction Silverstein considers three disturbances presently not ■ I. Theoretical classified as personality disorders in the DSM-IV nomenclature, Foundations one of which, depressive personality disorder, is under ■ Chapter 1. Theoretical Introduction ■ Chapter 2. Narcissistic Personality consideration for future inclusion. The others are consistent Disorder ■ II. Devitalization. The Unmirrored Self ■ Chapter 3. with the general definition of personality disorder as a pattern Descriptive Psychopathology and Theoretical Viewpoints ■ Chapter 4. A Self Psychological Viewpoint ■ III. Forestalling Fragmentation of deviant inner experience and behavior that is enduring, ■ Chapter 5. Descriptive Psychopathology and Theoretical Viewpoints inflexible, and pervasive and that produces clinical impairment ■ Chapter 6. A Self Psychological Viewpoint ■ Part IV. Alternative (i.e., harmful dysfunction). All are amenable to explanation Pathways for Preserving a Cohesive Self ■ Chapter 7. Descriptive from a self psychological framework, the author observes, and Psychopathology and Theoretical Viewpoints ■ Chapter 8. A Self notes that self psychologists have considerable experience Psychological Viewpoint ■ V. Other Disorders of the Self ■ Chapter 9. working with such disturbances. Depressive Personality Disorder ■ Chapter 10. Disorders of the Self and This thoughtfully prepared volume, the first to systematiSomatic Reactivity ■ Chapter 11. Disavowal and the Vertical Split ■ Afterword ■ References ■ Author Index ■ Subject Index ■ About the Author cally apply this theoretical framework to this broad group of disorders, offers readers fascinating and valuable insights into how undermined self-cohesion compromises patients’ functioning in their daily lives. 2007. 320 pages. Hardcover. Series: Personality-Guided Psychology

List: $69.95 • APA Member/Affiliate: $49.95 • ISBN 1-59147-430-2 • Item # 4317116 • ISBN-13: 978-1-59147-430-2

ALSO AVAILABLE SELF-RELATIONS IN THE PSYCHOTHERAPY PROCESS Edited by J. Christopher Muran 2001 • 391 pages • Hardcover • List: $49.95 APA Member/Affiliate: $39.95 • ISBN 1-55798-733-5 Item # 431648A • ISBN-13: 978-1-55798-733-4

RESOLUTION OF INNER CONFLICT An Introduction to Psychoanalytic Therapy Second Edition Frank Auld, Marvin Hyman, and Donald Rudzinski 2005 • 283 pages • Hardcover • List: $59.95 • APA Member/Affiliate: $49.95 ISBN 1-59147-195-8 • Item # 4317057 • ISBN-13: 978-1-59147-195-0

SELF AND MOTIVATION Emerging Psychological Perspectives Edited by Abraham Tesser, Diederik A. Stapel, and Joanne V. Wood 2002 • 226 pages • Hardcover • List: $49.95 • APA Member/Affiliate: $39.95 ISBN 1-55798-883-8 • Item # 4318999 • ISBN-13: 978-1-55798-883-6

APA Books Ordering Information

800-374-2721

www.apa.org/books In Washington, DC, call: 202-336-5510 TDD/TTY: 202-336-6123 • Fax: 202-336-5502 In Europe, Africa, or the Middle East, call: 44-207-240-0856

AMERICAN PSYCHOLOGICAL ASSOCIATION AD0469

Becoming Culturally Oriented Practical Advice for Psychologists and Educators NADYA A. FOUAD AND PATRICIA ARREDONDO

I

n Becoming Culturally Oriented, Nadya A. Fouad and Patricia Arredondo provide a comprehensive framework for helping psychologists to increase and improve culturally responsive practice, research, and education. Research shows that racial and ethnic minorities have less access to mental health services than do whites and are more likely to receive poor quality services. Compounding these problems is the fact that ethnic minority psychologists are poorly represented CONTENTS: Chapter 1. among psychologists as a whole, relative to their numbers Introduction to in the general population. To address these concerns, APA Multicultural-Centered Practices ■ Chapter 2. Evaluating Cultural has developed the Guidelines on Multicultural Education, Identity And Biases ■ Chapter 3. Psychologists in Cross-Cultural Interactions With Others ■ Chapter 4. Implications for Psychologists as Training, Research, and Organizational Change for Practitioners ■ Chapter 5. Implications for Psychologists as Educators Psychologists as a blueprint for psychologists pursuing their ■ Chapter 6. Implications for Psychologists as Researchers ■ Chapter 7. work in increasingly diverse communities. The Guidelines Psychologists as Organizational Change Agents ■ Chapter 8. Concluding Thoughts: Psychology as a Transformed Profession ■ Appendix A: are written in aspirational language, but to date, an Checklist for Culturally Competent Practice ■ Appendix B: The associated program for applying the guidelines has been Empowerment Workshops’ Workforce Diversity Audit ■ References ■ Index ■ About the Authors missing. In this book, Fouad and Arredondo show how educators, practitioners, administrators, and researchers can use each of the guidelines as a basis for consciousness-raising and self-examination as well as for broadening culturally responsive practices on an organizational level. Addressing each guideline in turn, the authors provide case studies, checklists, and questions for self-examination and discussion, designed to foster planning and implementation of more culturally informed psychological services and teaching practices. 2007. 208 pages. Hardcover.

List: $49.95 • APA Member/Affiliate: $39.95 • ISBN 1-59147-424-8 • Item # 4317114 • ISBN-13: 978-1-59147-424-1

ALSO AVAILABLE RACIAL IDENTITY IN CONTEXT The Legacy of Kenneth B. Clark Edited by Gina Philogène 2004 • 273 pages • Hardcover List: $59.95 • APA Member/Affiliate: $44.95 • ISBN 1-59147-122-2 Item # 4316032 • ISBN-13: 978-1-59147-122-6

ADDRESSING CULTURAL COMPLEXITIES IN PRACTICE A Framework for Clinicians and Counselors Pamela A. Hays 2001 • 239 pages •. Hardcover • List: $39.95 • APA Member/Affiliate: $34.95 ISBN 1-55798-768-8 • Item # 431773A • ISBN-13: 978-1-55798-768-6

DEFINING DIFFERENCE Race and Racism in the History of Psychology Edited by Andrew S. Winston 2004 • 303 pages • Hardcover • List: $49.95 • APA Member/Affiliate: $39.95 ISBN 1-59147-027-7 • Item # 4316011 • ISBN-13: 978-1-59147-027-4

APA Books Ordering Information

800-374-2721

www.apa.org/books In Washington, DC, call: 202-336-5510 TDD/TTY: 202-336-6123 • Fax: 202-336-5502 In Europe, Africa, or the Middle East, call: 44-207-240-0856

AMERICAN PSYCHOLOGICAL ASSOCIATION AD0460

Scientiﬁc Jury Selection JOEL D. LIEBERMAN AND BRUCE D. SALES iven the importance of trial consultants to the modern day practice of law, Scientific Jury Selection is designed to be informative for psychologists, other professionals interested in trial consulting (e.g., sociologists, communication experts, marketing researchers, psychiatrists, and social workers), and attorneys. The authors provide a thorough review of the most common techniques used to select jurors, and a critical evaluation of the ultimate effectiveness of these methods. This critique is CONTENTS: based upon an examination of the Chapter 1. History and social science literature. Overview of the Scientific Jury Selection Process ■ Chapter 2. The Purpose and Effectiveness of the Psychologists and other social Voir Dire ■ Chapter 3. Community Surveys ■ Chapter 4. The Influence scientists as well as practicing trial of Demographic Factors ■ Chapter 5. The Influence of Personality and consultants who read the book should Attitudes ■ Chapter 6. Incourt Questioning of Prospective Jurors gain a better understanding of the ■ Chapter 7. In-Court Observations of Nonverbal Behavior ■ Chapter 8. current state of research relevant to Overall Effectiveness of Scientific Jury Selection ■ Chapter 9. Additional Trial Consulting Techniques that Aid Jury Selection ■ Chapter 10. Ethical scientific jury selection, and areas and Professional Issues in Scientific Jury Selection ■ Chapter 11. Future where new research needs to be Directions for Scientific Jury Selection ■ References ■ Indexes conducted to advance the field. Attorneys who read the book should be better able to decide whether or not to hire consultants to assist in future litigation, and if so, what types of services these consultants should provide. 2007. 264 pages. Hardcover.

G

Series: Law and Public Policy: Psychology and the Social Sciences / Series Editor: Bruce D. Sales

List: $79.95 • APA Member/Affiliate: $49.95 • ISBN 1-59147-427-2 • Item # 4316081 • ISBN-13: 978-1-59147-427-2

ALSO AVAILABLE MORE THAN THE LAW Behavioral and Social Facts in Legal Decision Making Peter W. English and Bruce D. Sales 2005 • 272 pages • Hardcover • List: $69.95 APA Member/Affiliate: $54.95 • ISBN 1-59147-255-5 Item # 4316053 • ISBN-13: 978-1-59147-255-1

DETERMINING DAMAGES The Psychology of Jury Awards Edie Greene and Brian H. Bornstein 2003 • 238 pages • Hardcover • List: $49.95 • APA Member/Affiliate: $39.95 ISBN 1-55798-974-5 • Item # 431695A • ISBN-13: 978-1-55798-974-1

EXPERTS IN COURT Reconciling Law, Science, and Professional Knowledge Bruce D. Sales and Daniel Shuman 2005 • 162 pages • Hardcover • List: $49.95 • APA Member/Affiliate: $39.95 ISBN 1-59147-246-6 • Item # 4316051 • ISBN-13: 978-1-59147-246-9

APA Books Ordering Information

800-374-2721

www.apa.org/books In Washington, DC, call: 202-336-5510 TDD/TTY: 202-336-6123 • Fax: 202-336-5502 In Europe, Africa, or the Middle East, call: 44-207-240-0856

AMERICAN PSYCHOLOGICAL ASSOCIATION AD0464