COGNITIVE COGNITIVE SCIENCE SCIENCE PERSPECTIVES PERSPECTIVES ON ON PERSONALITY AND EMOTION PERSONALITY AND EMOTION
A...
40 downloads
1156 Views
14MB 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
COGNITIVE COGNITIVE SCIENCE SCIENCE PERSPECTIVES PERSPECTIVES ON ON PERSONALITY AND EMOTION PERSONALITY AND EMOTION
ADVANCES ADVANCES IN IN PSYCHOLOGY PSYCHOLOGY 124 124 Editors: Editors:
G. E. E. STELMACH G. STELMACH VROON P. A. A. VROON E
� m
�
ELSEVIER ELSEVIER Amsterdam A m s t e r d a m-- Lausanne Lausanne -- New New York Y o r k-- Oxford O x f o r d-- Shannon Shannon -- Singapore S i n g a p o r e-- Tokyo Tokyo
COGNITIVE COGNITIVE SCIENCE SCIENCE PERSPECTIVES PERSPECTIVES ON PERSONALITY AND AND EMOTION ONPERSONALITY EMOTION
edited by editedby
Gerald MATTHEWS MATTHEWS Gerald University University of of Dundee Dundee Dundee, Scotland Dundee, Scotland
Ji!! m � 1997
ELSEVIER ELSEVIER Amsterdam Amsterdam - Lausanne Lausanne-- New New York Y o r k- Oxford O x f o r d- Shannon Shannon-- Singapore Singapore-- Tokyo Tokyo
NORTH-HOLLAND NORTH·HOLLAND ELSEVIER ELSEVIERSCIENCE SCIENCEB.V. B.V. SaraBurgerhartstraat Burgerhartstraat25 25 Sara p.o. P.O.Box Box211, 21l,1000 I OOOAE AEAmsterdam, Amsterdam,The TheNetherlands Netherlands
ISBN: 0 444 82450 2
t;)9 11997 997 Elsevier Science B.V. B.V. All All rights reserved. No No part part of this publication may may be be reproduced, stored stored in in aa retrieval retrieval system system or transmitted transmitted in in any any form or by any means, means, electronic, electronic, mechanical, mechanical, photocopying, photocopying, recording or otherwise, otherwise, without without the the ermissions prior prior written permission permission of the publisher, publisher, Elsevier Science Science B.V., B.V., Copyright & & PPermissions Department, Department, P.O. EO. Box Box 52 52 1l,, 1000 1000AM AM Amsterdam, Amsterdam, The The Netherlands. Netherlands. Special Special regulations regulations for readers readers in in the the U.S.A. U . S . A . -- This This publication publication has has been been registered registered with with the the Copyright Copyright Clearance Center Inc. Inc. (Ccq, (CCC), 222 222 Rosewood Rosewood Drive, Drive, Danvers, Danvers, MA MA 01923. 01923. Information Information can be be obtained from from the CCC CCC about about conditions conditions under under which which photocopies photocopies of of parts parts of of this this publication publication may may be be made made in in the the U.S.A. U.S.A. All All other other copyright copyright questions. questions, including including photocopying photocopying outside outside of of the the U.S.A., U.S.A., should should be be referred referred to to the the copyright copyright owner, owner, Elsevier Elsevier Science Science B.V., B.V., unless unless otherwise otherwise specified. specified. No No responsibility responsibility is is assumed assumed by by the the publisher publisher for for any any injury injury and/or and/or damage damage to to persons persons or or property property as as aa matter matter of of products products liability. liability, negligence negligence or or otherwise, otherwise, or or from from any any use use or or operation operation of of any any methods, methods, products, products, instructions instructions or or ideas ideas contained containedin in the thematerial material herein. herein. This This book bookis is printed printed on on acid-free acid-freepaper. paper. Transferred Transferred to todigital digital printing printing2005 2005
List List of of Contributors Contributors
Jean P. P. Banquet·. Banquet*. Neuroscience Neuroscience et et Modelisation, Modrlisation, Institut lnstitut des des Neurosciences, Neurosciences, Jean UPMC, UPMC, 99 quai quai St St Bernard, Bernard, 75252 75252 Paris Paris cedex, cedex, France. France.
Anthony Beech·. Beech*. Department Department of of Forensic Forensic Psychology, Psychology, Fair Fair Mile Mile Hospital, Hospital,
Wallingford, Oxfordshire Oxfordshire OXIO OX 10 9H, 9H, England. England. Wallingford, Jean Claude Dreher. Equipe Equipe de de Traitement Traitement des des Images Images et et du du Signal Signal (ETIS), (ETIS), ENSEAlUCP, ENSEA/UCP, Universite Umversit6 de de Cergy-Pontoise, Cergy-Pontoise, 66 Avenue Avenue du du Ponceau, Ponceau, 95014 Cergy-Pontoise Cergy-Pontoisecedex, cedex, France. France. 95014
Kevin M M. Carlsmith. Department Department of of Psychology, Psychology, Princeton Princeton University, University,
Princeton, NJ 08544, 08544, U.S.A. Princeton, Gerald L. Clore. Clore. Deparment Deparment of Psychology, Psychology, University University of Illinois Illinois at Urbana UrbanaChampaign, Champaign, 603 East Daniel Daniel Street, Street, Urbana-Champaign, Urbana-Champaign, IL 61820, 61820, U.S.A.
Doug/as Douglas Derryberry·. Derryberry*. Department Department of Psychology, Psychology, Oregon Oregon State State University, University,
Corvallis, OR 97331, 97331, U.S.A. Chabot. Department Department of Psychology, University of New Heather Frasier Chabot. Psychology, University
Hampshire, Durham, Durham, NH 03824, 03824, U.S.A. Philippe Gaussier. Equipe Equipe de Traitement Traitement des Images Images et du Signal Signal (ETIS), ENSEAlUCP, Universit6 Universite de Cergy-Pontoise, Avenue du Ponceau, ENSEMUCP, Cergy-Pontoise, 6 Avenue 95014 Cergy-Pontoise Cergy-Pontoise cedex, France.
Wilfried Gtinther. Gunther. Neuroklinik Neuroklinik Bamberg, Bamberg, St Getreu Getreu Strasse Strasse 14-18, Wilfried 14-18, 8600
Bamberg, Germany. Rick E. Ingram. Department Department of Psychology, Psychology, San Diego Diego State University, University, San Rick
Diego, CA 92182-0551, U.S.A. C~dric Cedric Joulain. Joulain. Equipe Equipe de Traitement Traitement des Images Images et du Signal Signal (ETIS), ENSEAlUCP, Universit6 Universite de Cergy-Pontoise, 6 Avenue Avenue du Ponceau, ENSEA/UCP, 95014 95014 Cergy-Pontoise eedex, cedex, France.
Timothy Timothy Ketelaar*. Ketelaar·. Center Center for Adaptive Adaptive Behavior Behavior and Cognition, Cognition, Max Planck Institute Institute for for Psychological Research, Research, Leopoldstrasse Leopoldstrasse 24, 80802 Planck Munich, Germany. Germany.
Contributors
vi Vi
Kitayama*. Faculty Faculty of Integrated Integrated Human Human Studies, Studies, Kyoto Kyoto University, Shinobu Kitayama·. University,
Kyoto 606-01, 606-01, Japan. Kyoto Gerald GeraM Matthews·. Matthews*. Department Department of Psychology, Psychology, University University of Dundee, Dundee, Dundee Dundee DD11 4HN, 4HN, Scotland.
Mayer*. Department Department of Psychology, Psychology, University University of New Hampshire, Hampshire, John D. Mayer·.
03824, U.S.A. Durham, NH 03824, Edward Necka·. Necka*. Instytut Instytut Psychologii, Psychologii, Uniwersytet Uniwersytet Jagiellonski, Jagiellonski, ul. Golebia Golebia
13, 31-007 Krakow, Krak6w, Poland. 13,31-007 Mar jorie A. Reed. Marjorie Reed. Department Department of Psychology, Psychology, Oregon Oregon State State University, University, Corvallis, Corvallis, OR 97331, 97331, U.S.A.
Carien M M. van Reekum. Reekum. Department Department of Psychology, Psychology, Universite Universit6de Geneve, Gen6ve, 9, 9,
Drize, CH CH-- 1227 1227 Carouge-Geneva, Carouge-Geneva, Switzerland. route de Drize, Arnaud Revel. Revel
Equipe Traitement des Images Equipe de Traitement Images et du Signal Signal (ETIS), (ETIS), ENSEAlUCP, Universite de Cergy-Pontoise, ENSEA/UCP, Universit6 Cergy-Pontoise, 6 Avenue Avenue du Ponceau, Ponceau, 95014 Cergy-Pontoise Cergy-Pontoisecedex, cedex, France.
Klaus R. Scherer·. Scherer*. F.P.S.E. F.P.S.E. Section Section Psychologie, Psychologie, Universite Universit6 de Geneve, Gen6ve, 9,
Drize, CH CH-- 1227 1227 Carouge-Geneva, Carouge-Geneva, Switzerland. route de Drize, Siegle*. Doctoral Doctoral Training Facility, San Diego Diego State State University, 6363 Greg Siegle·. Training Facility, University, 6363
Alvarado Court, Court, San Diego, Diego, CA 92120, 92120, U.S.A. Tryon·. Tryon*. Department Department of Psychology, Psychology, Fordham Fordham University, University, Rose Rose Hill Campus, 441 East Fordham Fordham Road, Road, Bronx, Bronx, New York, York, NY 10458-5198, 10458-5198, Campus, U.S.A.
W. W. W.W.
Leanne Williams. University of New England, Williams. Psychology Psychology Department, Department, University England, Arrnidale NSW 2351, Armidale 2351, Australia.
*
* Corresponding Corresponding author
Preface Preface
We are are all cognitive cogmtlve scientists scientists now. now. Researchers Researchers routinely use the We language of cognition cognition in developing models of personality and emotion. emotion. Constructs such as as automatic processing, schemas, working memory, Constructs resources and the like are now part part of the essential fabric of attentional resources theory. The The popularity of information-processing models offers both a promise and a threat. threat. The The promise is that of a true true understanding understanding of how the faculties of perception, attention, memory and so forth different psychological faculties integrated are inter-woven to create the whole person, and to create the mtegrat~ adaptive reactions we call emotions. emotions. Contemporary cognitive science is at ease levels of description and explanation, ease with with multiple levels explanation, and so is especially well-suited to explaining the origins and expressions of emotion and personality. personality. But do we really speak a common language, or are we heading for a new Babel? Constructs Constructs such as schemas and strategies sometimes seem plastic enough to fit almost any theoretical conception, so that the verbal labels become private rather than shared. shared. As subjects of inquiry, emotion and personality are particularly vulnerable to the use of language as artifice rather than as scientific discourse. discourse. The decline of psychoanalysis as a scientific threat. In contemporary research, there enterprise illustrates the nature of the threat. is an evident risk of "cognitivism", dressing up untestable ideas in cognitive jargon. of cognitive jargon. The differing perspectives provided by different strands of research are a strength, not a weakness, but communication between different perspectives requires us to work from common scientific bases. This book aims to highlight the vigour, diversity and insight of the various cognitive science perspectives on personality and emotion. It aims also to emphasise the rigorous scientific basis for research to be found in the integration of experimental psychology with neuroscience, connectionism and the new evolutionary psychology. Collectively, the contributors to this book provide a wide-ranging survey of leading-edge research topics. It is, a little arbitrarily, divided into three parts, on general frameworks for cognitive science, on perspectives from emotion research, and on perspectives from studies of personality traits. In the first, introductory chapter, I begin Part I with a personal view of the impact of the cognitive revolution, and apply the "classical theory" of cognitive science to issues in personality and emotion. As the book took shape, I came to appreciate how much a cognitive science of personality and emotion is necessarily a science of motivation too. In
Pref ace Preface
Vlll
viii
Chapter 2, Mayer, Frasier Frasicr Chabot and Carlsmith inter-relate inter-relate these three constructs constructs in the context of the traditional "trilogy of mind": mind": conation, affect and cognition.
They procr~ proceed to outline a new "quatemity "quatcrnity of mind",
encompassing consciousness consciousness also. One of the most radical and exciting innovations of cognitive science is the use of connectionist models, and the innovations remaining two contributors to Part I provide two different perspectives on their application. Bidirectional Associative Memory (BAM) uses the application. Tryon's Bidirectional conncctionist metaphor of memory as wells in an energy surface as a source connectionist of insight into normal emotion and pathological conditions (Chapter 3). He also outlines how psychotherapy may be directed towards re-Iandscaping re-landscaping the energy surface, by shrinking memory wells whose diameter gives them too much power over the person's experiences, for example. In Chapter 4, Drehcr, Joulain, Revel and Gunther Banquet, Gaussier, Dreher, G0nthcr describe a more ncurologically-orientod conncctionist perspective on personality. They neurologically-oriented connectionist discuss how the person's sense of identity in space and time derives from spatio-tcmporal circuits in hippocampus and prefrontal cortex, supporting spatio-temporal processing, working memory, planning and goal propagation. Part II reviews perspectives derived primarily from emotion research, which explore the interplay between emotion as a common human characteristic and individual difference factors. One of the flaws in an overly unconscious, preattentive prcattcntivc cognitivistic conception of emotion is neglect of unconscious, processes which guide later, attentive processing. Kitayarna Kitayama (Chapter 4) presents the amplification model of affect-cognition affect-cognition interaction in early perceptual processing. The model model describes how the emotional content of rccognition, stimuli may either enhance or impede subsequent conscious recognition, "perceptual defence". dcfencc". Van Reekum Rcckum and explaining phenomena such as "perceptual Schcrcr of Scherer (Chapter 5) also address distinctions between different levels of processing, in the context of appraisal, which may be supported by sensory sensorymotor, schematic or conceptual conceptual processing routines. routincs. They review ncuroscicncr neuroscience bases for appraisal, and link personality to different appraisal Sicgle and Ingram explore connectionist conncctionist characteristics. In Chapter 6, Siegle modelling of the negative biases in cognition characteristic of of depression and other emotional disorders, expressed in appraisal, attention and memory. lcxical decision and valence identification as tasks They focus especially on lexical which bring to thc processing underlying the surface the abnormalities of processing pathology. The pcrspcctivc perspective from evolutionary psychology is presented in Chapter 7 (Ketelaar (Kctelaar and Clore), Clorc), which discusses the long-term adaptive significance of emotions, as informative and motivational signals. The significance authors review evidence suggesting that analysis of the evolved functions of of
Pref ace Preface
IX ix
emotions helps us to understand their more immediate effects on cognition in experimental studies. Part III is oriented towards research on personality traits, within a
superloosely Eysenckian framework, with contributions relating to the three super extraversion-introversion, neuroticism (anxiety) and psychotir factors of extraversion-introversion, psychoticism (schizotypy). Perhaps a future volume of this kind will be able also to cover
additional
dimensions
from
the five factor model; conscientiousness, agreeableness and openness to experience. In Chapter 9, 9, I present a cognitive cognitiveadaptive model of extraversion, extraversion, which reviews information-processing correlates of the trait in the context of adaptive specialisation. Extraverts may be superior in verbal facilities such as short-term recall, retrieval and multi multitasking because these cognitive characteristics contribute to coping with their preferred environments. Derryberry and Read (Chapter 10) discuss the relationship between motivational and attentional aspects of anxiety, anxiety, from the standpoint of cognitive neuroscience. Experimental data illustrate anxietyanxiety related biasing of specific attentional functions which may contribute to anxious individuals. shaping the higher-level cognitions and motivations of anxious Beech and Williams (Chapter 11) assess the cognitive bases for schizophrema schizophrenia and schizotypal personality. They develop a model of activation and inhibition processes which explains priming data obtained experimentally, and the positive symptomatology of schizophrenia such as delusions and hallucinations. Finally, Finally,· contemporary trait researchers are increasingly engaged with exploring the relationships between personality and ability traits. In Chapter 12, Necka links intelligence, extraversion and neurotir neuroticism to an attentional resource model. Both personality and ability arousal processes, processes, whose impact on cognition is shown in traits are related to arousal experimental studies of dual-task performance and memory scanning. scanning. I am grateful to the Medical Medical Research Council for their support for my research while this to thank the this book was in preparation. I would also like to contributing reading and re-reading re-reading the chapters, contributing authors. I have enjoyed reading chapters, and my schemas and networks networks are greatly enriched. This This is the book I would have liked to to have have read read when I first began began researching personality personality and emotion as as a doctoral doctoral student in the early 1980s. I hope it will serve as as an inspiration and a guide guide to to all those with an an interest in this this exciting new research area.
(Jerald Matdlews Gerald Matthcws
This Page Intentionally Left Blank
Contents Contents
PART P A R T I. I. FRAMEWORKS F R A M E W O R K S FOR F O R COGNITIVE C O G N I T I V E SCIENCE SCIENCE
Chapter 1. 1. An Introduction to to the Cognitive Science Science of Chapter An Introduction Personality and Emotion
.................. . .......... . ... ....... .... . ... . ......... 3 3 .......................................................................
.
.
.
.
..
.
.
.
.
.
.
.
Gerald Matthews Landmarks the Cognitive Landmarks of of the Cognitive Revolution Revolution .............................................. 3 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . .
A Cognitive Science A Cognitive Science Framework Framework
...................................................... 77 Towards aa Cognitive Neuroscience of Personality and Towards Cognitive Neuroscicnce of Personality and Emotion? Emotion? ......... 13 13 Developing Adaptive Adaptive Explanations Developing Explanations .................................................. 15 15 An Anxiety and An Example: Example: Explaining Explaining Anxiety and Cognition Cognition .............................. 20 20 Conclusions Conclusions .................................................................................... 24 24 . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . ... . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter 2. Conation, 31 Conation, Affect, and Cognition Cognition in Personality................... ................... 31
M. Carlsmith John D. Mayer, Heather Frasier Chabot and Kevin M The Relational The Relational Model Model of of Personality Personality .................................................. 32 32 Understanding and Cognition Understanding Conation, Conation, Affect, Affect, and Cognition ................................. 39 39 The of Mind Mind and and Personality The Quatemity Quaternity of Personality Dynamics Dynamics ............................ 52 52 Conclusions Conclusions and and Other Other Considerations Considerations .............................................. 60 60 . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .. . . . ..... . . . . . . . . . . . . . . .... . . . .
. . . . . . . . . . . . . . . . . . . . .. . . . . . .
. . . . . .. . . . . . . . . . . . . ... . . . .. . . .. . . . . . . . . . . . . . . .
Chapter 3. Introduction to the Bidirectional Associative Memory Model: Implications for Psychopathology,
Treatment, and Research ............... ....................................................................... 65 Treatment, ... . .... . . . . .. .. ...... ........... . . .. 65 .
.
.
.
.
. .
.
.
.
.
.. ..
..
..
�arren Warren � W. T�on Tryon Bidirectional Associative Memory Memory (BAM) Bidirectional Associative (BAM) ......................................... 67 67 . . . . . . .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .
Encoding Emotion Emotion ............................................................................. 70 70 Encoding . . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Implications for for DSM-IV DSM-IV Disorders Disorders .................................................. 75 75 Implications Treatment .........................................................................................92 Treatment 92 Research Strategies ........................................................................... 99 99 Research Strategies Conclusions .................................................................................... 1101 0I Conclusions Appendix: Description Description of of the Bidirectional Associative Associative Memory Appendix: the Bidirectional Memory ....... 109 109 . . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . ..... . . . . . . . . . . . . . . . . . ... . . ... . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . .
xii XII
Contents Chapter Chapter 4. 4. Space-Time, Space-Time, Order, Order, and Hierarchy Hierarchy in FrontoFronto-
...................... 123 Hippocampal Hippocampal System: A A Neural Neural Basis of of Personality Personality .... .......................... Jean dean P. Banquet, Banquet, Philippe Gaussier, Jean Claude Dreher, Dreher, Cedric Cddric Joulain, /fried Gunther Joulam, Arnaud Revel and Wi Wilfried G~tnther Hipp Hippocampal ocampal Function: Function: An Extended Extended View View ...................................... .. . .. 126 126 .
.................. ....
......
..
Working Memory as Both aa Cortical Working Memory as Both Cortical and and aa Hippocampal Hippocampal
System ...................................................................................... 129 . ..... . . . . ... ..... . . . .... . 129 System .
.
.......
...
..
.............
........ ....
.
.
.
....... .........
Neuropsychology, Neuropsychology, Brain Brain Imaging Imaging and and Working Working Memory Memory ................... 135 135 ...................
Neurophysiology: Neurophysiology: Human Human Versus Versus Animal Animal Working Working Memory Memory ............ 148 148 ............
Spatio-Temporal Processing Processing in in Hipp Hippocampus and Prefrontal Prefrontal Spatio-Temporal ocampus and
Cortex ...................................................................................... 151 Cortex 151 .... ............................. ............ ........ . . ................. . . . . . .. . . .....
Functional ... . .. .. ... ... . . . .. .... . . .. . 159 Functional Model Model .............................................................................. 159 ...
......
.
.....
.
..
..
.
.
...
..
.. .
.
...............
Fronto-Hipp ocampal Function . .... .. . . . 176 Fronto-Hippocampal Function and and Personality Personality ................................ 176 ......
........ .
.
.
..
...
Conclusion . . . .. . ... . .. . 179 Conclusion ..................................................................................... 179 ...........
.... . .
. .. . . .......
..................
.
.
..........
.
............
PART II. PERSPECTIVES PERSPECTIVES FROM FROM EMOTION EMOTION RESEARCH RESEARCH PART Chapter Chapter
5. Affective Affective Influence in Perception:
.................................... 193 Some Implications Implications of of the Amplification Amplification Model ...................................... Some
Shinobu Kitayama The Model of of Affect-Cognition Affect-Cognition Interaction Interaction ................. 196 The Amplification Amplification Model ..... .. 196 Evaluation Criteria Criteria of of the the Amplifi cation Model . ... . 202 Evaluation Amplification Model ................................ Experiment Experiment 11 .................................................................................. .................................................................................. 212 Experiment Experiment 2 .................................................................................. .................................................................................. 221 221 The Amplification Model .. . 230 The Amplification Model Evaluated Evaluated ................................................. Relations with with Extant Extant Theories Theories of of Attention Attention ..................................... Relations . .. . . . . . .... 232 Amplification Attention in in Other Other Domains Domains ................................... Amplification of of Attention .. .. . . 235 Perceptual Defense Defense and and Vigilance? Vigilance? .................................................... Perceptual . .. . 238 Future . . ...... .. .. .. 240 Future Research Research Directions Directions ............................................................. Concluding . . .. . . ... . 242 Concluding Remarks Remarks ... ....................................................................... .
.
....
.
.
....
..
.......
.
. ..............
.................... .. . . .
.................
..
.
.
.
....... .. . . . .. . .....
... .
.......
..... ...............
...................... . . ..
..
... ......
......... .. . .
....
. . .....
.... ........................
.
......... . . . . ...
.
.
............
.
.......
.............
.. ......
Contents Contents
Xlll xiii
Chapter 6. Levels Levels of of Processing Processing in in Emotion-Antecedent Emotion-Antecedent Chapter Appraisal Appraisal
.............................................................................................. 259 259
..............................................................................................
Carien M M. van van Reelcum Reekum and Klaus R. R. Scherer Scherer Carien of Appraisal Notions ......................................................... ......................................................... 260 260 Critique of of Processing in in Appraisal .................................................... ........................... : ........................ 263 263 Levels of in Related Traditions Traditions .......................... .......................... 266 266 Hierarchical Process Notions in in Rewriting Appraisal Theory .............................................. .............................................. 277 277 Issues in in Appraisal Appraisal Processes Processes .................................. .................................. 280 280 Individual Differences in Conclusions .................................................................................... .................................................................................... 289 289 Conclusions
Modeling Individual Differences in Negative Chapter 7. Modeling Information Processing Processing Biases Information
.............................................................. 301
..............................................................
Greg.1. J Siegle and Rick E. Ingram Greg Personality Research and Vulnerability Vulnerability to Depression: Depression: A History History ...... ...... 302 302 Personality Simulating Aspects Aspects of Depression Depression and Personality Personality on a Simulating Computer ....................................................................................... ....................................................................................... 304 304 Computer Simulating Personality Personality Factors Factors ........................................................ ........................................................ 320 320 Simulating Conclusion ......................................................................... ......................................................................... 348 348 A Brief Conclusion
Chapter 8. Reason" The Proximate Effects Effects and Chapter 8. Emotion Emotion and Reason: The Proximate Ultimate Functions of Emotions ........................................................... 355 Ultimate Functions of Emotions 355 .......................•...................................
Timothy Ketelaar GeraM L. Clore Timothy Ketelaar and and Gerald Why ............................................. 356 356 Why Does Does Emotion Emotion Affect Affect Cognition? Cognition? ............................................. Specific .......................................................... 358 358 Specific Aims Aims of this this Chapter Chapter .......................................................... Consequences of Mood ................................................................... 360 Consequences Mood ...................................................................360 Consequences .............................................................. 365 365 Consequences of Emotions Emotions .............................................................. Emotion-as-motivation and and Frank's Frank's (1988) (1988) Commitment Commitment Model Model ....... Emotion-as-motivation ....... 371 371 Affect-as-lnformation ................................................ 378 378 Affect-as-Information and and Behavior Behavior ................................................ The Processing ............................. ............................. 387 387 The Future Future of Affect Affect and and Information Information Processing Conclusion: ..................................... 388 388 Conclusion: Deficits, Deficits, Biases, Biases, and and Functions Functions .....................................
xiv XIV
Contents III. P PERSPECTIVES PPART A R T IIl. E R S P E C T I V E S FFROM R O M PPERSONALITY ERSONALITY RESEARCH TTRAIT RAIT R ESEARCH
Extraversion, Emotion Emotion and and Performance: Chapter 9. Extraversion, Chapter A Cognitive-Adaptive Cognitive-Adaptive Model Model................................................................ 399 399 A ................................................................
Gerald Matthews ..
Extravorsion .................................................................. 400 400 Extraversion and and Affect Affect .................................................................. Extraversion 405 Extraversion and and Performance Performance ......................................................... .. ... . .... . . ....... ..... . 405 Extraversion, 409 Extraversion, Arousal Arousal and and Attontion: Attention: Empirical Empirical Studies Studies ................... ... ....... . .409 An Adaptivr Adaptive Framowork Framework for for Cognitive Cognitive Correlates Correlates of of An Extraversion-Introversion ................................................................. .... .. .... . .... ... .. . ...... . .. . 426 426 Extraversion-lntroversion Conclusions .................................................................................... .. . .... . . ....... .. .. ... . . . ......... ... . . 434 Conclusions 434 .
. .......
..... ..
.
...
......
.
...
.
..
.. ..
.
.
.
.
.........
...
...
... .
...
.
........
.
.....
...
....
.
.
.
...
.
....
. ..
.
....
..
...
...
.
..
of Chapter 10. Motivational and Attentional Components of 443 Personality ............................................................................................ 443 ............................................................................................
Douglas Derryberry jorie A. Reed Derryberry and Mar Marjorie Biological .............................................. 444 Biological Approachos Approaches to to Personality Personality ..... .. . ... .... ........ ... .444 Assessing Attentional Processes in ......... . .. ...... ..... ...450 Assessing Attcntional Processes in Anxiety Anxiety ..................................... 450 Extensions to Complex . .. ... .......... ...... . .462 Extensions to Complex Cognitive Cognitive Processing Processing ................................... 462 Conclusions .. ............ . ........ .. .. . . ... ... . .. .... .. . ..... .466 Conclusions .................................................................................... 466 .
.......
..
....
..
.
...
.
.
.
.
.
.....
..
..
.
.
....
.
...
..
.
.
..
. ..
.
.
...
..
...
.
..
...
.
.
...
Chapter 111. 1. Investigating Cognitive Processes in Schizotypal Personality and Schizophrenia
.......................................... .475 475
...•....•.••••.••••••........•.•••.......
Anthony Beech Beech and Leanne Williams Williams Mechanisms . ... .. ...... .. . .. . . ....... ....477 Mechanisms of of Selective Selective Attention Attention .................................................. 477 Experimental Investigations Investigations of of Inhibitory Inhibitory Processes Processes ........ ......................... 478 Experimental ........ . ...478 Inhibitory Inhibitory Processes Processes in in Schizophrenia Schizophrenia ............................................. .............................................. .485 485 Towards aa "Reduced "Roducexl Cognitive Cognitive Inhibition" Inhibition" Model Model of of Towards Schizophrenic Schizophrenic Symptomatology Symptomatology ...................................................... ...................................................... 490 490 Revising .. . ... .... .. ..... . . ........... ........ . . . . ... .. 494 Revising the the Model Model .......................................................................... 494 Conclusion . . .. .. . .. ..... . ... ...... .. .. . ... .. . 497 Conclusion ....................................................................................... 497 ... .
.
.
.
.
......
..
.
..
.
..
.
. .......
.
..
.
.
..........
..
..
........
.
.
...
....
..
. .
......
.
.
.
.
..
.
.
.
.
.. .
.
.
.
.. .
.
.
xv xv
Contents Chapter
Working Memory Memory and Arousal: 12. Attention, Working
Concepts Concepts Apt to Account Account for the "Process "Process of Intelligence"
................. 503
.................
Edward Necka Theoretical Notions Notions ......................................................................... Theoretical Assumptions ................................................................................... Assumptions "The Process Process of o f Intelligence" Intelligence" ........................................................... "The Preliminary Empirical Data ............................................................. Preliminary Empirical Data Cognitive Science Science Perspectives Perspectives ........................................................ Cognitive
504 504 512 512 519 525 525 542 542
index ......................................................................................... Subject index
555
. . . . .. . . . . . . . .. . . . .. . . .. . . . . . . . . . . . . . .. . . . . . . . . ..... . . . . ... . . . . . . . . . . . . . .
. . . . .. . . . . . . . . . . . . . . . ... . . . . . . . . . . .. . . . . .. .. . . . . . . . . . . . ... . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . ... . . . . .. .. . . . .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . .
.. . . . . . . . . . . . . . . . . . . . . . . . . ... . . .. .. . . . ... . . . . . .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . .
.........................................................................................
This Page Intentionally Left Blank
PART P A R T II FRAMEWORKS FOR COGNITIVE FRAMEWORKS FOR COGNITIVE SCIENCE SCIENCE
This Page Intentionally Left Blank
Perspectives on Personality Personality and Emotion Emotion-G. Matthews (Editor) (Editor) Cognitive Science Perspectives G. Matthews -
1997 V. 1997 Elsevier Science B. B.V.
CHAPTER 1 CHAPTER!
An Introduction Introduction to the Cognitive Cognitive Science of Personality Personality An Emotion and Emotion GeraM Matthews Gerald
The cognitive revolution has transformed the face of research on personality and emotion. Information-processing theories spring up like poppies in a cornfield, and often wither just as quickly, crowded out by more recent growth. With cognitive approaches so firmly established, it is timely to stand back a little from the intellectual ferment, and take stock of the achievements and limitations of the research area. This book represents the leading edge of research on the cognitive science of personality and emotion, contributions concern a variety of specific topics. But for and so the contributions personality and emotion research to mean anything at all, it must above all be integrative. Both constructs hang on a multi-layered web of data and hypothesis, spanning the gamut of psychological phenomena from neuronal firing to social interaction. The aim of this introductory chapter is to outline the overall framework provided by cognitive science, and its place in personality and emotion research. research. In this chapter, I will sketch the progress so far of the cognitive revolution in personality and emotion (PE) research. research. I will then describe the classical model of cognitive science, and its three levels of explanation: the biological, the symbol-processing and the knowledge levels. Cognitive science emphasises that information-processing models are necessary but not sufficient for understanding, understanding. lI will show that cognitive explanations provide new perspectives on some old problems, and science explanations demonstrate its integrative potential by outlining its application to anxiety. anxiety.
Landmarks Landmarks of the Cognitive Revolution Emotion and Emotion and cognition cognition
The cognitive science of emotion has disparate roots, roots, which demonstrate the diversity of "cognitive" approaches. The information-processing approach is based on empirical, empirical, performance-based studies of emotion, addressing problems such as the deleterious effects of anxiety on attention. It
4
Chapter 1
accommodates the the various various conccptualisations conceptualisations of emotion: as as aa universal universal but accommodates situationally-contingent human response, as as an individual individual difference factor, and as as aa property of stimuli (valence). (valence). Emotion may be conccptualiscd conceptualised as as aa and dependent variable influcnce~ influenced by processes such as as appraisal, or as as an independent variable which itself influences influences information-processing. The independent more sophisticated applications 1984) build in applications of the approach (e.g. Ingrain, Ingram, 1984) feedback from appraisals into emotion. In linking appraisals of performance back into emotion to bchaviour, behaviour, the basic research tactic is to demonstrate moderation of effects of emotion on performance standard technique performance by task factors, the standard of experimental experimental cognitive psychology. Emotion •x task parameter interactions inspire processing models which may then be subjected to further test. The approach scores highly on scientific rigour, but like much cognitive psychology, risks degenerating into an account of the minutiae of a specific experimental paradigm with little wider relevance (cf. Neisser, Neisscr, 1976). 1976). An alternative approach is design-oriented: design-oriented: what might be the purpose of emotion within the cognitive system as a whole? Simon (1967) linked emotion to an interrupt function, contributing to people's capacity to to adapt to unpredictable environments by switching back and forth between different goals. This analysis begs the question of why the interrupt function requires all the various concomitants of emotion such as physiological arousal, biases in thinking, action tendencies and the like. like. Research in the Artificial Intelligence (AI) tradition simulates complex, goal-directed systems to ple, what other features discover basic design principles, indicating, for exam example, are required for interrupts to arc to work properly. This approach generates rich and thought-provoking data, but its scientific rigour is open to question. Argument tends to proce~ proceed by analogy and comparison of features of artificial and human systems, and it is unclear that the parallels drawn are arc open to falsification or to formal test against alternative explanations. A third tradition derives from stress and clinical research, and the observation that negative emotions derive from the way people interpret and manage events, rather than from fixed properties of the events themselves. It 1984) on is exemplified by the work of Lazarus (1991; Lazarus & Folkman, 1984) the transactional transactional model of stress, and the roles of appraisal and coping within specific, potentially stressful encounters. As theory, it has some of the characteristics of both the design and information-processing approaches. Like the design approach, it is explicitly systems-based, with emotion conceptualised as a "core relational theme" characterising conceptualiseA charactcrising the person personenvironment system as a whole. However, like the information-processing tradition, transactional theory attempts to establish local cause-and-effect
G. Matthews
5
relationships open to direct empirical test, such as the effect of appraisals on emotion. Similarly, clinical accounts of anxiety and depression which emphasise the role of the person's self-knowledge and reasoning processes in generating negative emotion as an overall indicator of system functioning (e.g. Beck, 1967; 1967; Ellis, 1962). 1962). The advantages of such approaches are depth of insight obtained into the experiences of people in real environments, and practical applications to stress management and cognitive behaviour therapy (Matthews & Wells, 1996). 1996). Their shortcomings relate, first, to emphasis on self-report data, which may present a partial and distorted view of underlying processing, and, second, as with the design tradition, to difficulties in rigorous theory testing. Finally, the neuroscience of emotion has become increasing cognitive in orientation, as traditional arousal theory has fallen from favour (e.g. Robbins, 1986). 1986). Increasingly, it has becomes possible to align specific neural circuits with information-processing and behavioural function (e.g., Gray, 1982). 1982). Emotion is notoriously difficult diffmult to localise, but advances in brain scanning technology, and in simulation of neural function are a source of optimism for the future. Pessimists focus on the extent to which feelings are intertwined with thinking, and consequent difficulties in discriminating neural and influences. There remain fundamental disagreements over the extent cognitive influences. to which psychological phenomena are reducible to neural processes (see 1984, 1991, 1991, and Gazzaniga, 1992, 1992, for the end-points of the Lazarus, 1984, continuum of views). At the least, though, computational theories permit testable predictions concerning neural influences on behaviour, contributing to the development of cognitive neuroscience models of emotion.
and cognition Personality and Much of personality research is structure- rather than process-oriented, and so unaccommodating to cognitive perspectives. The current popularity of the Five Factor Model owes much to the prodigious empirical progr programmes ammes of researchers such as Costa and McCrae (1992) in deriving the Big Five as a structural description of various data sets. Personality trait theories have often been based on somewhat naive biological or conditioning models, J.B. Watson rather than by contemporary research. inspired by Pavlov and J.B. blessing. The concept Arousal theory, in particular, has proved to be a mixed blessing. undoubtedly has integrative value (K.A. Anderson, 1990), 1990), and the basic principle that personality reflects biology is becoming increasingly securely (Loehlm, 1992: 1992: Lesch supported by behaviour and molecular genetic studies (Loehlin,
6
Chapter 1 Chapter
ct 1996). Eyscnck et al., al., 1996). Eysenck and and Eyscnck's Eysenck's (1985) (1985) application application of of arousal arousal theory theory to to personality personality has has scored scored some some notable notable empirical empirical successes successes in in predicting predicting extraversion-introversion effects effects on on sensory sensory thresholds thresholds and and simple simple cxtraversion-introvcrsion conditioning conditioning tasks. tasks. Unfortunately, Unfortunately, psychophysiological psychophysiological data data on on personality personality are confusing confusing and and inconclusive, inconclusive, and and arousal arousal theory theory has has proved proved to to be be aa poor poor arc basis for for predicting predicting personality personality effects effects on on cognitive cognitive tasks tasks (e.g. (e.g. Matthews, Matthews, basis Matthews & & Deary, Deary, in in press). press). 1985; Matthews Despite Despite the the conservatism conservatism of of much much personality personality research, research, there there arc are increasing signs that that the the cognitive cognitive revolution is taking taking root root in in this this area area also. also. increasing signs revolution is of emotion, emotion, its its expressions expressions arc are diverse. diverse. Information-processing Information-processing As As in in the the case case of analyses analyses of of personality personality effects effects on on performance performance arc are becoming becoming increasingly increasingly common. common. The The trail trail has been been blazexl blazed by by research research on on anxiety anxiety traits, traits, driven driven by by that cognitive cognitive worry worry is is more more predictive predictive of of performance performance than than the observation observation that the emotional and and physiological physiological tension. tension. Detrimental Detrimental effects effects of of anxiety anxiety arc are now now emotional routinely routinely explained explained in in terms terms of of constructs constructs such such as as attentional attentional capacity capacity (Sarason, Sarason, & 1995) and (Eyscnck, 1992). 1992). (Sarason, Sarason, & Pierce, Pierce, 1995) and working working memory memory (Eysenck, Humphmys Rcvr162 (1984) have Humphreys and and Revelle have proposed proposed an an ambitious ambitious integration integration of of individual differences research which anxiety individual differences research which links links achievement achievement motivation, motivation, anxiety and to arousal and effort, effort, which which in availability of and impulsivity impulsivity to arousal and in tum turn influence influence availability of multiple and working multiple resources resources for for performing performing attcntional attentional and working memory memory tasks. tasks. There is also aa rather rather separate tradition with with aa basis basis in in social-cognitive There is also separate tradition social-cognitive psychology, concerned concerned with with the the knowledge knowledge structures which support psychology, structures which support personality, the self-schema personality, such such as as the self-schema (Cantor (Cantor & Zirkel, Zirkcl, 1990). 1990). This This approach approach supports work, such supports some some information-processing information-processing work, such as as studies studies of of self-referent self-referent processing (Klein (Klein & Loftus, Loftus, 1988) 1988) and and priming priming (Bargh, (Bargh, Chaiken, Chaikcn, Govender, Govcndcr, processing & Pratto, Pratto, 1992), 1992), but but also also leans leans heavily heavily on on qualitative qualitative and and self-report self-report data. data. Hence, Hence, itit resembles resembles the the transactional transactional approach approach to to emotion: emotion: its its allegiance allegiance is is to to cognition cognition but but not not necessarily necessarily to to cognitive cognitive science. science. On On the the other other hand, hand, itit is is sufficiently to both both nomothetic sufficiently flexible flexible to to be br applied applied to nomothctic and and idiographic idiographic aspects aspects of of personality, personality, and and engages engages with with individuals' individuals' actual actual life life experiences. experiences. Integration of ofpersonality and and emotion emotion research research
The distinction distinction made made between between personality personality and and emotion emotion is is artificial artificial to to the the The extent extent that that much much personality personality research research has has an an explicit explicit trait-state trait-state orientation, orientation, within within which which personality personality effects effects are arc mediated mediated by by emotional emotional states states (e.g. (e.g. Spielberger's, Spiclbcrgcr's, 1966, 1966, anxiety anxiety theory). theory). We We cannot cannot do do personality personality research research without without consideration consideration of of emotion, emotion, but but the the converse converse also also applies. applies. Some Some studies studies of mood mood make make aa strong strong equation equation between between positive positive and and negative negative affect affect on on the the of
G. Matthews G. one hand, and extraversion and neuroticism on the other.
77 Individual
differences in mood may substantially reflect individual differences differences in reward differences and punishment systems said to be the basis for extraversion and neuroticism (Watson
& Clark, 1992). Unfortunately, taken to the extreme, this approach
leads to a dreary tautology, such that some unfortunates have negative genes, negative brains, negative emotions and negative personalities, and little more can be be said.
More promising are interactionist approaches which emphasise that
differences in emotional response are not mechanically linked to individual differences personality, but depend on a more complex interplay between person and
environment. Within the transactional model, personality is seen as biasing environment. the appraisal and coping processes which are perhaps more direct influences on emotion (Matthews & Deary, in press). Interactionism can easily degenerate into an unfalsifiable everything-affects-everything position, but can potentially supply much needed precision to theory computational models can in this area. Information-processing analyses of performance frequently effects on different different processing attempt to discriminate trait and state effects trait components. Eysenck's (1992) review of the area suggests that simple traitstate models, within which trait effects are entirely mediated by gross state constructs, are not viable: trait anxiety may sometimes influence cognition and behaviour even with state anxiety controlled. Integration of trait and state research requires a more sophisticated view, such that traits affect stable parameters of processing systems which moderate their reactions to to stimuli. We might link traits traits to knowledge structures in long-term memory (LTM) (LTM) which feed into appraisal and coping (Wells & Matthews, 1994), 1994), or, from a connectionist perspective, to parameters of networks which govern the spread of activation (Matthews (Matthews & Harley, 1993). 1993). In either case, moderating effects of traits traits are apt to be subtle, and require careful modelling.
A Cognitive Framework Cognitive Science Framework The brief overview above demonstrates the vigour of the cognitive cogmtIve approach to P PE. approach E. It also shows that progress has been uneven, and the diversity of differing "cognitive" approaches. We require a general framework for examining where cognitive research has been most successful, and where its impact has so far been limited. "classical theory" of limited. Fortunately, the "classical cognitive science provides a ready made framework, discriminating different levels of explanation. Next, I will outline these levels, and discuss their application to PE research.
8
Chapter 1I
Pylyshyn (1984) presents a detailed analysis of knowledge, symbol symbolprocessing and biological levels of explanation, from which the following account is derived (see Figure 1). 1). The central point is that psychological events are open to qualitatively different explanations. Suppose we observe an extraverted man man at a party, engaging in cheerful social interaction. How do behaviour? One approach is to refer to we explain this bohaviour? to his motives and goals. Perhaps he is a newcomer, and wishes to to make new friendships from which he will benefit. This level of explanation is the knowledge-based or semantic level. It is concerned especially with the way the cognitive system is designed for adaptive interaction with the external environment, in pursuit of its goals. It has been developed in PE research through AI approaches approaches to understanding emotion, through work on the adaptive functions of PE, and through social knowledge approaches approaches to personality. Alternatively, we might present an account based on the formal cognitive architecture: a computational description of the processing structures and operations linking inputs to social behaviours. We may then identify spee~ of accessing individual differences in specific computations, such as speed items of social knowledge, which explain the individual's social behaviour at the processing level. Explanations of this kind are concerned with the formal characteristics characteristics of processing, rather than with the adaptive significance of processing routines. They provide the basis for much of the extensive research on information-processing information-processing models of emotion and personality previously described. Classical theory requires the architecture to be based on discrete symbols, expressing propositions. Pylyshyn distinguishes sub-levels of algorithm and functional architecture, which differentiate the logical operations performed on symbols form the cognitive structures implementing symbol processing. The centrality of symbols is a controversial area. Some authors place symbol-based accounts of processing centre stage, due to identifiability problems of modelling functional architecture (J.R. (J.R. Anderson, 1990). 1990). Conversely, connectionist models see network implementations as a more powerful method for modelling behavioural data than symbolic accounts, and may even reject symbolic representations as irrelevant to theory accounts, (Smolensky, 1988). 1988). I will take the view that, in the light of the successes of connectionism, an a priori commitment to to symbolic accounts may be too constraining for PE research. I will use the term "architectural explanation" to refer to explanation in terms of the formal properties of the processing machinery, irrespective of whether or not it is symbolic in nature.
G. Matthews Matthews G.
Knowledge = Knowledge
Symbol Symbol
processing processing
Biology Biology
<
Goals, intentions intentions and and personal personal Goals, meaning, meaning,supporting supporting adaptation adaptation to to external environments environments external
Algorithm
= Formal Formalspecification specification of of program program for for symbol symbol manipulation manipulation
Algorithm
=
Functional Functional
_
/
=
99
Architecture Architecture
Real-time Real-timeprocessing processing operations operations supporting supporting symbol symbol manipulation manipulation
Physical, neuronal neuronal Physical, representationof of representation processing processing
1. Levels of explanation in cognitive science. Figure 1.
Finally, we may look to the functioning of the neural hardware for
explanation. We might use brain scanning techniques to investigate investigate which interaction, and develop neural structures and circuits are active during social interaction, a theory linking the individual's social behaviour to the activity of the circuits concerned. We must then tackle transduction transducaon problems; the conversion of analogue physical events events into symbolic codes (Pylyshyn, (Pylyshyn, 1984), 1984), or other abstract codes. In the next section of this chapter, I develop the position that information-processing models of PE are necessary but not sufficient sufficient for understanding. Processing models possess the rigour provided by understanding. computational specification, specification, and, if adequately formulated, are readily testable against empirical data. However, a processing description of PE phenomena requires supplementation with explanations which look both downwards, to architecture and cognitive neuroscience, neuroscience, and upwards to knowledge-level knowledge-level explanations. Information-processing Information-processing models." models: Strengths Strengths and and limitations limitations
Information-processing models of personality and emotion have an impressive track track record in characterising empirical phenomena in terms terms of constructs constructs such as as resources, processing stages and activation of network units. units. The The application of such models is demonstrated throughout this this
10 10
Chapter I1 Chapter
volume. volume. Multi-level Multi-level models, models, distinguishing qualitatively qualitatively different types types of processing, processing, such such as as stimuhs-drivcn stimulus-driven and and strategic strategic processing, processing, have have been particularly particularly successful successful in explaining empirical data data (see van Rcekum Reekum & & Scherer, this this volume). volume). Processing Processing models models are are essential essential for predicting and and Scherer, understanding E, and they arc increasingly finding understanding the correlates correlates of P PE, are increasingly applications in the clinical clinical domain (see Beech & & Williams, Williams, Sicgle Siegle & Ingrain, Ingram, applications Tryon, important to be clear about what such Tryon, this this vohmc). volume). However, However, it is important models models provide and do not provide. provide. Most models provide a snapshot snapshot description description of processing at a single time epoch, although there is growing Kanfer & & Ackerman, Ackerman, 1989). Such a interest in learning models (e.g. Kanfcr description leaves leaves open alternative alternative types types of explanation. explanation. The first question is description whether effects effects of PE PE factors factors on on processing processing reflect genuine differences differences in whether cognitive architecture, or differences differences in strategy, i.e. how the same architecture is used to support different processing sequences within a given context. It is unlikely that PE has has dramatic effects on architecture; we would not expect syntactic deep structure to vary across individuals, for example Perhaps more likely are quantitative cross-individual (cf. Pinker, 1994). Perhaps cross-individual or cross-occasion differences in system components such as resource cross-occasion availabilities, short-term memory (STM) slots and speed of execution of key availabilities, processes (e.g. Nccka, Necka, this vohmc). volume). The architecture may also handle processes neutral stimuli (Kitayama, emotional stimuli stimuli differently to neutral (Kitayama, this volume). Architecture Architecture as as a a source source o of f variation variation needed in showing that variance in processing reflects variation Care is needezl in architecture, as opposed to variation in strategy and intention (Pylyshyn, 1984). A strategy may be defined as a goal-directed, goal-dircctexi, voluntarily-initiated voluntarily-imtiatcd processing routine. routine. Typically, a strategy is implemented implcmcnteA and regulated through executive processes processes which bias involuntary processing (see Norman & Shallice, 1985). There are are. rather few instances of attempts to establish systematically whether PE phenomena are strategy-dependent, although effects of emotion on strategy-insensitive processes such as early stimulus Picketing, & analysis (Kitayama, this volume) and procedural learning (Corr, Pickering, Gray, 1995) are suggestive of architectural differences. differences. More generally, processes processes of interest depend on both the fixed architecture and strategy, and it is difficult to disentangle the two types of influence. For example, extraverts tend to show greater STM STM capacity than introverts (Matthews, 1992), but this effect might reflect either individual differences in cognitive architecture, perhaps derived from physiological processes (Eysenck & Eysenck, 1985), or
G. Matthews G.
11 11
coding strategies strategies which tend tend to enhance short-term short-term from extraverts' choice of coding 1975). recall at the expense of long-term recall (Schwartz, 1975). is established, explanatory questions If a PE effect on architecture is in architecture architecture reflects reflects relatively relatively remain. One possibility is that PE variance in straightforward properties of the brain. The neural substrate for emotional timestates may influence the formal properties of processing over short time scales. Given the heritability of personality traits, including traits related to emotionality, it is plausible that genes code for individual differences in architecture. Alternatively, the architectural difference may be more readily conceptualised as a learning effect, such as changes in control structure 1982). We may associated with "proceduralization" of knowledge (Anderson, 1982). also ask if individual differences result from biological bases for learning, or from socially-influenced exposure to learning opportunities: each level of explanation poses further questions. Strategy choice and adaptation
PE effects on processing may derive not from architecture but from choice. Architectural accounts of strategy implementation which strategy choice. 1988) are important but describe specific executive functions (e.g. Shallice, 1988) incomplete. We need also to address knowledge level questions concerning the person's goals, and choice of strategy to meet those goals. Again, answers generate new questions. How has the person acquired the goals concerned? How does the person's knowledge of strategies, such as strategy efficacy in the current context, feed into strategy choice? At one level we can answer such questions through addressing the cognitive and social factors which 1977). influence motivations and associated learning (e.g. Bandura, 1977). Understanding strategy choice may requires understanding of the shaping Shaping of cognition within the wider social matrix, through the person's attempts to meet social norms, negotiate shared identities with others, and generally adapt to social demands (Hampson, 1988). 1988). Mayer, Frasier Chabot and Carlsmith (this volume) provide a detailed discussion of the inter-relationship between motivation, emotion and cognition. cognition. A radically different perspective is provided by evolutionary psychology (Tooby 1992). The person's most important life goals are (Tooby & Cosmides, Cosmides, 1992). influenced by by the set of genetically programmed programmed mechanisms for solving specific evolutionary problems. Some proximate goals such as "stay warm" may be directly coded. generally, the individual's goals are indirectly coded. More generally, influenced by the structuring of experience imposed by the set of adaptive
12
Chapter J1
mechanisms, which, at the least, is likely to signal that certain types of stimuli and encounters are of special significance. In particular, the motivations which tend to accompany emotional states (e.g. avoidance as a correlate of anxiety) are likely to reflect adaptive pressures. pressures. There is an argument too that anxiety) specific strategies, such as the decision rules used in "Prisoner's Dile mma" Dilemma" social encounters (Ketelaar & Clore, this volume), may be directly encoded (Cosmides & Tooby, 1992). 1992). However, the evolutionary psychologist's description of a "strategy" carries no commitment to to a particular information informationprocessing mechanism. The strategy might be implemented through architecture, or, alternatively, through coding cxxling for motivational factors. factors. Evolutionary psychologists have perhaps shown insufficient interest in whether strategies in the evolutionary sense are contingent upon implementation of strategies in the information-processing sense previously defined. Strategies for processing reflect voluntary control and potentially defined. complex, contingent decisions which may not be related to genetically geneticallyprogrammed programmed adaptations in any simple way. Evolution is an essential part of the backdrop to understanding the inter-relationship of PE and cognition, but it is simplistic to imagine that every such relationship may be traced back to the operation of an adaptive mechanism (cf. (of. Lazarus, 1991). Two qualifications are required here. here. First, definitions definitions of "adaptation" differ confusingly. To evolutionary psychologists the term refers to genetically-programmed mechanisms. I prefer Lazarus' 1991) broader usage genetically-programmed Lazarus' ((1991) of the term to refer to any attempt to manage the demands and opportunities of an environment, which leaves open the utility of an evolutionary analysis. I will use "adaptation" subsequently in this broad sense, unless otherwise indicated. indicated. Second, in emotion research especially, it is important to uman characteristic from distinguish explanations for emotion as a hhuman individual individual differences in emotion and associated behaviour. Adaptive explanations at the species level do not necessarily generalise to explanations for individual differences. ummary, processing models are only the beginning of the cognitive In ssummary, science enterprise. enterprise. For further explanation, we may look either towards a reductionist approach of seeking PE effects on the cognitive architecture, which may be supported by neural mechanisms. mechanisms. Alternatively, we may adopt a more systems-orientexl systems-oriented holistie holistic approach of establishing strategy effects, and their role in the person's adaptation adaptation to the physical and social environment. environment. We may also nee~ need to consider how neural systems, processing and evolution. The new evolutionary psychology motivations have been shaped by evolution. provides a different kind of adaptive, knowledge-level explanation to that
G. G. Matthews Matthews
13 13
choice. Next, the prospects for for afforded by motives for personal strategy choice. levels of of explanation are discussed further. developing these complementary levels Towards a Cognitive Cognitive Neuroscience Neuroscience of Personality Personality and Emotion? Emotion? Towards
neuroscience of PE have been dogged by two Investigations of the neuroscience fundamental problems: the use of over-generalised constructs, exemplified by
status of general arousal theory, and nagging doubts about the causal status physiological constructs. Criticisms of arousal theory are familiar. In brief, there are four sources of difficulty (Matthews & Amelang, 1993). 1993). Empirical Empirical criticisms focus on the failure of arousal theory predictions: the supposed relationship between arousal and performance is simply not robust inverted-U relationship 1985; Neiss, 1988). 1988). Methodological Methodological criticisms relate to (Matthews, 1985; weaknesses in inference from empirical data, such as the difficulty in falsifying arousal theory within typical stressor-interaction designs (Hockey, 1984). PsychometriC Psychometric criticisms point to the failure of alternative arousal 1984). measures to intercorrelate, implying that the construct cannot be 1967). Conceptual Conceptual criticisms concern the construct operationalised (Lacey, 1967). validity of "arousal" and "performance", "performance", both of which are multi-faceted 1984; Robbins, 1986). 1986). Hockey's cognitive critique of arousal theory (Hockey, 1984; important: "arousal" effects vary across stressors and processing is especially important: functions, and are often associated with subtle strategic effects rather than functions, changes in parameters of the architecture. architecture. None of these considerations considerations rule out the possibility of a better arousal theory. Such a theory would require the discrimination discrimination of different circuits whose overall activity might influence processing, a description of the specific functions sensitive to each circuit, and satisfactory information-processing functions methods for manipulating and measuring these multiple arousal dimensions independently. Various multi-dimensional multi-dimensional arousal theories have been proposed (e.g. (e.g. Sanders, 1990), 1990), but none have succeeded in explaining more than a small part of the empirical data. A severe barrier to theory development is the sheer complexity and interactivity of neural systems. In the personality context, Zuckerman (1991) points out that there is no one-toone-to one mapping between neural systems and personality traits. He sees each trait as supported by several systems, and, conversely, each system feeds into several traits. Hence, even if neurological neurological reductionism reductionism is correct in principle, principle, it may be difficult to establish in research practice. The other basic criticism of the psychobiological enterprise may be traced back to peripheralist views of emotion and the Jamesian view that
114 4
Chapter 1I
emotion derives from perceptions of of physiological physiological reactions, perhaps through of autonomic nervous system activity (Schachter the appraisal and evaluation of of this approach may be extended by denying & Singer, 11962). 962). The logic of physiological reactions any special status. Emotions may be constructed from appraisal of of a variety of of cues, from the external physical and social environment, as well as from physiological reactions. In contemporary research, this position has 1 984, has been expressed most forcefully by Lazarus ((1984, 11991) 99 1 ) who argues that the influence of of physiology is always shaped by appraisal and cognition. Lazarus ((1991) 199 1 ) does suggest that there may be qualitatively different types of of appraisal, trading off off speed of of processing against depth and complexity, which might be loosely associated with different brain structures. However, explaining how different modes of of appraisal influence emotion is a cognitive- rather than a brain-level question: the distinction is between two different cognitive cognitive modules. In tenns terms of of the current framework, framework, the explanatory questions are how the architecture different types of of appraisal, and how appraisal and emotion are supports different driven by adaptation to the environment. Van Reekum Reehan and Scherer (this that multiple levels of of processing must be distinguished in volume) argue that relating appraisal to emotion and brain mechanisms. Despite the difficulties outlined, there are several promising lines of of research which elucidate mappings between brain and cognitive processes. of neural pathways (e.g. (e.g. Banquet et One approach is the fine-grained analysis of 1 982; leDoux, 995). Gray's ((1982) 1 982) account of al., this volume; Gray, 1982; LeDoux, 11995). of the septo-hippocampal system (SHS) as the basis for anxiety and behavioural of this approach. He explicitly describes inhibition demonstrates the potential of the SHS as performing processing functions, functions, such as calculation of of the mismatch between current sensory events and expectancy. Processing is mapped onto brain circuitry to an impressive impressive degree. degree. However, as with other animal models, fundamental questions concerning the coding of of information are left open. Some system components are clearly non-propositional, non-propositional, such as of the "enabling signal" which gates output from the SHS, and biasing effects of ascending afferents associated with arousal. The system must also make and verify predictions predictions about the world, a process which, in humans, we might imagine to be prepositionally propositionally coded. coded. Rats and people may process information of course, but, in any case, it is difficult information differently, of difficult to develop the theory as a cognitive account of of human emotion when the computational basis for comparator function is uncertain. Predictions from Gray's theory have met with mixed success, in part, because of of difficulties in operationalising its constructs constructs in human subjects (Pickering, Diaz & Gray,
G. Matthews
115 5
has been most powerful when modified through integration with 11995). 995). IItt has (see Derryberry Derryberry & Reed, this volume). human cognitive neuroscience (see solution to the coding coding problem is is the use of of Perhaps the most promising solution al.,. , this volume). It is is emphasised that connectionist models (see Banquet et aI connectionist models do not necessarily correspond to actual neural net
processes: Smolensky ((1988) 1988) describes a variety of important differences between neural functioning and the connectionist architectures typically applied to psychological problems. However, connectionist models do possess some of the key formal properties of nerve cell assemblies. They comprise linked elementary processing units representing analogue information only ("activation"), which is transmitted through associative pathways. There is no direct representation of symbols, which is assumed to be distributed across units, and learning is a direct consequence of the formal properties of the net. Modelling allows testable predictions to be derived concerning behavioural consequence of neural function, as illustrated by Cohen and ServanServan Schreiber's ((1992) 1 992) work on the consequences of abnormality in dopamine function for attention in schizophrenics. However, connectionism is scientifically valuable irrespective of whether activation corresponds directly to neural functions such as rate of firing; typically, activation is best treated (of. J.R. Anderson, 11990). 990). as a formal attribute of the cognitive architecture (cf.
Developing Explanations Developing Adaptive Adaptive Explanations Knowledge level explanations in PE P E research address questions of adaptation. I will take an "adaptive explanation" as a demonstration that expressions of emotion or personality arc are functionally useful in achieving personal goals or dealing with environmental demands. Traditionally, personality theory has been much concerned with the challenges posed by the interplay between basic drives such as sex and power-seeking in an often threatening and unaccommodating world, as expressed in in various psychodynamic theories. Although "adaptive", such explanations are unsatisfactory unsatisfactory because of their failure to specify mechanisms in testable form (e.g. Popper, 1957). 1 957). In contemporary emotion research, Lazarus Lazarus (1991) ( 1 99 1 ) places places adaptation at at the heart of emotion processes" processes: emotions map onto "core relational themes" themes" describing the adaptational adaptational relationship between person and environment. Somewhat similarly, social-cognitive approaches approaches to personality are concerned with the person's person's strivings to implement "personal projects" projects" through interaction with the social environment (Cantor & Zirkel, 1990). 1 990). Such theories theories are testable, testable, by and large, but rarely computational. computational.
16 16
Chapter 1
Contemporary Contemporary research research on on adaptive adaptive models models of of PE PE is is open open to to the the criticism criticism that can such such that it it is is poorly poorly integrated integrated with with information-processing information-processing models. models. How How can an an integration integration be be effected effected in in future future research? research? The The processing processing construct construct bridging bridging architectural architectural and and knowledge knowledge levels levels of of explanation explanation is is strategy. strategy. We We can describe strategies terms of processing constructs selection of can describe strategies in in terms of processing constructs such such as as selection of processing codes, codes, criterion-setting criterion-setting and and so so forth. forth. Understanding Understanding of of strategy strategy use use processing also the person one strategy also requires requires understanding understanding of of why why the person chooses chooses one strategy over over another; another; the the motivational motivational guidance guidance of of strategy strategy choice. choice. Thus, Thus, adaptation adaptation understood cognitively refers understood cognitively refers most most straightforwardly straightforwardly to to the the acquisition, acquisition, selection which selection and and implementation implementation of of computationally-specified computationally-specified strategies strategies which aim within aa given aim to to facilitate facilitate the the person's person's goals goals within given environment. environment. The The primary primary source then experimental source of of data data is is then experimental and and simulation simulation studies studies which which allow allow computional models of of strategy strategy use use to to be be developed. developed. computional models Transient adaptation and strategy selection Explanations for experimental experimental data an understanding understanding of how and and Explanations for data require require an of how
why recent why PE PE factors factors are are related related to to strategy strategy selection. selection. For For example, example, much much recent research on on distressing distressing environmental stressors such such as research environmental stressors as loud loud noise noise suggests suggests that their pronounced than effects on on basic that their effects effects on on strategy strategy are are often often more more pronounced than effects basic structural of the the processing system (Hockey, (Hockey, 11986). Noise structural parameters parameters of processing system 986). Noise appears enhance use of the the dominant for whereas appears to to enhance use of dominant strategy strategy f or performance, performance, whereas fatigue is associated with with aa switch switch to to low-effort low-effort strategies strategies.. Ecological theories is associated Ecological theories fatigue 989) see of of stress stress (Hancock (Hancock & Warm, Warm, 11989) see behaviour behaviour in in performance performance contexts contexts as as driven both by strivings to perform well and strivings to maintain a driven both by strivings to perform well and strivings to maintain a comfortable task task load. load. Negative Negative emotion emotion and and performance performance degradation degradation are are comfortable influence~ by by the the success success or or failure failure of of the the strategies strategies which implement such such influenced which implement motivations (see (see Kluger Kluger & DoNisi, Strategy choice choice under under motivations DeNisi, 11996). 996). Strategy environmental stress stress reflects reflects the the subject's subject's immediate immediate motivations, motivations, beliefs beliefs environmental about the the personal personal significance significance of of the the stressor, stressor, and and beliefs beliefs about about the the efficacy efficacy about of strategy strategy use in meeting meeting salient salient goals. goals. Beliefs Beliefs vary vary dynamically, dynamically, and and use in of perhaps even even on on aa trial-to-trial trial-to-trial basis, basis, as as the the person person modifies modifies strategy strategy in in perhaps response to to error error ffeedback (cf. Rabbitt, Rabbitt, 11979). These "state" "state" variables variables are are response eedback (cf. 979). These influenced by "trait" representations in in L LTM of the the person's person's goals goals and and general general influenced by "trait" representations TM of beliefs relevant relevant to to the the particular particular situation situation (Matthews (Matthows & Wells, Wells, 11996). For beliefs 996). For example, detrimental detrimental after-effects alter-effects of of noise on performance performance may may derive derive from from example, noise on reduced of active active coping strategies, resulting resulting from from appraisals appraisals of of the the reduced use use of coping strategies, stressor/task environment environment as as uncontrollable uncontrollable and and the the limited relevance of of the the stressor/task limited relevance laboratory situation to personal personal goals goals (see (see Cohen, Individual laboratory situation to Cohen, 11980). 980). Individual
G. Matthews
117 7
differences differences in susceptibility to noise may reflect the individual's beliefs about the threat and controllability of 984). of noise stimuli (Jones, 11984).
Stabilities of of adaptation There are different timescales for adaptation (Revelle, 11993). 993). In addition to "single-occasion" "single-occasion" instances of of strategy-driven behaviour, there are stabilities of of adaptation associated with PE evident over periods up to a single stabilities al., this volume). The key question here is the nature life time (see Mayer et aI., of the representation which maintains stability, and there are several options. of Emotion effects on performance may often be somewhat context context-specific, -specific, and contingent upon context-bound appraisals and motivations (Matthews, contingent Sparkes, & Bygrave, 11996). 996). At the same time, data from widely diverse contexts suggests that that emotions such as anxiety and depression may have some cognitive correlates which are intrinsic to the emotional state (Martin & Jones, 11995), 995), or at least prototypical of of the emotion. Oatley and Johnson Johnson1 987) hypothesis that Laird's ((1987) that emotions signal the status of of current action plans implies a degree of of context-independence. Sadness indicates failure of of a major plan (a description of of adaptive status), which in tum turn constrains cognitions and action. As Lazarus ((1991) 199 1 ) states, sadness is associated with appraisals of of irrevocable loss, and an action tendency for withdrawal from the environment, so that a given emotion entails a given representation of adaptive status. The basis of of emotions in adaptive status forces at least some emotion-cognition relationships relationships across individuals and consistency in emotion-cognition of contextual factors. occasions, despite the influence of reflect stabilities stabilities of adaptation. Similarly, personality traits may reflect Dora ((1995) of the diversity of of Matthews and Dom 1 995) present an adaptive account of independent information-processing functions associated with traits such as independent that personality traits represent extraversion and neuroticism. They argue that of environment, defined in terms of of fitnesses for adapting to certain kinds of of traits provide the their informational properties. Cognitive correlates of of the skills necessary for success in the building blocks for acquisition of environments concerned. For example, extraverts are adapted to environments environments characterised by high information information flows, including social environments (see Matthews, this volume). Correlates of extraversion such as high STM capacity, low response criterion and efficient dual-task performance facilitate the development of skills and strategies for handling rapidly-changing inputs. of information-processing alone, extraversion is associated Viewed in terms of with an arbitrary collection of of cognitive correlates. The link between
18 18
Chapter I1
processing and the central characteristics of such as impulsivity and the of extraversion, such and sociability, requires the adaptive perspective. Representation of of the and of adaptive potentials associated with traits is distributed over aa number of distract distinct processing characteristics. for adaptation Genetic basesfor of adaptive The final element element of adaptive explanations is the evolutionary perspective, operating over a time scale of of many lif lifetimes. etimes. At the species level Darwin recognised that emotional responses may be understood at the species species tenns of of their functional properties in aiding survival and level in terms reproduction. Evolutionary psychologists argue that emotions solve the regulatory problems posed by the need to to coordinate multiple processing modules to to handle imperative situations (Tooby & Cosmides, Cosmides, 11992). 992). Instructions for building modules during development are represented within the genes. Modules may then be characterised both computationally eomputationally and in terms of of their design for solving adaptive adaptive problems. Perhaps a more contentious question is how how individual differences differences in genotype are expressed as individual differences differences in module functioning. functioning. Despite the controversial nature of of the research, there is now convincing evidence from structural modelling of of behaviour genetic data to suggest that that major personality traits such as negative emotionality are partially inherited inherited 992), and the beginnings of (Loehlin, 11992), of a molecular genetics of of personality are emerging (e.g. Lesch Leseh et al., 1996). 1996). The thinking of of psychobiological researchers often seems unduly linear: the implicit model seems to be that that the random random outcomes of of the genetic dice feed forward powerfully into personality, with perhaps a little modification by gene-environment interaction. This model leads naturally to the naive good genes/bad genes perspective previously criticised. criticised. It is hardly possible to to estimate the the selection pressures on the various traits. However, even traits which are socially devalued, such such as neuroticism and psychoticism presumably have have adaptive value in some circumstances, circumstances, or the genes coding for them would have been selected out. Matthews and Dom Dora ((1995) 1 995) argue that that neuroticism is adaptive adaptive when the environment is characterised by disguised or subtle threats, especially social threats. Similarly, psychoticism may facilitate creativity (Eysenck, 1995), perhaps mechanisms such as those described by Beech and through attentional mechanisms Williams (this volume). Thus, while genes may feed f orward into the forward cognitive correlates of traits at the level of the the individual, individual, the cognitive cognitive
G. Matthews Matthews
119 9
of traits traits represent represent feedback from the environment over many components of generations. If If aa person is to to function as an extravert, by relying on social generations. interaction interaction
to to
promote promote
survival,
for f or
example,
then
those
cognitive
characteristics supporting social interaction characteristics supporting interaction skills will be selected for. This
process in m tum turn entails selection ffor the neural neural net parameters parameters associated associated with process or the characteristics. The patterning of of cognitive/neural functions the cognitive characteristics. associated with traits traits represents, represents, in part, part, the toolkit of of functions required for 11997). 997). Thus, natural natural selection links the adaptive and biological levels of of Thus, explanation: individual diff differences brain functioning support support individual explanation: erences in brain differences of environment. We We can reconceptualise reconceptualise the ladder of of diff erences in choice of explanation as aa loop, as shown m Figure 2, with connectionist connectionist networks, explanation in Figure and natural natural selection as the key key constructs constructs bridging the levels of of strategies and Adaptation, in the broad broad sense, sense, is not solely solely driven by natural natural explanation. Adaptation, of course. Learned adaptations adaptations may be equally or more important, selection, of although it is uncertain how how much learning influences basic parameters parameters of of neural net functioning. The present account emphasises the importance of of skills rather rather than than processing processing components components in determining adaptation. adaptation. Good SSTM TM for words does not necessarily assist a person person to function as an but being able able to to remember ongoing conversations conversations most likely does. extravert, but must be learnt, aa process process which reflects the the interaction interaction between the Skills must person's choice of for acquiring knowledge (knowledge level) and of strategies strategies for the processing processing routines routines which implement learning (architectural (architectural level). adapting to the the environments associated associated with the the trait trait (Matthews, (Matthews, adapting
Adaptation (environmental (environmental fitness)
I
Strategies Strategies (performance (performance and learning) and/earning) Information processing
---
Knowledge
Natural Natural selection selection
Architecture
Connectionism
~--
Biology
t
Neuroscience
Figure
2.. Levels of explanation explanation reconceptualised reconccptualised as a loop. 2
20 20
Chapter 1 Example: Explaining Anxiety and Cognition An Example:
In In discussing discussing levels levels of of explanation explanation for for PE PE phenomena, phenomena, quite quite aa lot lot of of ground has has been been covered, and and the the scope scope for for confusion confusion and and over-complexity over-complexity in ground explanation will be evident. What What the the researcher researcher must must do, of of course, course, is to to explanation select levels of of explanation explanation appropriate appropriate for for the research research problem at at hand. In the application of of the cognitive science approach to to this section, I illustrate the explaining to be explaining relationships relationships between between anxiety and attention. attention. The phenomena phenomena to of attention, attention, bias of of selective attention explained are well-known: impairment of threat stimuli, and the relationship relationship between abnormality in attentional to threat function and clinical anxiety anxiety disorder (see Wells & Matthews, 1994, 1 994, for for a of well-regarded information-processing information-processing models review). There are are aa variety of area (e.g. Bower, 1981; 1 98 1 ; Ingram, 1984; 1 984; Williams, Watts, MacLeod, Macleod, & in this area Mathews, 1988). press a) address the question of 1 988). Matthews and Wells (in press question of how phenomena how we can can go beyond the the information-processing description of of phenomena to explain associations between emotion and attentional functioning, and their for clinical disorder. implications for Anxiety and in/ ormation-processing information-processing The first step is to to decide what what kind of of explanation explanation is sought. There is a
of anxiety-related bias but but it has proved difficult to to integrate psychobiology of with studies of of selective attention in humans (Wells & Matthews, 1994, 1994, pp. ple, Gray's ((1982) 1982) SHS influences attention to threat 325-332). For For exam example,
stimuli (punishment (punishment cues), but anxiolytic drugs which act on the SHS fail to influence attentional attcntional bias on the emotional Stroop test (Golombok, Stavrou,
& Bonn, 99 1 ) . Thus, while Bonn, 11991). while acknowledging that biological (and evolutionary) evolutionary) factors may be important, the most straightforward approach is to focus on the architectural and knowledge levels. levels. The next next step step is to to characterise the performance
correlates
of anxiety
in processing terms.
The
central central
architectural issue here is the extent extent to which anxiety influences strategic and/or automatic processing. The The distinction distinction between between plan-driven plan-driven strategic strategic control and and stimulus-driven stimulus-driven "automatic" "automatic" control control of of processing processing has has been been developed in in considerable detail detail (Norman (Norman & Shallice; Shallice; 1985). 1985). Anxiety might might influence both the the processing routines implementing implementing strategic strategic or or executive executive control, control, and and parameters parameters of involuntary involuntary processing. processing. Matthews Matthews and and Wells Wells (in (m press press b) b) review review the the evidence evidence on on the the automaticity automaticity
of attentional attentional bias, bias, and and conclude conclude that that bias bias is is predominantly predominantly strategic. strategic. There There is is of considerable or context-sensitivity considerable evidence evidence ffor context-sensitivity of of bias bias (e.g. (e.g. Calvo Calvo & & Castillo, Castillo,
G. Matthews
2211
11997), 997), even with subliminal 996). Similarly, deficits evident on subliminal stimuli (Fox, 11996).
tasks with neutral stimuli, demonstrated in test anxiety research (Sarason (Samson et al., appear to be associated with loss of attentional resources or aI., 11995), 995), appear working memory (Eysenck, 1992), 1992), constructs associated with strategic rather than automatic processing. The clinical literatUre literature too tends to emphasise the strategies that anxiety patients develop for interpreting interpreting and coping with a 985; Wells, world appraised as threatening (Beck, Emery, & Greenberg, 11985; differently, people with anxiety traits have developed "skills" "skills" for 11995). 995). Put differently, handling threat, which are sometimes maladaptive. One effect of state anxiety may be to bias retrieval of the processing routines routines controlling these skills. 1 996) investigated Matthews Matthews and Harley ((1996) investigated the computational basis for attentional bias using a connectionist connectionist simulation of the emotional Stroop. Stroop. The network was trained to discriminate colour and semantic inputs inputs using the backpropagation algorithm. Bias towards negative emotion semantic content was introduced introduced through various mechanisms, mechanisms, and the performance of of the network compared with real data. The most satisfactory mechanism mechanism was a "threat-monitoring" task demand unit strategic one: one: low-level activation of a "threat-monitoring" during colour-naming and word reading. In other words, strategic processes modulate the spread of activation from input to output units. "Automatic" mechanisms, such as sensitivity of input units to negative stimuli, and over overlearning of response to negative stimuli, generated patterns of performance incompatible with real data. data. Siegle Siegle and Ingram (this volume) and Tryon (this volume) discuss alternative connectionist architectures for modelling phenomena relating to negative emotion. Investigation of the underlying architecture through experiment and simulation simulation suggests that bias is more than just an "accidental" over-sensitivity over-sensitivity of automatic threat-processing mechanisms. However, various explanatory questions are left open. open. It is conceivable that the primary consequences of anxiety are architectural, such as loss of resources, and anxiety effects on strategy are an attempt to "work around" these limitations. limitations. Alternatively, Alternatively, anxiety may not affect the architecture at all, but, instead, it influences personal goals and motivations which directly impinge on strategy choice and acquisition of threat-management threat -management skills. skills. Questions Questions also remain about the inter-relationship of the various performance correlates of anxiety, which are sufficiently sufficiently diverse that multiple processing mechanisms are likely to be 992). Diversity in component processes may be involved (Eysenck, 11992). associated with unity at the knowledge level Dom, 11995). level (Matthews & Dorn, 995). For serve example, the various processing characteristics of anxiety may all sub subserve an overall orientation towards hypervigilance (Eysenck, 11992). 992).
22
Chapter 11
A multi-level explanatory model The Wells and Matth ews ((1994, 1994, 11996) 996) Self-Referent Executive Function Matthews
model of of attention attention and negative emotion emotion integrates architectural architectural and (S-REF) model knowledge levels levels of of explanation explanation within a multi-level multi-level framework. Three main components of of the architecture architecture are distinguished: distinguished: stable self-knowledge self-knowledge
LTM form, automatic processing processing encoded in L TM in procedural f orm, stimulus-driven automatic networks, and a supervisory executive. executive. In response to various various internal and or coping external threat stimuli, stimuli, the executive retrieves generic procedures ffor to the specific demands of of the with threat from LTM, LTM, and tailors them on-line to situation. As in most models of 985), of this kind (e.g. (e.g. Norman & Shallice, Shallice, 11985), influence behaviour indirectly, indirectly, though routines under executive control influence of the biasing automatic processing. In the S-REF configuration, operation of characterised by self-focus of of attention, cognitive cognitive interference executive is characterised generated generated by worry, and the pursuit of of self-regulative goals, such as maintaining self-esteem. The S-REF model maintaining self-esteem. model also emphasises the dynamic interplay of of components: components: self-knowledge self-knowledge drives processing processing of of threat stimuli, but is itself itself often often modified by self-appraisal. self-appraisal. Clinical Clinical disorder is generally associated with dynamic disturbances, disturbances, such as perseverative perseverative cycles of of rumination which ffail self-beliefs adaptively adaptively (cf. (of. Siegle Siegle & Ingram, rumination ail to modify self-beliefs this volume; volume; Tryon, this volume). model, architectural architectural and knowledge knowledge levels levels of of understanding understanding Within the model, are linked linked through coping strategies (Matthews & Wells, 11996). 996). The level specifies the personal goals and beliefs about goal attainment attainment knowledge level which influence strategy choice. choice. For example, generalised anxiety patients are motivated to to protect themselves against various (often unrealistic) unrealistic) threats, motivated metaeognitive belief that worry is a successful strategy ffor and they hold the metacognitive or so doing (Wells, 1995). architectural level level delineates delineates the specific 1 995). The architectural processing routines which which implement implement coping. coping. The S-REF model model makes two general statements about processing processing in distress states, consistent with reviewed by Wells and Matthews Matthews (1994). First, processing processing empirical evidence evidence reviewed both the internal activities associated with worry tend to interfere with both operations of of the executive system, such as formulating coping strategies, and operations implementing and regulating the strategies themselves, themselves, if they are with implementing attentionally ddemanding. although there is considerable variability in emanding. Second, Second, although attentionally of coping, distressed individuals often choose the task-focused strategy of for congruent with personal concerns. Threat monitoring monitoring f or threats congruent necessarily fully conscious) is (which is voluntarily initiated but not necessarily responsible for emotional emotional Stroop effects. It remains for future research to to responsible
G. Matthews
23
determine the specific processing routines involved: involved: the Matthews and Harley
((1996) 1 996) simulations illustrate how this might be done computationally.
3 summarises levels of or inter-relationships of explanation ffor between cognition and anxiety (and (and other negative emotions) emotions).. The three of explanation provide alternative ways of of describing anxiety classical levels of phenomena. phenomena. At the knowledge knowledge level, anxiety relates to self-knowledge self-knowledge and goals, as in Beck et al.'s ((1985) 1985) schema theory. There may also be anxiety effects on processing specified at the architectural level (de-emphasised effects within the S-REF model). A full account must accommodate the neuroscience of of anxiety, which is becoming increasingly integrated with architectural volume).. In descriptions (see Derryberry & Reed, this volume; Kitayama, this volume) this section, we have argued that deeper understanding is obtained through use of of constructs which bridge the levels, especially strategies which control of the architecture to serve personal goals, and neural nets which the use of describe processing phenomena using constructs broadly compatible with that anxiety is genetically-influenced, genetically-influenced, we need neurophysiology. To the extent that also to consider how the neural basis of of anxiety has developed through individuals are sensitive to threat stimuli, but often natural selection. Anxious individuals they are conspicuously poor at handling the demands of of threatening Figure
Threat-driven Threat-driven
sself e l f --regulation regulation
Coping strlltegies strategies
� monitoring ",-onitoring
- thre threatt
rumination
Attentional Attentional processes
processes -- resource r e s o u r c e loss loss
�__
-
~-
I Knowledge Knowledge
(
Geneti c adaptBtion Genetic adaptation
Architecture Architecture
etc. - bias, bias. etc.
Neural Neural net net parameters parameters
to to t!nvironmf!nts environments characteri sed by characterised
\ � •
-- e.g. activation activation of of threat threat
monitoring monitoring units units
/
subtle subtle threBts threats
Biological Biological
I
Cortical and subcortical Cortical and subcortical circuits circuits activated a c t i v a t e d by by threat t h r e a t stimuli stimuli
33.. Lcvds Levels of for associations between anxiety and of explanation for cognition. Figure
Chapter 11
24
environments. environments.
Matthews and Dom ((1995) 1 995) argue that the processing
correlates of of anxiety serve the adaptive goal of of maintaining vigilance for
disguised threats (especially social threats), and neural correlates of of subtle or disguised for anxiety may have evolved f or this purpose. Conclusions I have suggested that the multi-f aceted emerging cognitive science of multi-faceted of personality and emotion may be clarifi ed by distinguishing inf ormation clarified information-
explanations of of the phenomena those models models processing models from explanations describe. Sperry ((1993) 1 993) has has claimed that cognitive science introduces a new model model of of causal determinism, combining traditional microdeterminism with the top-down influences of of emergent, macro mental state variables variables.. Consistent
with this view, reductionistic and holistic explanatory strategies may be distinguished ocus on the transient distinguished in PE research. Reductionism requires a ffocus (state) or fixed (trait) diff differences architecture which may be (state) erences in cognitive architecture associated with emotion and personality factors. Architectural erences Architectural diff differences may in tum turn be traced to properties of of neural circuits. For reductionism to be scientifically valid, the mappings between these diff erent levels different levels must be sufficiently simple that novel, novel, testable predictions of behaviour may be derived from theory. Predictions include include those derived derived from connectionist models, which may provide an important bridge between neural and architectural levels levels of of explanation. The range of phenomena open to cognitive neuroscience explanation remains to be determined. seeks to explore the The alternative, holistic approach to explanation seeks adaptive basis of of emotion and personality, in the broad sense proposed by Lazarus. We require an understanding understanding of how state and trait characteristics subserve the goals associated with emotions and personality. That That is, the functional design of the processing system f or implementing and acquiring of for implementing contextualised contextualised skills skills may vary across individuals individuals and across occasions. Cognitive science requires that adaptive explanations explanations are linked linked to computational accounts of phenomena. Over short time-scales, the link of phenomena. link may be achieved through specifying the strategies which allow goals to be met through implementing implementing specific processing routines. Over longer time scales, there are several approaches to explaining stabilities of of adaptation. First, representations TM may drive representations of of generic genetic strategies in L LTM drive consistency in computation. computation. Second, Second, representations representations of of adaptive status may be intrinsic to emotional states. Third, personality traits may be associated with bundles of
25
G. Matthews G. relatively stable,
functionally independent
computational
characteristics
which support successful adaptation to specified environments.
reduetionist and holistic explanations may feed into Finally, both reductionist evolutionary explanations. To the extent that reductionism results in neural accounts of personality and emotion, evolutionary psychology may explain
how the brain systems concerned have been shaped by the pressures of and/or the strategies natural selection. In addition, the person's goals, and/or available for satisfying those goals, may be directly or indirectly related to genetically-programmed mechanisms, It it is likely that the adaptive ammed adaptive mechanisms. genetically-progr characteristics of personality and emotion reflect some complex interplay between social learning and genetics. However, as the example of anxiety choosing levels of of explanation. research shows, it is wise to be selective in choosing cognitive science framework framework are Different levels within the overall cognitive appropriate to different problems in personality and emotion research. References
R 1 982). Acquisition of cognitive skill. R. ((1982). skill. Psychological Review, Review, 89, 369-406. Anderson, 1. J. R R. ((1990). adaptive character oof thought. Hillsdale, NJ: Anderson, 1990). The adaptive f thought.
Anderson, 1. J. Anderson,
Erlbaum. Anderson, K. 1. J. ((1990). 1 990). Arousal and the inverted-U hypothesis: A critique of Neiss's "Reconceptualizing Arousal". Psychological Bulletin, Bulletin, 107, 107, 96100. 1 00. Bandura, A. Englewood Cliffs, NJ: PrenticeA (1977). ( 1 977). Social learning theory. theory. Englewood Prentice Hall. J.. A., Chaiken, S., Govender, Pratto, F. ((1992). A, Chaiken, Govender, R., R, & Pratto, 1 992). The generality Bargh, 1 of the automatic attitude activation effect. Journal of of Personality and Social Psychology, 62, 893-912. 893-9 12. Beck, A. treatment. Philadelphia: Philadelphia: A T. (1967). ( 1 967). Depression: Depression: Causes Causes and treatment. University of Pennsylvania Press. Beck, A T., T. , Emery, G., & Greenberg, R. R L. (1985). ( 1 985). Anxiety disorders and Beck, A. phobias: phobias: A cognitive perspective. New York: Basic Books. Bower, G. G. H. (1981). ( 1 98 1). Mood and memory. American Psychologist, 36, 36, 129148. Calvo, M. M.. D. M . G., & Castillo, M D . (1997). ( 1 997). Mood-congruent bias in interpretation of ambiguity: Strategic processes and temporary of Experimental Psychology, Psychology, 50A, 50A, 1631 63activation. Quarterly Journal of 182. 1 82.
26
Chapter 1I
Personality, cognition, cognition, and purposive Cantor, N., & Zirkel, S. ((1990). 1990). Personality, Pervin (Ed.), (Ed. ), Handbook o fpersonality: Theory of Theory and behavior. In L. A A. Porvin research. New York: Guilford. 1 992). Context, cortex and dopamine: Cohen, 1. J. D., & Servan-Schreiber, Sorvan-Sehreibor, D. ((1992). A connectionist conne~ionist approach to behavior and biology in schizophrenia. Psychological Review, Review, 99, 45-77. Cohen, S. ((1980). 1 980). After effects of stress on human perfo rmance and social performance ,Psychological Bulletin, Bulletin, 88, behavior: A review of research and theory. Psychological 821 08 . 82-108. Cosmides, L 1 992). Cognitive adaptations L.,. , & Tooby, 1J.. ((1992). adaptations for social exchange. In J. 1. H. Barkow, L. Cosmides & 1. J. Tooby (Eds.), (Eds.), The adapted mind: f culture. culture. Oxford: Oxford Evolutionary psychology psychology and the generation oof University Press. 1 992). Four ways five factors are basic. Costa, P. T., Jr., & McCrae, MeCrae, R. R. ((1992). Personality and Individual erences, 113, 3, 653-665. IndivMual Diff Differences, Ellis, A 1 962). Reason and emotion emotion in psychotherapy. New York: York: Lyle A. ((1962). Stuart. Smart. 1. ((I1 995). Creativity as a product of intelligence and personality. Eysenck, H. J. f In D. H. Saklofske & M. Zeidner (Eds.), International handbook o of personality and intelligence. New York: York: Plenum. J., & Eysenck, M. W. ((1985). Eysenck, H. 1., 1 985). Personality and individual differences: A natural science approach. York: Plenum. approach. New York: Anxiety: The cognitive cognitive perspective. Hillsdale, NJ: 1 992). Anxiety: Eysenck, M. W. ((1992). Erlbaum. anxiety: The role Fox, E. ((1996). 1 996). Selective processing of threatening words in anxiety: 10, 449-480. of awareness. Cognition and Emotion, 10, Nature'ss mind: mind: The biological biological roots roots oof thinlang, Gazzaniga, G azzaniga, M. S. ((1994). 1 994). Nature' f thinking, emotions, sexuality, language and intelligence. intelligence. Harmondsworth: emotions, sexuality, language Penguin. Bonn, J. ((1991). diazepam on A, & Bonn, 199 1 ) . The effects of diazepam Golombok, S., Stavrou, A., anxiety-related cognition. cognition. Cognitive Cognitive Therapy Therapy and Research, Research, 15, 459467. Gray, J. 1. A. ((1982). 1 982). The neuropsychology f anxiety: An enqUiry neuropsychology oof enquiry into into the of the septo-hippocampal system. system. Oxford: Oxford: Oxford University functions of Press. Hampson, S. E. ((1988). of personality personality (2nd ed.). ed.). London: 1 988). The construction of Hampson, Routledge. Hancock, P. A., & Warm, J. model of stress and Hancock, 1. S. ((1989), 1 989), A dynamic model 19-537. sustained sustained attention. Human Factors, Factors, 31, 5519-537.
G. Mattnews
z27I
J. ((1984). of attentional attentional state: state: The eff effects of the 1 984). Varieties of ects of Hockey, G. R. 1. environment. In R. Parasuraman & D .), Varieties of of D.. R. Davies (Eds (Eds.), attention. New York: Academic. of adaptation to stress and Hockey, G. R. J. 1. ((1986). 1 986). A state control theory of differences individual diff erences in stress management. In G. R. 1J.. Hockey, A. W. G.. H H.. Coles (Eds (Eds.), K. Gaillard, & M. G .), Energetics and human inf ormation processing. Martmus Nijhoff. information processing. Dordrecht: Martinus Humphreys, M 1 984). Personality, M.. SS.,. , & Revelle, W. ((1984). Personality, motivation motivation and perf ormance: A theory of erences performance: of the relationship relationship between individual diff differences and inf ormation processing. 5 3 - 1 84. information processing. Psychological ReView, Review, 91, 1153-184. of Ingram, R. E. ((1984). 1 984). Toward an information-processing information-processing analysis of depression. Cognitive Therapy and Research, 8, 443-47 8. 443-478. Jones, D. M 1 984). Individual and group diff erences iin n the response to M.. ((1984). differences .), Noise and SOCiety. noise. In D. M. Jones & A. JJ.. Chapman (Eds (Eds.), society. New York: Wiley. Kanfer, Kanf er, R., & Ackerman, P. L. (1989). ( 1 989). Motivation and cognitive abilities: An integrative/aptitude-treatment interaction interaction approach to skill skill acquisition. of Applied Psychology, 74, 657-690. 657-690. Journal of S.. B B.,. , & Loftus, Loflus, 1. J. ((1988). of self-ref self-referent encoding: The Klein, S 1 98 8). The nature of erent encoding: f contributions of of elaborative and organizational processes. Journal oof 1. Personality and and Social Psychology, 55, 5-1 5-11. Kluger, A 1 996). The eff ects of eedback interventions on A.. N., & DeNisi, A. ((1996). effects of ffeedback performance: performance: A historical review, review, a meta-analysis, and a preliminary 1 9, 254-284. f eedback intervention theory. Psychological Bulletin, 1 feedback 119, 254-284. Lacey, 1. 1 967). Somatic response patterning and stress J. II.. ((1967). stress:: Some revisions of of activation theory. In M. H. Appleby & R. Turnbull Tumbull (Eds. (Eds.),), Psychological stress. New York: Appleton-Century-Crofts Appleton-Century-Crofts.. Lazarus, R. S 1 984). On the primacy of S.. ((1984). of cognition. cognition. American Psychologist, Psychologist, 37, 3 7, 1019-1024. 1 0 1 9- 1 024. 1 99 1 ). Emotion and ord: Oxf ord University Lazarus, R. S. ((1991). Lazarus, and adaptation. Oxf Oxford: Oxford University Press. 1 984). Stress, appraisal and and coping. New Lazarus, R. SS.,. , & Folkman, S. ((1984). York: Springer. 1. E. ((1995). 1 995). Emotion: Emotion: Clues from the brain. Annual Review o f LeDoux, J. of Psychology, 46, 209-235 209-235.. Psychology, Lesch, K. -P. -P.,, Bengel, D D.,. , Heils, A., Sabol, S. Z., Greenberg, B. D D.,. , Petri, SS.,. , J., Muller, Miiller, C C.. R R.,. , Hamer, D. H. & Murphy, D D.. L. ((1996). Benjamin, 1., 1 996). Association of of anxiety-related traits with a polymorphism in the 527- 1 53 1 . serotonin serotonin transporter gene regulatory region. region. Science, 274, 11527-1531.
28 Loehlin, J. C. (1992). (1992). Genes Loehlin,
Chapter 11
and environment in personality development. Newbury Park, CA: Sage. Sage. Martin, M., & Jones, 1 995). Integral Jones, G. V. ((1995). Integral bias in the cognitive cognitive processing f Psychology, 86, 44191 9of pictures. British Journal o of of emotionally emotionally linked linked pictures. 436. Matthews, G. ((1985). 1 985). The effects effects of of extraversion and arousal on intelligence intelligence test perfo nnance. British Journal of performance. of Psychology, 76, 479-493. Matthews, G. (1992). ( 1 992). Extraversion. Extraversion. In A. P. Smith Smith & D. M. Jones Jones (Eds.), Handbook of human per f ormance. Vol. 3: State and trait. of performance. Vol. trait. London: London: Academic. Matthews, Matthews, G. (1997). (1997). Intelligence, Intelligence, personality personality and information-processing: information-processing: Kingsma (Eds.), (Eds.), Advances An adaptive perspective. perspective. In W. Tomic Tomic & J. Kingsma in cognition (Vol. 4), pp. 475-492. 475-492. cognition and educational practice (Vol. Greenwich, Greenwich, CT: JAI Press. Matthews, 1 993). Extraversion, Matthews, G., & Amelang, Amelang, M. ((1993). Extraversion, arousal theory and nnance: A study of individual perfo performance: individual differences differences in the EEG. EEG. Personality and erences, 114, 4, 347-364. and Individual Indi~dual Diff Differences, Matthews, Matthews, G., & Deary, Deary, I. 1. J. (in press). press). Personality traits. traits. Cambridge: Cambridge: Cambridge University University Press. Matthews, G., & Dorn, Dorn, L. ((1995). Cognitive and attentional attentional processes processes in 1995). Cognitive Saklofske & M. Zeidner Zeidner (Eds (Eds.), personality and intelligence. intelligence. In D. H. Saklofske .), International handbook of of personality and intelligence. New York: York: Plenum. Effects of extraversion and self-report 1 993). Effects Matthews, G., & Harley, T. A. ((1993). semantic priming: priming: A connectionist connectionist approach. approach. Journal of of arousal on semantic Personality and Social Psychology, 65, 735-756. Matthews, G., & Harley, Connectionist models models of emotional Harley, T. A. ((1996). 1 996). Connectionist distress and attentional bias. bias. Cognition and Emotion, 10, 561-600. Matthews, G., Sparkes, Sparkes, T. 1., J., & Bygrave, Bygrave, H. M. ((1996). Stress, attentional 1996). Stress, overload and simulated formance, 9, simulated driving driving performance. performance. Human Per Performance, 77101. 77-101. Wells, A A.. ((1996). Attentional processes, processes, coping coping strategies Matthews, G., & Wells, 1996). Attentional intervention. In M. Zeidner Zeidner & N. SS.. Endler Endler (Eds.), (Eds.), and clinical clinical intervention. Handbook of of coping: Theory, Theory, research, research, applications. applications. New York: York: Wiley. Wells, A. (in press a). The cognitive cognitive science science of of attention Matthews, G., & Wells, and emotion. emotion. In T. Dalgleish Dalgleish & M. Power (Eds.), (Eds.), Handbook of of cognition and and emotion. New York: York: Wiley. Wiley.
G. Matthews Matthews G.
29 29
Matthews, G., & Wells, A. A. (in press b). b). Attention, Attention, automaticity automaticity and and aff affective ective disorder. disorder.
Behavior Modification. Modification.
Reconceptualizing arousal: arousal: Psychobiological Psychobiological states states In in ((1988). 1988). Reconceptualizing Bullean, 1103, 345-366. motor perfo performance. motor rmance. Psychological Bulletin, 03, 345-366. Neisser, U. U. ((1976). San Francisco: Francisco: Freeman. Freeman. Neisser, 1976). Cognition and reality. San Norman, D. D. A., A., & & Shallice, Shallice, T. T. ((1985). Attention to to action: action: Willed Willed and and 1 985). Attention Norman,
Neiss, R. R. Neiss,
of behaviour. In In R. R. 1J.. Davidson, Davidson, G. G. EE.. Schwartz & &D D.. automatic control of Shapiro (Eds.), (Eds.), Consciousness and self-regulation: Advances in Shapiro (Vol. 4). New York: Plenum. research (Vol. K.,, & Johnson-Laird, P. ((1987). of Oatley, K 1987). Towards a cognitive theory of emotions. Cognition and Emotion, 11,, 29-50. K. ((1957). of Popper, K 1 957). The poverty o f historicism. London: Routledge & Kegan Paul. D.,. , Diaz, A., & Gray, 1. J. A. ((1995). Personal@ and Pickering, A. D 1 995). Personality reinforcement: reinf orcement: An exploration using a maze-learning task. Personality 541-558. and Individual Differences, 18, 54 1 -558. S.. ((1994). Pinker, S 1 994). The language instinct. Harmondsworth: Penguin. foundation Pylyshyn, Z. W. ((1984). 1 984). Computation and cognition: Toward a f oundation for f or cognitive science. Cambridge, MA: MIT Press. 1 979). Current paradigms and models in human Rabbitt, P. M. A. ((1979). information D.. M. Warburton Warburton (Eds (Eds.), .), inf ormation processing. In V. Hamilton & D and cognition: An inf information Human stress and ormation processing approach. London: Wiley. Revelle, W. (1993). motivation: ( 1 993). Individual differences differences in personality and motivation: "Non-cognitive" "Non-cognitive" determinants of of cognitive performance. performance. In A. Baddeley & L. Weiskrantz (Eds.), Attention: Selection, Selection, awareness and and control control. Oxford: Oxford Oxford University Press. Robbins, and neurobiological aspects ( 1986). Psychopharmacological and Robbins, T. W. (1986). of of the energetics of of information information processing. In G. R. J. J . Hockey, A. W. W. K. K Gaillard, & M. G. H. Coles (Eds.), Energetics Energetics and and human information Dordrecht: Martinus Nijhoff. Nijhoff. information processing. Dordrecht: Sanders, A. F. F . (1990). ( 1 990). Issues and trends in the debate on discrete versus continuous continuous processing processing of of information. information. Acta Acta Psychologica, Psychologica, 74, 123-167. 1 23-167. Sarason, Sarason, I. I . G., Sarason, Sarason, B. B . R., & Pierce, Pierce, G. R. (1995). ( 1 995). Cognitive interference: interference: At At the inteUigenee-personality intelligence-personality crossroads. crossroads. In D. H. H. Saklofske, Saklofske, D. D . H., H . , & M. M . Zeidner, M. M . (Eds.), (Eds.), International International handbook handbook of of personality personality and and intelligence. intelligence. New New York: York: Plenum. Plenum. Schachter, & Singer, Singer, J. 1. E. E. (1962). ( 1 962). Cognitive, Cognitive, social, and and physiological Schachter, S., S., & determinants determinants of of emotional state. state. Psychological Psychological Review, ReView, 69, 379-399. 379-399.
30
Chapter 1I
Individual differences in cognition. Journal of of Research Schwartz, S. (1975). ( 1975). Individual in Personality, 9, 2217-225. 17-225 . Shallice, T. ((1988). 1 988). From neuropsychology to mental structure. structure. Cambridge: Cambridge University Press. 1 967). Motivational and emotional Simon, Simon, H. A A. ((1967). emotional controls of cognition. cognition. Psychological Review, Review, 74, 29-39. Smolensky, P. (1988). (1988). On the proper treatment of of connectionism. eonnectionism. Behavioral and Brain Sciences, Sciences, 111, 1, 1-74. 1-74. ( 1993). The impact and promise of Sperry, R. W. (1993). of the cognitive cognitive revolution. American Psychologist, 48, 878-885. of anxiety on complex complex learning and Spielberger, C. D. ((1966). 1966). The effects of academic achievement. achievement. In C. D. Spielberger (Ed.), Anxiety and behavior. London: London: Academic Press. Tooby, J., 1., & Cosmides, 1992). The psychological foundations Cosmides, L. ((1992). foundations of of culture. In J. 1. H. Barkow, L. Cosmides 1. Tooby (Eds.), Cosmides & J. (Eds.), The adapted mind: Evolutionary psychology psychology and the generation oof f culture. culture. Oxford: Oxford: Oxford University Press. A. ((1992). Watson, D., & Clark, L. A 1992). On traits and temperament: General and specific factors of emotional experience experience and their relation to the five of emotional five1 -476. factor model. Journal of of Personality, 60, 44 441-476. Wells, A 1 995). Meta-cognition A. ((1995). Meta-eognition and worry: A cognitive cognitive model model of of generalised anxiety disorder. disorder. Behavioural and Cognitive Psychotherapy, 23, 30 1-320. 301-320. Wells, A, 1994). Attention and emotion: A., & Matthews, G. ((1994). emotion: A clinical perspective. Hove: Hove: Erlbaum. Edbaum. A, & Matthews, G. ((1996). 1996). Modelling Wells, A., Modelling cognition cognition in emotional emotional disorder: The S-REF model. Behaviour Research and Therapy, Therapy, 34, 888181888. Williams, J. M. G., Watts, F. N., Macleod, MacLeod, C., & Mathews, A A. (1988). Cognitive Cogniave psychology psychology and emotional disorders. Chichester: Chiehester: Wiley. Zuckerman, M. (1991). Psychobiology o f personality. Cambridge: of Cambridge: Cambridge University Press.
Perspectives on Personality Personality and Emotion Emotion -G. Matthews (Editor) (Editor) Cognitive Science Perspectives G. Matthews -
9 1997 1997 Elsevier Science Science B B.V. fights reserved. reserved. © .V. All rights CHAPTER 2 CHAPTER
in Personality Conation, Affect, and Cognition in M. Carlsmith John D. Mayer, Heather Frasier Chabot and Kevin M
During much of the 20th Century, personality psychology has been a field divided into competing schools of psychodynamic, trait, humanistic, and other perspectives, with little communication among perspectives, and no common language. Recently, however, a consensus view of the field has been developing which considers personality from a systems perspective and attends to (a) the location of of personality, (b) its parts, (c) its organization, b, Pcrvin, and (d) its development (Mayer, 11993; 993; 11995a, 995a,b; Pervin, 11980; 980; Sears, 11960). 960). For instance, personality's pcrsonality's location is defined in relation to such neighboring For systems as biology and sociology. arc relatively basic such as Personality's parts include components that are hunger, happiness, and working memory, and more complex components as self and the ego. ego. Thousands Thousands of of parts of of well, including extraversion, the self, of these thousands, at personality have been proposed (Allport, 1958), and of least 400 parts are regularly discussed (Mayer, 11995b). 995b). Keeping 400 parts of personality in mind is a near impossibility, so one alternative strategy is to groups or classes (e.g., Barratt, 1985, consider them in groups 1985; Buss & Finn, 1987; 1 987; Mayer, 1995a,b). I 995a,b). Most classification systems for these components employ one or more of of three categories of of mind that that have a centuries-old tradition: the conaave, cogmtive - what Hilgard ((1980) conative, affective, affective, and cognitive 1 980) has referred to as the -
trilogy of ofmind. According to this division, c~nation conation (or motivation) includes components for that propel or move the organism such as the hunger drive, and the need for achievement. The The affect group, group, principally containing emotion, includes such basic feelings as anger related parts parts such as the anger and happiness, along with related programs for emotional facial faCial expressmns. expressions. The cognition group, mental programs containing thought-related processes and and mechanisms, includes such elements as worlang working memory, judgment, judgment, and reasoning. The The division of of the mind into rconation, affect, and and cognition is so embedded embedded in our our discipline that that many of of our our journals journals are named a~er after those parts: Cogmtion, Cognition, Motivation Motivation and and Emotion, Cogmtion Cognition and and Emotion, and so of us would would be hard-pressed to to recall the origin of of this on. Despite this, many of classification system, or to to describe the differences among the three three
32
Chapter 2
categories. Along these lines, Henle (cited in Hilgard, 11980, 980, p. 1115) 15) remarked:
absorbed in our own specialties we often become as we become absorbed cryptosystematists, that that is, our beliefs are embedded in larger systems of of thought that are not explicit but may serve to perpetuate errors. Indeed, the differences among motivation, affect, and cognition can become paper thin. A person's associations to the word "success" may reveal her need for achievement (conative), while also being influenced by her mood (affect), and memory (cognition). To accommodate such blended areas of of perfo rmance, there exist blended areas of performance, of study such as "cognition and affect", and "motivation and emotion." Still, in what sense is one such class of of mental process to be distinguished from the others? the meaning of of this tripartite tripartite division. We will In this chapter we clarify the begin by examining a general systems model of of personality (already introduced at the outset). This model's further development relies in part on the distinction among classes of of conative, affective, and cognitive components. The systems model illustrates how the three spheres of of conation, affect, and cognition, can be used to classify aspects of of personality psychology. The usefulness of of the three spheres, however, relies on a clear understanding of of each one's meaning. Following description description of of the systems model, we focus on conation, affect, and cognition, including (a) their historical origins, (b) their changing description description across time, (c) their conceptualization, and (d) a recommended update of of their meaning. Finally, we return return to questions of of conation, affect, and cognition in personality and in contemporary research, and discuss how the trilogy may be integrated into a picture picture of of the person as a whole.
Relational Model of Personality The Relational Several contemporary models of of personality employ one or more classes of of conation, conation, affect, and cognition in their their construction (e.g., Barratt, Barratt, 1985; Buss & Finn, 1987; Mayer, I1995a,b). 995a,b). Examination of of one such model demonstrates one way the trilogy of of mind is used today, and highlights some of of the issues surrounding its use. The specific model employed here is the relational model model of of personality, so-called because personality and its parts are all described in relation to one another and their neighboring
e /
I
-
'
I
PERSONALITY
INTERNAL PERSONALITY
NERVOUS SYSTEM
EXTERNAL SITUATION
SITUATIONAL ELEMENTS
[
,I.D. Mayer, H. Frasier Chabot and K.M. Carlsmith
L
.
GROUPS INCLUDING OR INTERACTING WITH
Figure Figure 11 . An An view view of of the the personality system amidst personality system amidst its its neighboring neighboring systems, systems, including includmg biology, and sociology, biology, sociology, and situations. situations. A A molecular-molar molecular-molar IS represented dimension dunension is represented vertically, vertically, an an internal-external internal-external dimension dimension horizontally, horizontally, and and an an organismic organismic dependent-construct dependent-constructed ed dimension dunension depthwise. depthwise.
33
34 34
Chapter 2
systems systems (Mayer, (Mayer, 1995a,b). 1 995a,b). This This relational relational model is typically typically developed developed according according to to four four systems-oriented systems-oriented topics: topics: that that is, according according to to personality's personality's location, components, components, organization, organization, and and development. One aspect aspect of of the the location, relational model model that that makes makes itit particularly particularly worth worth discussing is its highly highly relational integrative developed by integrative aspects; aspects; it contains contains or or subsumes subsumes several models developed by & Finn, 1987). others (e.g., Buss & others 1 987). Certain conceptual conceptual dimensions can can be employed to distinguish Certain personality from fields of of scientific important personality from its neighboring fields scientific study. The The most most important of these these include aa molecular-molar molecular-molar dimension, that of that distinguishes more molecular brain sciences which which underlie personality from personality itself, molecular and also distinguishes distinguishes personality personality from structures that that and also from more molar molar social social structures "contain" it such such as as the family family and and society. A second, internal-external internal-externaI "contain" mental processes from outside observable dimension, distinguishes inside mental behavior. To behavior. To this, a third, organismic-constructed organismic-constructed dimension can be added, which distinguishes between those of personality personality that are most those parts parts of constrained by by the biological organism (i.e. basic motivations) from those that of three that are most independent (i.e., formal reasoning). The use of dimensions makes possible of possible aa three dimensional pictorial representation of I). personality parts (see Figure 1). personality and its component component parts The purpose of this initial picture is to orient personality amidst its purpose of neighboring system in the three-dimensional space. Internal personality is contained contained in aa box labelled "personality" on the the left-hand side side of of the figure, beneath it, and family and social mid-way between nervous system substrates beneath systems above it. In the picture, this vertical dimension represents the molecular-molar molecular-mOlar continuum in the sense that the lower brain sciences are more molecular than personality whereas the family and other social groups above personality are more molar. The second, horizontal dimension, represents the internal-external continuum with internal personality to the left, external manifestation (i.e., its interaction with the and personality's external environment) to the right. Finally, the third, depth dimension, distinguishes more organismic parts of of personality (to be added momentarily) in the foreground from more constructed parts (also to be added) in the background. The empty personality box can now be filled with classes of personality components in a manner that is consistent with each of the three dimensions. For example, in Figure 2, conation, affect, and cognition are placed along the floor of of the cube, near the biological level, with a slight rise toward the back greater molarity of of cognition relative to conation. This indicating the greater particular particular placement implies that conation, affect, and cognition refer to
�
J.D. Mayer, H. Frasier Chabot and K.M. Carlsmith
GROUPSlNCLUDlNG OR INTERACTING WITH PERSONALITY
Figure Figure 22.. A second second view view of the personality system including personality system including the enablers: enablers: Conation, Conation, affect, affect, cogmtlon, cognition, and and consciousness consciousness (modified (modified from from Mayer, Mayer, 1995a, 1995a, Figure Figure 2). 2).
�
�
�
."'1
;:x::
� � ;;; . "'I
� C
-
�
�
� g
"'I
1:;"' :s
§:
35
36
Chapter 2
internal, more molecular components of of mind - that is, close to the biological level, or, only minimally influenced by learning. Notice that toward the innermost
part part
of of personality
a a
fourth
category
has
been
added,
consciousness. ect, and consciousness. The placement of of consciousness near conation, aff affect, cognition suggests that consciousness, like them, is a more molecular,
Too little is known about consciousness to place it definitively anywhere, of of course. One very respectable and influential tradition views consciousness as analogous to an image imago in aa hologram, in that it emerges from layered information within the cerebral cortex (Pribram, 11971, 97 1 , p. 1171). 71 ). This view would place consciousness at at the ceiling of of the personality box. But the relational rdational model puts it to the left bottom f or reasons to be developed later. for Within this relational model, conation, aff ect, cognition, and affect, consciousness are arc subgroups of of a class of of personality components containing Enablers are mechanisms that carry out, or enable, them, termed enablers. enablers. Enablcrs the basic functions of are our broad of personality. The enablers r arc one of of ffour classifications that collectively contain all the parts of of personality. The other three classes are establishments, themes, and agencies. arc establishments, Establishments are arc so-called because they are established (or learned, leamod, or the self, the the world, and and the self in the world. Examples constructed) models of of the of establishments include the the self self-conc~t, self-esteem, of -concept, self -esteem, attachment patterns, and expert knowledge. knowledge. Establishments develop from experience and learning, and and utilize the enablers' cnablcrs' functions to operate. For example, the self concept's self -love or self-hatred will be generated and interpreted self-love intcrprcteA by emotional enablers; enablers, its self assessment will require cognitive enablers. cnablers. The connection cnablcrs and establishments is often limited, however, to the fact that between enablers cnablers support establishments. At the establishment level, for example, enablers expert knowledge can can be fairly independent of of a good or bad memory at at the That is, children may construct expert expert knowledge about about enabler level. That of whether they possess an impoverished or superior dinosaurs independent of can be defined dcfine~ primarily according to its memory. Thus, the establishment can arc illustrated in Figure 3, as the three specific content. Establishment models are of internal personality. They are more molar than than the enablers, enablcrs, floating cubes of and are arc more independent of of the organism as they proceed back toward and viewed as models of of the world. Note that all parts of of personality are arc vicwod connected to all others; no arrows or connections are drawn in, however, as a thicket of of connections would obscure the rest of of the depiction. depiction. such a biological phenomenon, phenomenon, which may interact with the other three.
,I.D. Mayer, H. Frasier Chabot and K.M. Carlsmith
N E R V O U S S Y S T E M
37
Figure Figure 3. 3. A more more complete complete view of the personality system system now including including all all four major classes classes of personality components. components. The The enablers (conation, affect, enablers (wnation, affect, cognition, cogmtion, and consciousness) consciousness) are are on the the floor floor of the The box. personality box. establishments establishments (models (models of the self, self, world, world, and self-in-world) are are represented as as boxes floating floating in the inside inside of the cube. cube. The The in themes themes combine combine features features of enablers enablers and establishments; establishments; one one theme, extroversion, IS theme, extroversion, is illustrated illustrated toward the the back center center of the Figure. Figure. Finally, Finally, agencies agencies are are larger super superof individual composites composites individual components components that collectively act as as sub-personalities; sub-personalities; one one such such agency, agency, James' James' self-as-knower, self-as-knower, is is represented, represented, as as a cloud cloud that intersects intersects with the "Models "Models of the Self' Self' box (modified (modified from from Mayer, 995a, Figure Mayer, 11995a, Figure 2). 2).
3388
Chapter 2
of components, the themes, represent thematic The third class of connections between establishments and enablers. Themes combine features from enablers and from establishments so as to form conceptually related mixtures that reveal themselves to observers in a coherent fashion. Whereas establishments are focussed on contents, themes are focussed on common or integrated features across enablers, across establishments, or across the two combined. Thus, a need for stimulation alone is an enabler; a model of of "joining friends for a party," is an establishment. But the two can be viewed as thematically related. Thus, extroversion, according to Eysenck ((1982), 1982), involves both a need for stimulation, (the conative eonative enabler), and establishment models of of things such as how to throw a party. Extroversion is illustrated as elliptical features found in both conation eonation and in models of of the world; these features are labelled ("extroversion features") to the right of of the internal personality cube. The fourth class of of components, the agencies, refer to large subdivisions of personality that carry out much of of a personality's activities, but in partial of of the whole; these include the id, ego, or superego. Another independence of example of -as-knower, which comes of an agency is James' concept of of the self self-as-knower, close to a self-conscious free spirit or free will. The self-as-knower is of the Self. represented as a cloudlike column that runs through the Models of A more comprehensive discussion of the classification of of of personality components into enablers, establishments, themes, and agencies, and their components twenty-one subcategories can be found elsewhere (Mayer, 11995a,b). 995a,b). Here, we are particularly interested in conation, affect, cognition, and consciousness, the subgroups subgroups of of enablers. Enablers, as already noted, are viewed as close to the biological level in the relational model. For that reason, there must be plausible biological bases for the operation of of these parts, and their division. parts form a larger class that describe mechanisms that carry Moreover, these parts of personality. Hence, the enablers must be divided and out the functions of understood foremost according to what they enable, that is, what functions of they perform. Because enablers are so basic, and perform basic functions of personality, almost all other parts of of personality rely on them and are influenced by them. Better defining conation, affect, and cognition, and understanding the rationale underlying these concepts, can clarify understanding of of personality as a whole.
J.D. J D.
H. Frasier Chabot and K.M K.M. Carlsmith Mayer, H.
39
Understanding Conation, Conation, Affect, Affect, and and Cognition Cognition Understanding affect, recent history Conation, aff ect, and cognition through recent Mind..... recounts the rise Hilgard's (1980) classic article The Trilogy oof fMind 1700's to early 1900's, 1900's, and and fall of these three concepts from the early 1700's resurrection. Surprisingly, offers a rationale and recommendation for their resurrection. Hilgard's work omits virtually any discussion of the meanings of conation, affect, or cognition, aside from their special status as a three-fold classification for the overall mind. Nonetheless, his article provides a basis for such an exploration by tracing the major figures who developed the trilogy over its history. .
and the trilogy oof Faculty psychology and fmind Hilgard (1980, p. 108) 108) starts with the German faculty psychologists of the 18th 18th century. He credits, in particular, particular, Moses Mendelssohn's Letters on Sensation for bringing together .the the three concepts for the first time. Mendelssohn distinguished conation, affect, and cognition according to the fact that they operated differently from one another and that they might even another. For example, when reason (cognition) "laboriously interfere with one another. investigates the origin of pleasure," destroyed" investigates pleasure," he wrote, "pleasure may be destroyed" (Mendelssohn, 1755/1971, p. 66) 1. There is both a phenomenological quality (Mendelssohn, 1755/1971, I . There to this conscious experience of this statement, indicating a sensitivity to the inner conscious cognition and affect, and also a functional notion, identifying that cognition "investigates" pleasure. "investigates" pleasure. Mendelssohn also noted the independent behavior of the three components, writing that "convictions...belong that "convictions ...belong in the realm of man's cognitive psychology," psychology," and that that "by their very nature, [convictions] cannot be influenced by coercion or bribe" 1983/1969, p. 44). On the bribe" (Mendelssohn, 1983/1969, other hand, will or motivation motivation could be encouraged encouraged or discouraged by "reward 1983/1969, p. "reward and punishment" punishment" (Mendelssohn, (Mendelssohn, 1983/1969, p. 44). Mendelssohn's Mendelssohn's approach specifying the approach is aa partly partly functional functional one in the sense that he is specifying the conditions spheres can conditions under under which which operations operations of of the the three three spheres can be teased teased apart. apart. The 18th century Germany gradually The faculty faculty psychology psychology of late 18th gradually spread 1I Mendelssohn's Hans G. Hirsch was kind enough Mendelssohn's work is not yet translated in English. English. Hans enough to translate fragments of the work which at least suggest some flavor of the original writings fragments suggest (see also Mayer, 1995b). 1 995b).
40
Chapter 2
to England and 9th century. A number of and Scotland in the early 119th of psychologists contributed to classifying aspects of of the mind during this period. For example, Thomas Thomas Reid, the great great Scottish faculty psychologist, divided mental faculties into the intellectual (cognitive) and the active (motivational), dropping out emotion. By the late 119th 9th century a summary of of these British works works was published in Alexander Bain's two-volume English textbook on critical of of attempts such as Reid's to reduce the psychology. Bain was fairly critical trilogy to only two categories. He wrote that Reid's "submerged department of Emotion," could not be made to disappear but rather that its parts, such as of emotions, feelings, and so on, "will be found partly taken in among the Intellectual Powers . . . and partly treated among the Active Powers," (Bain, Powers...and 11855/1977, 855/1977, pp. 6-7), where they did not plainly fit. "Mind," wrote Bain ((1855/1977, 1 855/1977, p. 11)) at the outset outset of of The Senses and the Intellect, ....possesses . . possesses three three attributes attributes or capacities.
I. It has Feeling, in which tenn term I include what is commonly called Sensation and Emotion. II. It can can Act Act according to Feeling. III. It can IIl. can Think. Bain's trilogy, however, differs from the contemporary. For Bain, sensation, whereas today's mental divisions typically group Feeling included sensation, sensation with perception, outside the trilogy. Additional information concerning Bain's views on each member of of the trilogy appear in the top of Table 11,, which has three columns. Table 11 indicates the views of of portion of columns several central theorists, beginning with Bain. The three col umns are divided so as to represent that that theorist's view of of conation, affect, and cognition. For Table ll's's affect example, in Table affect column, Bain says feeling and consciousness are "one and the same;" a statement which appears untenable today given research on unfelt, unexpressed, or unconscious emotions (e.g., contemporary research At the same time, Bain successfully develops a contemporary Taylor, 11984). 984). At understanding of of conation as he distinguishes between mental actions, which part of of the mental sphere, and those external actions that that are not (Table I1,, are part under "conation") "conation")..
Chapter 2
411 4
I . Historical and Contemporary Views of Conation, Affect, and Table 1. Quotations and Brief Summaries from Key Figures. Cognition: Direct Quotations
.
.
Conation .
.
.
.
.
.
.
.
.
.
.
.
.
Emotion .
.
.
.
.
.
Cognition
.
BAIN ((1855/1977) 1 8551 1 977)
"Action is is...The "Action . . .The putting
of power to execute forth of
terms, "The three tenns,
""...discriminating . . . discriminating with
Feeling, Emotion, and
preference, and the perfonnance performance of of
perfonn some some work or perform
Consciousness, will, I think think:
operation . . . In speaking of operation...in of
be found in reality to
intermediate actions to intennediate
Action, however, as a characteristic of mind, we must render explicit the distinction between between mental actions and such as are not mental...mental mental . . .mental actions [are]... the prompting {are) . . . under the and guidance of Feeling." Feeling. " (pp 2-3) ""...There . . .There are in the human system movements and tendencies to movement .9 . . The eyes may open of ..The themselves, the voice may break forth into utterance ....Yet . . Yet those movements belong to the sphere of of Volition The tenn term Volition mind. The applies . . . to the entire range applies...to of of mental or feelingfeelingprompted actions ." ." (p. 5)
the same express one and the
attain an end, are the most
mind..." ... fact or attribute of mind
"
(p. 11)) ""...for . . . for a notion of what refer each feeling is, I must refer person to their own experience. The warmth ffelt elt fragrance iin n sunshine, the f ragrance of of flowers, the sweetness of honey..." honey . . . " (p. 2)
universal aspects of intelligence, inasmuch as they pervade the whole of of the animal kingdom." (p. 6) ""...the . . . the intellect {is) . . .a [is]...a distinct endowment following laws of its own, being sometimes well developed and sometimes feeble without regard to the force or degree of the other two attributes. attributes."" (p. 6)
Intellect is distinct from emotion and volition because it allows for sensations and ideas to be relived without the stimulus (pp. 3 1 5-3 16) 315-316) "Reason "Reason without affect would be impotent, affect without reason would be blind." (p. 1112) 1 2)
MACLEAN ((1990) 1 990) "The protoreptilian
"In the evolutionary
"The neocortex [can [can be
fonnation formation is represented by
transition f rom reptiles to from
described as) . . . ballooning as]...ballooning
a particular particular group of
ammals, three cardinal m mammals,
out progressively in
ganglionic structures
behavioral developments
evolution and reaching its
located at the base of of the
were ((1) 1 ) nursing in
greatest proportions in the
42 42
Chapter 2
Table 11 continued. continued. Table forebrain in in reptiles, reptiles, birds, birds, forebrain and mammals...these mammals. . .these and ganglia must must be be of of ganglia significance' for 'enormous significance'for otherwise they they would would not not be be otherwise found as as aa constant constant feature feature found in the the vertebrate vertebrate in forebrain. . . [It is involved in] forebrain...[It such basic behavior as the the struggle for power, for adherence to routine, adherence 'imitation,' obeisance to deception. " precedent, and deception." 1 5-16) (pp. 15-16)
with maternal maternal conjunction with care, care, (2)audiovocal (2) audiovocal commcomm unication for maintaining maintaining unication matemal-offspring contact, contact, maternal-offspring and (3) (3) play...The play...The limbic limbic and basic role in in system plays aa basic functions thymogenic functions reflected as emotional reflected behavior...Two behavior...Two evolutionarily older subdivisions. . .have proved subdivisions...have be involved, respectively, respectively, to be in oral and genital functions...The third functions...The subdivision, for which there appears to be no
human human brain...[it] brain. . . [it] has has afforded afforded aa progressive capacity capacity for problem solving, learning, learning, and and memory of of details... details. . . linguistic linguistic translation translation and and communication of of subjective SUbjective states..." states... " (p. 17) 1 7)
counterpart in rudimentary counterpart reptiles ... [involves] reptiles... [involves] parental parental care, care,audiovocal audiovocal communication, and play behavior" 6-17) behavior" (pp. (pp. 116-17) TOMKINS 1 962) TOMKINS ((1962) "In the human being the drive system plays a central role in ... self -maintenance in... self-maintenance and reproduction. reproduction."n (p. (p. 29) The The system's system's primary function is to provide "motivating information" "information "information that drives and aa drive orms" drive that that inf informs"specific specific to to survival. survival. (pp. (pp. 3030331) 1) It It communicates n"...where where and and when when to do what what-- when the the body body does does not not know know otherwise otherwise how how to to help help itself." itself." (p. (p. 3311 )) •••
"The aff ective system affective [possesses] ...numerous [possesses]...numerous invariant instigators of any particular affect... [and] numerous invariant invariant reducers of the same affect...It is this differentiated coupling and uncoupling characteristic characteristic which permits permits the affect system system to to assume assume aa central central position position in in the the motivation motivation of of man." man." (p. (p. 23) 23) "Affects "Affects are are sets sets of of muscle muscle and and glandular responses responses located located in in the the face face and and also also widely widely distributed through through the the
[Not compared]
J D. Mayer, H. J.D. H. Frasier Chabot Chabot and and K.M K.M. Carlsmith Carlsmith
Table 11 continued. continued. Table "The drive drive system system with with its its "The
body, which which generate generate body,
relatively primitive primitive signal relatively
sensory ffeedback which is sensory eedback which
mechanisms and ffeedback eedback mechanisms
either inherently inherently 'acceptable' 'acceptable' either
will work work well enough enough will
or 'unacceptable'." kmacceptable'." (p. (p. 243) 243) or
internal changes] changes] [signalling internal
Affects (associated (associated with with the the Affects
predictable because of this predictable
reticular activating activating system, system, reticular
the and small variability of the
p. 90) such such as as interest, interest, p.
internal environment." environment." internal
enjoyment, surprise, surprise, ffear, enjoyment, ear,
(p. 1124) 24) ""...a . . . a variety of materials
shame, arise in response to to learned or unlearned
regularly must be regularly
p. 337). triggers (p. 22, p.
transported in and out of of the the transported
There is aa partly invariant invariant There
body and thus drive signals
trigger-affect relation trigger-aff ect relation
and wane. wane."" (p. 125) wax and
(p. 23). Affect is partially
the independent of the motivational system; it can
amplifyy mask motivation, or amplif the drive system so as as to the the individual motivate the (p. 22). "This [affect] system is the primary
of blueprints for provider of cognition . . . " (p. 22) cognition..." "There is here no essential rhythm as there is
the drive with respect to the system."" ([(p. 125) system. PLUTCHIK ((1984) 1 984) Aroused by changing Aroused by internal internal states of of the the organism" (p. 214) 2 1 4) "Aroused the absence "Aroused by by the of of homeostatically homeostatically significant stimuli" (p. 214) significant 2 1 4) "There are are specific specific 'natural' objects objects toward toward which motives motives direct direct the the organism organism (e.g., (e.g. , food, food, water)" water),' (p. 214) 2 1 4)
"Aroused by external external stimuli" (p. 214) 214) "Aroused bbyy the the presence of a survival-related survival-related event" of (p. 214) 2 1 4) "There are few few 'natural' objects objects in in the the environment environment toward toward which which emotions emotions are are automatically automatically directed" directed" (p. 214) 214)
[Not compared] compared]
43 43
44
Chapter 2
Table Table 11 continued. "Induced before the process of of search is begun" (p. 214) "Tend to have a rhythmic rhytlunic character" (p. 2 14) 214)
"Induced after aRer an object is seen seen or evaluated" (p. 2 14) 214) "Depend on events in environment which may occur on a random basis" (p. 2 l4) 214)
IZARD 1 993) IZARD ((1993)
"Drives such as hunger, thirst, sex ...are cyclical in sex...are nature." (p. 72) 72) "[Drives are] dependent dependent upon peripheral physiological processes" (e.g., stomach growling; p. 73) "Drives provide specific inf ormation regarding the information the time and place that something needs to be done ... " (p. 73) done..." Drives, "cue a relatively specific set of . . ." of responses responses..." (p. 73)
An An emotion has has no temporal cycle (p. 73) ""...an . . . an emotion .. .is not emotion...is dependent dependent on peripheral physiological processes" (e.g., stomach growling) (p. 73) ""...can . . .can be associated with a virtually limitless variety of phenomena" (p. 73) Emotions "can "can motivate an equally wide range of cognitions cognitions and actions" (p. 73) "the emotions system preceded the cognitive system in evolution and outpaces it in ontogeny" (p. 73)
"Clearly, inf information "Clearly, ormation processing consists of several types or levels... levels . . . ranging from that which leads to the color of of an eye to that which produces a Mona Mona Lisa or a theory of relativity" (p. 73) "I propose four differentiable sorts of of information processing: processing: cellular, organismic, biopsychological, and cognitive . . . the first first three cognitive...the three of the forgoing categories involve types of noncognitive information processing" (p. 70) Cognition is about knowledge - learning, knowledgememory, symbol manipulation, thinking, and language (p. 73) Emotion-cognition interactions occur in all the many coping activities that require stimulus appraisal and judgment judgment before action (p. 73)
J.D. Mayer, H. H. Frasier Frasier Chabot Chabot and and K.M K.M. J D. Mayer,
Carlsmith Carlsmith
45 45
Hilgard ((1980, pp. 1113-114) concludes his his survey survey of of the the trilogy trilogy of of mind mind Hilgard 1 980, pp. 1 3-1 1 4) concludes shortly after after his his discussion discussion of of Bain, Bain, with with the the psychologists psychologists of of the the 11920's and shortly 920's and He comments: comments: 11930's. 930's. He Those in in America America who who were were proposing proposing aa new new experimental experimental or or Those psychology rejected rejected faculty faculty psychology psychology and and along along with with itit laboratory psychology the classification classification of of mental mental activity activity into into three three categories categories.., with [the [the the . . . with McDougall the the history history of of the the trilogy of of American psychologist] McDougall mind appears appears to have have ended, nearly nearly two centuries centuries after after itit began began in in mind Scotland. In In part, the the fading fading of of such such aa "generally "generally Germany and Scotland. accepted" view may have coincided with the the decline of a felt need mental processes. for such a comprehensive classification of mental mind was was still To be sure, Hilgard ((1980, 1 980, p. 1113) 13) wrote, "the trilogy of mind familiar in the vocabulary of psychology," but psychologists of the time were more interested in experimental advances than in the classification systems of the past. We believe that Hilgard's own interest in the trilogy suggests that its history was - and is - not over, although it may no longer occupy so central a place in the field. For that reason we proceed to more recent developments.
MacLean and and the influence oof psychiatry on the trilogy o/mind of mind MacLean /psychiatry By the mid-20th century century enough enough had been learned about about the brain structure and function function that that some initial statements could be made made regarding structure its relation to mental faculties. Of Of course, this had been attempted earlier. Phrenologists had attempted such as learning or Phrenologists had attempted to to connect mental faculties such or feeling to to specific brain brain areas, areas, for the the purpose purpose of of charting personality according cranial according to to aa shape shape of of an an individuars individual's cranium. Thus, Thus, someone with aa cranial indentation imagination would indentation alongside alongside the the presumed presumed brain-site for for imagination would be regarded as as having having aa stodgy, uncreative personality. But But phrenology was was regarded based on on pure pure speculation, speculation, and based and as as aa consequence, consequence, was was discredited. discredited. Brain localization localization became became aa reality, reality, however, however, with with the the identification identification of of Brain some some language language abilities abilities in in Broca's Broca's area. area. And And it it was was shortly shortly thereafter, thereafter, with with the the writings writings of of Paul Paul MacLean MacLean (e.g., (e.g., 1949, 1 949, 1973, 1 973, 1990), 1 990), that that the the trilogy trilogy of of mind mind found found aa possible possible home home in in brain brain science. science. MacLean MacLean inferred inferred from from the the structure structure of ofthe the human human brain brain the the existence existence of of three three partially partially independent independent subsub brains, or or brain brain divisions, divisions, which which reflected reflected three three distract distinct epochs epochs in in the the human human brains, brain's brain's evolutionary evolutionary development. development. The The first first such such brain, brain, which which was was structurally innermost, innermost, was was shared shared in in all all its its essentials essentials with with the the complete complete brain brain structurally
46
Chapter 2
of reptiles. The second brain, which corresponded to the limbic system, was of mammals. third brain, which corresponded ammals. The third shared in common with most m mans . MacLean to the cerebral cortex, was most highly developed in hu humans. ((1990, 1 990, p. 9) wrote: In popular popular terms the three evolutionary fonnations formations might be imagined as three interconnected biological computers, with each having its own special intelligence, its own subjectivity, its own sense of of time and space, and its own memory, motor, and other functions. Although MacLean never emphasized the point, parallels exist between ammalian brain, and conation and the reptilian brain, affect and the old-m old-mammalian ammalian brain. For example, the reptilian brain had cognition and the neo-m neo-mammalian associated with it, "such genetically constituted forms of of behaviour as selecting homesites, establishing territory, engaging in various types of of display, hunting, homing, mating, breeding, bree~ing, imprinting, forming social hierarchies, and selecting leaders." (MacLean, 11973, 973, pp. 9-10; 11990; 990; see also Table 1). The old mammalian brain, "plays an important role in elaborating emotional feelings that that guide behaviour with respect to the two basic life principles of ..." of self-preservation and the preservation of of the species species..." neomammalian mammalian brain, is concerned (MacLean, 11973, 973, pp. 12-13). The third, neo with higher cognitive processes. MacLean suggests a number of of innovative comparisons among the three brains. He notes that "the limbic system might be imagined as particularly designed to amplify or lower the intensity of of feelings involved involved in guiding behavior required for self-preservation and preservation of of the species." ((1991, 1 99 1 , p. 117). 7). He He further notes that the different brains vary as to their mammalian (cognitive) brain most external external orientation, with the neo neomammalian in that it receives its information through signals conducted from the eyes, ears, and somatic receptors (MacLean, 11991, 99 1 , p. 119). 9). MacLean's writings were influential in the 11950's 950's and it is not surprising that they turned up, shortly thereafter, in psychological writings more explicitly identified with the mental trilogy.
Modern psychologists fmind psychologists and and the trilogy oof Silvan Tomkins, an evolutionary emotions psychologist, focussed on the function of of psychological processes and may have been influenced influenced by
J D. Mayer. J.D. Mayer, H. H. Frasier Frasier Chabot Chabot and and K.M K.M. Car/smith Carlsmith
47 47
MaeLcan's writings. Recall Recall that that Maclean MacLean saw saw the the limbic system, system, which which was was Maclean's largely emotional, emotional, as as amplifying amplifying survival-related survival-relatod feelings; feelings; Tomkins Tomkins raised raised this largely idea again, again, arguing arguing that that the emotion emotion system's system's role role was was to to amplify motivation. motivation. idea Recall also also that Maclean MacLcan described described the the neo-mammal nco-mammalian brain as as more more Recall ian brain closely connected to the the outside world than were were the the paleopalco-mammalian mammalian or closely was perhaps influenced by this this comparison comparison when he he reptilian brains. Tomkins was notexi that the the emotion system was was directed toward toward the outside outside world whereas noted conativc system was directed to the the internal internal world. Finally, Finally, Tomkins the conative MacLcan and others of the time the use of an information informationshared with MacLean r processing metaphor, describing conation, for example, as providing "readouts" of the organism's internal states. Tomldns, conation has evolutionary significance in that it "plays a For Tomkins, in...self and reproduction" (Tomkins, 11962, central role in . . . self maintenance and 962, p. 29) as well as an information-processing aspect in which "primitive signal and feedback mechanisms" provide a readout of the internal homeostatic rhythms of the organism (Tomkins, 1962, p. 124). 124). Tomkins went on to carefully earcfuUy detail of of the characteristics that distinguished the conative conativc system from the some of affectivc. For example, Tomkins noted that "internal states" trigger conation, affective. and that conation is typically rhythmic. In contrast, "external stimuli" trigger emotion, and emotion follows no particular set timeline. timclinc. These ideas have Phtchik's ((1980) side-bybecome generally accepted. For example, Robert Plutchik's 1 980) side-by of conation and and affect included those and other distinctions side comparisons of that had been outlined by Tomkins. Plutchik's comparisons can also be found that in Table Table 1. 1. Tomkins and Plutchik both distinguish conation from emotion, with less attention paid to columns of of Table Table 1l are arc essentially to cognition (the cognition columns empty for these theorists). The distinction was likely viewed empty The conation-affcct conation-affect distinction as requiring more theoretical attention because motivation motivation and and emotion are are so inextricably intertwined in behavior. There is something so different between intertwined There conation conation and and affect, on the the one hand, and cognition, on the other, that that the difference was was often unattended unattended to to (Bain, 1855, 1 855, p. 6, made this same point). Nonetheless, Nonetheless, there there arc are some some difficulties involved in distinguishing distinguishing conation conation and A central central problem problem is caused caused by by the the frequent frequent use of of and affect affect from from cognition. A an information-processing metaphor metaphor to to describe describe both both the the functions of of an information-processing conation conation and and affect. affect. If If both both conation conation and and affect affect arc are processing information, information, what what is unique unique about about cognition? cognition? Tomkins' Tomkins' former former student, student, Cal Cal Izard, Izard, recently recently addressed addressed this this problem problem by by distinguishing distinguishing between between non-cognitive non-cognitive and and cognitive cognitive information information processing. processing. Non-cognitive Non-cognitive information information processing processing inchdes includes that that accomplished accomplished by by genetic genetic
48
Chapter 2 Chapter
codes, codes, chemical reactions, and and "reflective instinctive, instinctive, and and biologically prepared prepared or or genetically disposed disposed behavior" behavior" (Izard, 1993, 1 993, p. 70). Cognitive processing, processing, in contrast, contrast, "involves "involves more general and and flexible processes that that operate on experience based based learning learning and and memory. Cognitive activities operate involves judgment, planning, problem-solving and understanding."
Trends in thinking on the trilogy across time Trends in thinking Considerable shifts in meaning of of the tfilogy's trilogy's categories have taken place, even from Alexander Bain's writings in the late 19th 1 9th century to the present. This progression reflects (to us) aa cumulative understanding of of the utility of of the trilogy, and of of the differences among the tripartite areas. Several appear to to best best describe this progression: aa trend toward trends appear toward identifying the trilogy as taking place exclusively internal to personality, a trend toward of the trilogy in one or more brain areas, a trend localizing each member of toward an information-processing metaphor to describe them, and a of each class so as to create a more reformulation of more meaningful trilogy. distinguishing the internal from The trend trend toward distinguishing from the external. There constant recognition that conation, affect, and has been a more or less constant cognition are internal mental events, i.e., associated with brain function rather than with external events. Mendelssohn's comments that pleasure and pain not their cognition suggests that that cognition is change a person's will but but not something intrinsically private, hidden and autonomous (Mendelssohn, 11755/1971, 755/197 1 , p. 66). A century later, Alexander Bain struggled to define will's internal location. Bain ((1855, 1 855, p. 2) referred to will as conative action that of power to execute some work." Bain ((1855, 1 855, pp. required the "putting forth of Action...as of mind, we must 2-3) noted that, "In speaking of of Action . . .as a characteristic of render explicit the distinction between mental actions and such as are not mental." Bain's clarification that that action was "a characteristic of of mind," and therefore internal, was probably necessitated by his description of of mental action as "putting forth power," which could readily be mis-understood as taking place externally. This metaphorical difficulty evaporated with Maclean's MacLean's switch to the use of of information processing metaphors for brain function, which suggested an internal computer. The trend toward brain localization and inf ormation processing. informaaon Consistent with the internalization of these three processes was the attempt to find serious associations between the three classes and brain function. Although a non-scientific beginning to this pursuit originated with the phrenologists, serious connections awaited the works of of Maclean, MacLean, in
J D. D. Mayer, Mayer, H H. Frasier Frasier Chabot Chabot and and KM K.M. Carlsmith Carlsmith J
49 49
biological psychiatry, and Tomkins, in in psychology. psychology. Although Although Maclean's MacLean's work focussed focussed on on brain brain localization, localization, Tomkins' Tomkins' work work provided provided an an interesting interesting work supplemental supplemental conception conception by extending extending localization localization to to the the larger larger nervous nervous
system. For For example, "affects" "affects" were were "sets "sets of of muscle muscle and and glandular glandular responses responses system. the brain's reticular reticular activating activating system" system" (Tomkins, (Tomkins, closely associated with the p. 243). 11962, 962, p. Along with the the increased focus on on the the brain and and nervous nervous system system was was the the Along aforementioned industrial machines to an information information af orementioned shift in metaphor from industrial "putting forth of of processing paradigm. Bain's view of conative action as the "putting of mechanical power to execute some work" seems embedded in his own era of engines, whereas Tomkins' and
((1962, 1 962, p. 1124) 24) view that conation provides "signal
feedback mechanisms"
of internal organismic
information,
seems seems
computers.. Although Although the information processing embedded in an era of computers metaphor is today dominant it is still possible that multiple metaphors can best describe the phenomenon, just as in physics, light is both described as a wave and a particle (Bohr,
11963). 963). For example, conation seems best
described by combining Bain's and Tomkins' descriptions, so that conation is said to provide "a primitive readout of the internal, more or less homeostatic work."" rhythm of the organism", and generates "power to execute some work. trend toward finding finding more more homogeneous homogeneous categories categories at a common The trend level ooffunction. ffunction. There has also been an important narrowing of the trilogy's members such that each category is individually more circumscribed, and so that they operate collectively at a common level of function. For example, concepts of feeling, Bain's category of affect originally included the three concepts consciousness and sensation, whereas contemporary views have essentially restricted the category to emotions and closely related feeling states such as calmness and arousal. This This narrowing of focus represented a growing recognition that consciousness, sensation, and affect are incommensurate processes processes that perform different functions, are localized separately, and therefore are best treated separately. In today's Introductory Psychology has been paired off with perception, and consciousness is books, sensation has treated, if at all, all, in its own chapters. The remaining affect category retains treated, only emotion emotion and closely related feelings. This This narrowed version of the affect category seems more parallel parallel to the the similarly narrowed categories of conation and and cognition. similar and and no less important transition transition occurred for conation, conation, which A similar originally originally referred referred to to will, will, but but with with the the transition transition from from Mendelssohn Mendelssohn to to Tomkins Tomkins has has come come to to refer to to more-or-less more-or-Iess basic, basic, unlearned unlearned motivations. The The conation conation category category now now includes includes only basic basic motivations, motivations, which which are, are, once once
50 50
Chapter 22 Chapter
again, both both more more homogeneous, homogeneous, and and easier easier to to compare compare to to the the similarly revised again, category of of affect. category The concepts concepts of of "will" "will" and and "consciousness", although although excluded excluded from from the the The trilogy, were were not not plainly plainly grouped grouped with with other other parts parts of of the the mind. "Will" "Will" is trilogy, perhaps perhaps covered in personality personality in discussing self-control and and selfself could perhaps perhaps form aa fourth fourth category management. Consciousness, however, could to the the trilogy of of conation, affect, and and cognition - aa possibility we added to examine shortly. examine trend toward toward emphasizing emphasizing unlearned unlearned or or innate qualiaes. qualities. As the the The trend of eonation, conation, affect, and and cognition have categories of have been more narrowly the focus has has been directed toward their unlearned or innate focussed, the to distinguish these three mental categories has almost qualities. The effort to always best suece, exted when succeeded when descriptions of of them focus on their developmentally early, unlearned states. Thus, to say that that motivations are "rhythmic"~ "rhythmic'" whereas emotions are not, is to emphasize such motivations as hunger, thirst, and than more learned, less rhythmic motivations and sex, rather than such as aa desire for education or achievement. Similarly, to focus on the fact that emotions are triggered by external events is to emphasize their basic that might be triggered by nature rather than more complex, learned emotions that reminiscence. This lower level, more mechanical conception was yet another reason to homogenize the categories and dispense with those parts, such as consciousness and will, that did not fit well. What What remains in each category is a set of of mechanisms, or basic functions of personality. Recall that it was their basic mechanical qualities that led to the label of enablers for conation, affect, cognition, because they help personality get the job done. The reason this emphasis on innate, or minimally learned qualities of the enablers is so important, is that as learning increases, more complex structures are created created that are less plainly divisible into the three three categories. For, as the enablers engage together in more complex functions it it is clear that they become inexorably combined and intertwined. There exist a relatively few pure psychological enablers: pure conative urges for food and water, or pure affective joy or sadness, and pure memory memory networks. Soon after atter these enablers begin work, they construct a much larger set set of of established established thoughts that combine them. them. For example, a person develops develops models models of of the self, self, or a self concept, that that includes conation (what I want), want), affect affect (what I feel feel about about myself), myself), and cognition (what I know about myself). myself). But the the general general self selfconcept, which which includes includes all all three, by by necessity necessity integrates integrates the the enablers. enablers. It It was was concept, sensitivity sensitivity to to this this point point that that led led McDougall (1923, (1923, p. p. 266) 266) to to say say that that the the trilogy trilogy work work cooperatively cooperatively rather rather than than individually:
J.D. Mayer, H. Frasier Chabot and K.M K.M. Carlsmith J. D. Mayer.
5511
We often speak of an intellectual or cognitive activity; or of an act of willing or of resolving, choosing, striving, purposing; or again of that all mental a state of feeling. But it is generally admitted that activity has these three aspects, cognitive, conative, and affective; of these three adjectives to any phase of and when we apply one of mental process, we mean merely that the aspect named is the most of the three at that moment. Each cycle of of activity has prominent of this triple aspect; though each tends to pass through these phases in turn most which cognition, conation, and affection are in tum prominent; as when the naturalist, catching sight of a specimen, recognizes it, captures it, and gloats over its capture. trend toward more limited inclusiveness. Through the time of Bain, The trend some claim was made that the trilogy encompassed all mental function. With of the three classes of of mentation, it became the increasingly focussed meaning of easier to eject some concepts outside the trilogy. As has already been noted, sensation and perception were paired outside the trilogy. Similarly, will and of the consciousness were moved outside. The trilogy is no longer a trilogy of entire mind, perhaps, but remains a critical trilogy operative within the more molecular, basic aspects of of personality - and remains of of considerable research importance. emptor This particular reading of of the history of of the trilogy of of Caveat emptor. mind is, of are possible. The relational of course, our our own, and alternatives are model of of personality was constructed in part part according to this reading of of the evolution of of the categories and employs those categories according to their are possible. Still, the relational model has outline here; alternative models are that have been very evident strengths in relation to classification models that developed before (see Mayer, 1995b), 1 995b), and it is worth, therefore, further considering how the trilogy of of mind can be developed within it. Clarifying quaternity of of mind mind Clarifying the trilogy m in an expanded expanded quaternity
Although conation, have been narrowed and Although conation, affect, and and motivation have and clarified across across time, many of of the original distinctions among them still apply, even more above discussion, after atter all, has distinguished distinguished the even more clearly. The The above the three realms in several important ways. Phenomenological distinctions focus on different different conscious experiences of of the trilogy - that that conation, affect, and and motivation all "feel" differently from one another. Structural Structural brain brain distinctions focus focus on differences in brain brain localization of of the the trilogy.
52 52
Chapter 2
Functional Functional distinctions distinctions focus focus on on the the different different actions actions of the the three three systems, systems, and and so so on. These These distinctions, distinctions, as as well well as as aa number number of others, others, can can be be summarized summarized across across theorists theorists in aa new, new, enlarged enlarged format. format. To create create this this summary, summary, we we chose chose the the clearest clearest statements statements from from the the Table Table To 1, edited them, them, and and supplemented supplemented them them where where necessary, necessary, in Table Table 2. 2. Although Although 1, edited Table Table 2 was was constructed constructed on the the basis basis of the the above above discussion discussion of the trilogy trilogy of mind, the the table table denotes denotes aa quatcmity quaternity - consciousness consciousness has has been added. Some comment comment is necessary necessary on this. this. As As noted, noted, Bain Bain joined joined consciousness consciousness to to feeling, but but consciousness consciousness nowadays is just just as as likely to to be joined joined to to cognition (e.g., Bower, 1981), nowadays 1981), or & Cohen, Cohen, 1982). denoted as as aa blackboard blackboard to to represent represent all three three (e.g., Bower & denoted 1982). In implicated whenever any In fact, fact, consciousness consciousness is implicated any of the the three three systems systems reach reach aa high high enough enough level of activation. activation. For For these these reasons, reasons, it seems seems useful useful to separate separate consciousness consciousness from any single one of the other three and provide it interpretation of consciousness consciousness is that it with a place of its own. Because one interpretation is basic basic and elemental, elemental, a place place among the enablers enablers seems one possibility. possibility. Such a classification systemic viewpoint because, Such classification is useful useful from a systemic because, just just as the conative-enabler class class includes includes urges, instincts, conativc-cnabler instincts, and mental energy, so a conscious-enabler class class could include such components conscious-enabler components as the stream stream of consciousness, the phenomenal field, and so on. This consciousness, This provides provides a strong classificatory nothing else, for provisionally converting the classificatory rationale, if nothing quaternity, with trilogy into a quatcmity, with the addition of consciousness. consciousness. The Quaternity of of Mind and Personality Dynamics
If the discussion that discussion until now seems removed from contemporary contemporary concerns concerns that is one of the problems problems frequently encountered encountered with discussions discussions of classification. classification. Contemporary Contemporary research research is concerned with dynamics dynamics - causal causal or mutual influences mutual influences among among different parts parts of personality. personality. Another difference thus far and contemporary between the classification classification thus contemporary research research is the sheer generality of the discussion. discussion. So far, we have talked of all affect as if it were a single entity, when in fact, it is divisible into many parts. The contemporary researcher, in contrast, typically is interested in more specific personality parts parts and their dynamics. So, whereas up-to-now we have discussed the interaction between affect and cognition, the researcher researcher might be more interested in the influence influence of happiness happiness on memory. Discussion Discussion at the global level has indisputable indisputable value, however, because it can make clear the conceptual conceptual background within which more specific research is conducted. conducted.
JD. J.D. Mayer, H. Frasier Chabot and and K.M K.M. Car/smith Carlsmith
53 53
Table T a b l e 2. 2. Conation, Conation, Affect, Affect, Cognition, Cognition, and and Consciousness Consciousness Compared. Compared. Characteristic
Conation
Affect
Cognition
Consciousness
FUNCTION
To direct the organism to carry out basic acts so as to satisf satisfyy survival and reproductive needs needs
To organize a limited munber number of of basic responses quickly, adaptively, and in an organized fashion; to link those responses to complex situational environments
To learn from the environment and to problem solve so as to assist with motives and emotions
To assign mental activity where needed; to intervene flexibly in conation, affect, or cognition, where new new responses are called for
CONSCIOUS MANIFESTATIONS
If If conscious,
Direct consciousness itself; also reflective awareness of existence
specific urges, eat, to e.g., to toe.at, drink drink
If If conscious, the
pleasure and pain of of objects and stimuli; also, specific emotions such as happiness, fear, anger, etc.
Conscious and unconscious parts; conscious examination of of problem
AGENCY
Involuntary
Partly involuntary; partly voluntary
Mostly voluntary Partly voluntary; partly involuntary
DEVELOP-
Basic urges present immediately, including hunger, thirst, comfort comfort... .
Two or more basic emotions (e.g., pleasure, pain) present immediately; later development includes more complex emotions
Concrete reasoning early on, later the ability to reason with abstract information
MENTAL MENTAL ONSET
Unknown; selfawareness from 118 8 months; continuous conscious identity ffrom rom around around age 3 with the end of infantile amnesia
54 54
Chapter 2
Table Table 2 2 continued. continued. i,
Characteristic
Conation
Aff ect Affect
INITIATION INrITATION OF
Predominantly
Predominantly
Responsive
Responsive to
RESPONSE
responsive to internal bodily states
responsive to external environment
either to internal
non-habituated,
Cognition
Consciousness
or external
i.e., novel, or
environment
unusually intense, internal or external events
TEMPORAL 1EMPORAL
Motivations
Emotions often
Occurs any time; Alternates
CHARACTERCHARAC1ER-
precede action;
respond to
no set timeline
ISTICS
rise and fall
events; they
sleep-wake
rhythmically or
possess no set
cycle.
cyclically
timeline
INFORMINFORM-
Specific as to
ATIONAL
what is lacking
SPECIFICITY SPECIFICITY
and what what must
Identifies a class of possible that must events that be addressed, without necessarily being specific
be done
Either Either specific
according to the
Can incorporate
or general
and become
depending
aware of of a wide
upon problem
variety of
requirements,
inf ormation; is information;
work accompli-
very plastic in
shed, and and mental how it interprets capacity
inf ormation and information proceeds
BRAIN LOCALIZATIONS ATIONS
The limbic
Emotion is
Inf ormation Information
processing can can be distinguished associated with subcortical the limbic from higher structure, near level cognition. the center of cognition. of the system, particularly with Although the cerebral entire brain hemispheres. It the amygdala, It the encircles the top and secondarily processes inf ormation, we of information, of the brainstem. with the reserve cognition hypothalamus. It is commonly It There is also divided into to encompass flexible recent evidence three tracts, or processing based that the f rontal frontal circuits, cortex of composed of of the left on learning and hemisphere may memory; this different system is a
commonly
May be located in the the reticular activating system, or may be an emergent property of the mind as a whole
JD. J.D. Mayer, H. Frasier Chabot and K.M K.M. Carlsmith
55 55
Table Table 2 2 continued. continued. Consciousness Consciousness
Conation
Affect
Cognition Cognition
structures. structures. One mechanism of importance involves the hypothalamus which controls hormones hormones that target various parts of the body and may regulate drives, e.g., of hunger and sex (Reeve, 11992). 992).
specialize in processing processing positive emotion, the right hemisphere in negative. negative,
includes judgment, planning, planning, problem solving, and understanding. understanding. These are commonly viewed as dependent dependent upon the association cortex and the cerebral cortex.
Unmotivated Motivated
Unemotional Unemotional Emotional
Unthinking Thinking
Unconscious Conscious
SOCIALLY Constructive vs. DESIRED AIMS Destructive Motivations
Pleasant vs. Unpleasant Emotions
Intelligent vs. Unintelligent Thinking
Spiritually conscious vs. self-conscious
OPEN VERSUS Accepting Accepting vs. CLOSEDIINACCLOSED/INAC- Repressed CESSIBLE
In Contact vs. Out of Contact with Feelings.
Flexible vs. Rigid
Receptive versus Unreceptive
JOINT
*Basic emotions; e.g. happiness, anger, anger, fear **Complex emotions, e.g., shame, guilt, mixed emotions ***Sentiments (emotions attached to objects) e.g., loving one's country, country.
cognition: *Basic cognition: sensory motor operations, operations, learning ***Middle *Middle cognition: cognition: concrete operations, symbol learning ***Complex *"Complex cognitions: formal operations, operations, abstract thought.
*Basic consciousness consciousness **Reflective "Reflective consciousness consciousness ****Higher "Higher consciousness consciousness (e.g., reflective, spiritual, etc.).
Characteristic
DESCRIPTION OF QUANTITY
MOLECULARMOLAR DEVELOPMENTAL CONTINUUM
*Basic urges, e.g., hunger, hunger, thirst, physical contact; **Learned motivations: e.g., pleasing others, others, achievement ****Functionally * *Functionally autonomous autonomous motives, motives, e.g., doing a good job, helping others. others,
56
Chapter 2
For example, such such a general discussion can can provide hints as to where the important enabler-to-enabler interactions will take take place. Treating Treating more important conation, affect, cognition, and consciousness as equals would suggest there exist 6, i.e., (4 x• 3)/2, equivalently important sets of of interactions to cover. An An interesting interesting alternative alternative view, view, however, however, suggests suggests that that the the central central interactive interactive areas areas among among the the classical classical trilogy will be be more more limited. limited. Recall Recall Maclean's MacLean's triune triune brain brain that that emerges emerges in in stages stages from from conation conation to to affect affect to to cognition. cognition. If If we we assume adjoining adjoining areas areas (in (in tenns terms of of brain brain localization) localization) have have more more interactions, interactions, greater greater interactions should occur occur between between the the adjoining adjoining areas areas of of conation conation and and affect, affect, and and affect affect and and cognition, cognition, than between between conation conation and and cognition. cognition. This seems seems borne out out by by (our admittedly admittedly SUbjective subjective impression impression of) of) today's research literature, literature, which focusses focusses on on the the fonner former two two interactions. interactions. Limitations Limitations of of time time and and space space have have encouraged encouraged us us to to focus focus on on the the central central conative-affective, and affective-cognitive interactions. The interactions between between consciousness consciousness and and the the trilogy will be be considered briefly briefly at at the the end. end.
Conation and a ffoct affect To To recap, recap, conative conative phenomena phenomena concern concern include include hunger, hunger, thirst, thirst, and and reproduction. reproduction. Conative Conative functions functions chart homeostasis homeostasis in the the body body and and alert alert the the organism us we organism about about needs needs for for survival survival and and reproduction. reproduction. Thus, Thus, hunger hunger tells tells us we should eat; eat; thirst tells us us we we should drink, and and so so forth. forth. In In contrast, contrast, affect affect is concerned joy, and concerned with with such such feeling feeling states states as as happiness, happiness, joy, and alertness. alertness. Its Its primary primary concern concern is is to provide us us with with signals signals about about our our relations relations with with external individuals individuals and and objects. objects. Thus, happiness happiness tells tells us we we are are in hannony harmony with others, and anger that we are treated unjustly. with others, and anger that we are treated unjustly. It is plain that that conation and and affect must must serve the the same same master to some extent extent (e.g., (e.g., overall personality). personality). Thus, Thus, basic-level basic-level motivations motivations provide provide constraints ensure survival. constraints on on emotions emotions that that ensure survival. Say Say you you agree agree to to eat eat your your bagged bagged lunch lunch with with someone someone late late in in the the day. day. Then, Then, during during aa walk walk in in the the woods woods you you become become hungry hungry and and think think of of the the bagged bagged lunch lunch you you brought brought along. along. You You are likely to are likely to feel feel frustrated, frustrateA, but but you you won't won't eat eat immediately immediately because because you you know know it will make you feel guilty later. Should the motivation motivation to eat cat become stronger, however, however, most most people will eat, cat, so so as as to to promote promote their their energy energy and and clear-headedness their likelihood clear-headedness -- their likelihood for for survival. survival. In In the the above above instance, instance, motivation motivation (conation) (conation) and and emotion emotion work work together, together, assessing assessing different different necessities, necessities, and and balancing balancing one one against against another. another. In In that that example, example, whether whether motivation motivation or emotion "wins" "wins" is is aa matter of of which which signal signal (i.e., hunger or guilt) is the strongest. Often, however, more sophisticated sophisticated
J.D. Mayer, H. Frasier Chabot and K.M K.M. Car/smith Carlsmith JD.
57
is the interactions may take place. For example, the emotion system (which is more flexible) may "filter" motivations by allowing expression of those that are adaptive in a given situation, and by (at least temporarily) disallowing or suppressing those needs that are inappropriate. For example, if one is hungry, and there are people around who are eating, but none offer food, the original sense of of hunger may be replaced by a feeling of injustice. An An angry injustice might be a motivator for requesting food even though the act could be viewed as impolite or even improper (making a request might be suppressed by guilt). Say that, in this instance, the anger does replace the original hunger motivation and redirects the individual to ask for food. This is in part what Tomkms ((1962, p.. 22) meant when he wrote that "Affect.. "Affect...can Tomkins 1 962, p .can mask motivation, or amplify the drive system so as to motivate the individual." Similarly, Oatley and Johnson-Laird ((1987) 1 987) view emotions as coordinating motivational urges and plans. and emotion may contribute to one another more Finally, motivation and directly. Say you become happy because you have accomplished an important goal. You may need companionship as a consequence, and the motivational system may provide urges - phenomenological bursts of of energy - to assist you to pursue social companionship. As another example, you may suddenly become sad; motivationally you may need to return to your own territory, or as the present idiom has it, you "need space." Helpful or harmful though this hard to change its motivational accompaniment may be that moment, it is hard directional quality. Research on the the interaction between motivation and and affect often reflects explorations in physiological, non-verbal communication, and evolutionary psychology. A review of of such literature can be found in the chapter, "Motivation and and Emotion," in Mook's (1996) ( 1996) textbook, Motivation. Because this area has been reviewed so recently, and because a large portion portion of of it lies outside our own areas of ahead to the relation of expertise, we will move ahead between affect and cognition.
Affect Affect and and cogmtion cognition We already recapped We have have already recapped the the affect system, focussing on its depiction of of relationships between oneself oneself and the external world. The The cognitive system, on on the other hand, is useful for more flexible understandings understandings of of the the world and and the events in it. One of of affeet's affect's most most important contributions to to cognition is to to prioritize it (Mandler, (Mandler, 1984). 1 984). Thus, Thus, when working on aa project, aa fear home, although distracting at fear of of something going on at at home, at first,
58 58
Chapter 2
may turn tum one's one's attention attention to to what what is, ultimately, aa higher higher priority priority to to one's one's may survival. survival. Not only only do do affects affects interrupt interrupt cognitions, cognitions, but but they they can can also also change change them them Not in ways ways that that may may promote promote better better judgment judgment and and creativity. One One of of the the major major in influences of of affect affect on on cognition cognition is through through that that of of the mood-congruent mood-congruent influences cognition effect. effect. Modified slightly from Mayer, Gaschke, Braverman, & cognition Evans (1992, ( 1 992, p. 129), the the mood-congruent mood-congruent cognition effect: Evans . . . states that that people's cognitions are are sensitive to the the correspondence ...states sensitive to between pleasant-unpleasant quality of of their mood and the between the pleasant-unpleasant pleasant-unpleasant connotations connotations of of their ideas. An affective match pleasant-unpleasant between aa person's person's moods and ideas increases both the and the judged judged merit, broadly defined, of of those ideas. memorability and For example, example, mood-congruent concepts will be more more readily learned and recalled. In addition, mood-congruent mood-congruent ideas will be judged and richer in their associations, mood-congruent attributes will be judged as more applicable, mood-congruent examples of of categories judged will be judged as more typical, and mood-congruent causes and outcomes will be judged more plausible.
way mood facilitates It is possible to read into this effect another way cognition: As a person's moods shift, the shift will force changes in a person's turn, allows perspective on the surrounding world. Changing perspectives, in tum, for creative thinking about a problem, and the construction of a greater of action. Such mood shifts drag the cognitive number of alternative courses of system along with them, forcing alterations in thinking and motivating changes in perception, and potentially enhancing planning and creativity (see 1995). 986, or discussion in Mayer, McCormick, & Strong, 1995). Mayer, 11986, At a still broader level, cognitions seem to keep affects tolerable. That is, much thinking involves doing something for the emotion system, and consequently, for the motives those feelings relate to. This is what Tomkins p. 22) meant when he wrote that, tt"...this ((1962, 1962, p. . .this [affect] system is the primary cognition..." at least loosely related to provider of blueprints for cognition . . . tt It is also at notion that the ego derives its energy energy from from the the id. id. The The more more one's Freud's notion emotions are satisfied, the less directive they are and the more chance the cognitive system has to operate well according to its own rules of logic, propositions, and formalism. cognition follows the the blueprint of of affect, affect, itit can can also also tum turn Although cognition around and change change affect where where affect affect (or (or motivation) motivation) seems seems around .
d.D. Mayer, Mayer, H. H. Frasier Frasier Chabot Chabot and and K.M K.M. Carlsmith Carlsmith JD.
59 59
counterproductive. For For example, example, cognitions cognitions can can help help manage manage affects affects when when counterproductive. they get get out out of of hand, hand, and and separate separate good good or or useful affects, affects, from from misleading misleading they ones. So-called So-called metameta-or reflective experiences of of mood mood (e.g., (e.g., "This "This mood mood is is ones. or reflective clear clear to to me," "This "This feeling feeling is unacceptable," unacceptable," etc.) etc.) involve involve cognitive cognitive attempts attempts their responsiveness beyond to evaluate and regulate moods so as to improve their Mayer & & Gaschke, Gaschke, 11988; a simple reflexive attempt at survival (e.g., Mayer 988; Mayer Stevens, 11994; Salovey et aI. al.,, 11995). & Stevens, 994; Salovey 995). of emotional emotional intelligence (e.g., (e.g., Mayer & & The recently developed concept of Geher, 1996; Mayer & Salovey, in press; 1993; Salovey & Mayer, 1990) Geher, Salovey, 1 993; Mayer, 1 990) is the areas in which which emotion facilitates thought, and basically a compendium of the thought improves emotion. One recent definition of emotional intelligence (Mayer & Salovey, in press) describes it as including four broad classes of abilities: perceive accurately, appraise, and and express emotion: ....the . .the ability to perceive the ability to access and/or generate feelings when they facilitate thought; the ability to understand emotion and emotional knowledge; and the ability to regulate emotions to promote emotional and intellectual growth. of consciousness consciousness The role of
It is hard at this point. To To hard to resist aa mention of of consciousness' function at us, consciousness that of of a family bulletin board board upon consciousness plays a role similar to that which messages are placed (cf., Bower & Cohen, 11982, The 982, pp. 309-310). 309-3 1 0). The consciousness "bulletin board," more specifically, receives messages from consciousness conation, conation, affect, and motivation: urges, such as "need water," emotions, such as "anxiety", "I should talk talk more at at my upcoming "anxiety", and and thoughts, thoughts, such as "l meeting meeting to to appear appear more assertive." assertive. " Just Just as in aa family, each member has has different and cognition, cognition, have have their different handwriting, handwriting, so too, conation, conation, affect, affect, and their own individually individually recognizable recognizable modalities, their their signature signature phenomenology. An An integrated integrated personality personality recognizes recognizes messages messages from from each each source source because because it it experiences own terms, experiences each each differently, differently, and and evaluates evaluates each each system on its own terms, much much as as one one evaluates evaluates messages messages from from family family members members on on the the basis basis of of their their recognizable recognizable styles. styles. That That is, an an adult adult personality personality uses uses consciousness consciousness to to recognize recognize that that an an urge urge is an an urge, urge, and and as as such, such, has has aa different different status status than than aa logical proposition. Ideally, Ideally, itit weighs weighs the the urge urge ("l ("I am am increasingly increasingly hungry") hungry") logical proposition. with with the the thought thought ("This ("This project project would would best best be be finished finished before before lI eat") eat") and and wisely wisely chooses chooses which which to to follow follow depending depending on on circumstances. circumstances.
60
Chapter 2 Conclusions and Other Considerations
cnablers of of conation, affect, cognition, and consciousness The four enablers represent only the lower level portions of of personality. Emerging from them arc establishments, establishments, including models of of the self, the world, and the self-in self-inare the-world, and themes, coherent coherent collections of of features drawn from enablers cnablcrs and establishments that that are arc expressed as behavioral traits. Conation, affect, and cognition work closely together to support these more complex structures. For example, research on cognition and affect as they extend into a person's models of of the self and world (i.e., establishments) are being conducted by Fiske Fiskc and her colleagues on affect-triggered schemata (Fiske, (Fiskc, 11982); 982); by Higgins and his colleagues on self-schema and affect (Higgins, 1987), Pricstcr & Petty, 1 987), and by Petty and his colleagues on attitudes (e.g., Priester 11996). 996). Summary. Researchers in the area of of cognition and affect are, by virtue of of their interdisciplinary interest, unusually broad in the problems they pursue. Successful research across affect and cognition may be facilitated by of affect and cognition, the distinctions better understanding the scope of between them, and their relationship to personality. To better understand cognition and affect, their original grouping: conation, ffect, and cognition conaaon, a affect, the so-called trilogy of of mind - was was examined in considerable detail. We A and attempted to discover provided a historical review of of the the trilogy of of min min~! some trends in their evolving meaning. The definitions dofufitions of of conation, affect, cognition, were refined and updated. An An alteration of of the trilogy to a quatemity quaternity was recommended so as to include consciousness. This quatemity/trilogy was located within one possible contemporary model of of personality, the relational model. Finally, the relevance of of the quatemity and the interactions among its members were briefly applied to a discussion of of some contemporary research in cognition and affect. -
References
1 958). What Allport, G. W. ((1958). What units shall we employ? employ7 In G. Lindzey (Ed.), (Ed.), Assessment of of Human Motives (pp. 239-260). 239-260). New York: Rinehart & Company, Inc. and the intellect. London: John W. Parker 1 855/1977). The senses and Bain, A. ((1855/1977). & Son. [Reprinted [Roprintexi in D. N. Robinson (Ed.), (Ed.), Significant contributions to the history o fpsychology: 11750-1920 750-1 920 [Series A: Orientations; of Orientations; Vol. Vol. 4]. Washington, DC: University Publications of of America.
J.D. Mayer, Mayer, H. H. Frasier Frasier Chabot Chabot and and K.M K.M. Carlsmith Carlsmith JD.
6611
Barratt, EE.. SS.. ((1985). Impulsiveness defined defined within within aa systems systems model model of of Barratt, 1 985). Impulsiveness personality. In C. C. D. D. Spielberger Spiclbcrgcr & & JJ.. N. N. Butcher Butcher (Eds.), (Eds.), Advances Advances in in personality. (Vol. 5, 5, pp. pp. 1113-132). HiUsdalc, NJ: NJ: Lawrence Lawrence personality assessment (Vol. 13-132). Hillsdale, Erlbaum. Erlbaum. N.. ((1963). Bohr, N 1 963). Essays, 11958-1962, 958-1 962, on atomic physics and human knowledge. New York: Wiley. 36, 1129Bower, G. H. ((1981). Mood and and memory. American Psychologist, 36, 29Bower, 1 98 1). Mood. 148. Bower, G. H., & Cohen, P P.. R. ((1982). 1 982). Emotional influences in memory and Fiskc (Eds.), Aff Affect thinking: Data and theory. In M. SS.. Clark & SS.. T. Fiske ect Hillsdalc, NJ: Lawrence Erlbaum. and cognition. Hillsdale, S. E. ((1987). Buss, A. H., & Finn, S. 1 987). Classification of personality traits. of Personality and Social Psychology, 52, 432-444. Journal of Fiskc, SS.. T. (1 (1982). 982). Affect and cognition: The seventeenth Clark, M. S., & Fiske, Hillsdalc, NJ: Lawrence annual Carnegie Symposium on cognition. Hillsdale, Erlbaum. Eyscnck, H. J. ((1982). Eysenck, 1 982). Personality, genetics, and behavior. New York: Pracgcr. Praeger. Fiskc, S. T. ((1982). Fiske, 1 982). Schema-triggered affect: Applications to social Affect perception. In M. S. Clark & S. T. Fiskc Fiske (Eds.), Aff ect and cognition. Hillsdalc, NJ: Lawrence Erlbaum. Hillsdale, Higgins, E. T. ((1987). 1 987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94, 319-340. 3 19-340. Hilgard, E. R. (1980). and ( 1 980). The trilogy of of mind: Cognition, affection, and of the Behavioral Sciences, Sciences, 116, conation. Journal of of the History of 6, 107117. 1 17. Izard, C. E. E . (1993). ( 1 993). Four Four systems for emotion activation: Cognitive and noncognitivc ReView, 100, 1 00, 68-90. noncognitive processes. Psychological Review, MacLcan, MacLean, P. D. (1949). ( 1 949). Psychosomatic disease and the 'visceral brain'. Recent developments bearing on the Papcz of emotion. Papez theory of Psychosomatic Psychosomatic Medicine, 11, 1 1 , 338-353. MacLean, MacLean, P. D. (1973). ( 1 973). A A triune concept of of the brain and and behaviour. Toronto: Toronto: University of of Toronto Toronto Press. MacLcan, brain m MacLean, P. D. (1990). ( 1990). The triune brain in evolution: Role Role in paleocerebralfunctions. paleocerebraljUnctions. New New York: Plenum Plenum Press. Press. Mandlcr, Mind and and body: Psychology Psychology of ofemotion and and stress. New New Mandler, G. (1984). ( 1 984). Mind York: York: W. W. W. W. Norton Norton & & Co. Co. Mayer, Mayer, J. D. (1993). ( 1 993). A A system-topics system-topics framework framework for for the the study study of of personality. personality. Imagination, Imagination, Cognition, Cognition, and and Personality, 13, 1 3, 99-123. 99- 1 23 .
62
Chapter 2
Mayer, J.1. D. (1995a). ( 1 995a). The The system-topics framework and the structural structural of systems within within and and around around personality. personality. Journal Journal of 0/ arrangement of Personality, 63, 459-493. 459-493 . Mayer, J. 1. D. (1995b). ( 1995b). A framework for the the classification classification of of personality o/Personality, components. Journal of Personality, 63, 819-877. 8 1 9-877. Mayer, J. ( 1988). The experience and recta-experience meta-experience 1. D., & Gaschke, Y. N. (1988). of of mood. Journal of 0/Personality and Social Psychology, Psychology, .55, 55, 102-111. 102-1 1 1 . Mayer, J. 1. D., & Geher, G. (1996). (1 996). Emotional Emotional intelligence and the identification of of emotion. Intelligence, 22, 89-113. 89-1 1 3 . identification 1 . , & Strong, S. E. E . (1995). ( 1 995). Mood-congruent 1. D., McCormick, L. J., Mayer, J. mood: New evidence. evidence. Personality and Social recall and natural mood: 21, 736-746. Psychology Bulletin, 21,736-746. Mayer, J. of emotional ( 1 993). The intelligence of 1. D., & Salovey, P. (1993). Intelligence, 117, 7, 433-442. intelligence. Intelligence, 1. D., & Salovey, P. (in press). What is emotional intelligence? intelligence? In P. Mayer, J. (Eds .), Emotional development and emotional Salovey & D. Sluyter (Eds.), Implications/ or educators. New York: Basic Books. intelligence: Implications for Mayer, J. of the 1. D., & Stevens, A. (1994). ( 1 994). An An emerging understanding of 0/ Research reflective (meta-) experience experience of of mood. mood. Journal of reflective Research in 3 5 1 -373. Personality, 28, 351-373. Mendelssohn: Gesammelte Gesammelte Schrifien Schriften Mendelssohn, M (1971). ( 197 1). Moses Mendelssohn: Mendelssohn, 1 : Schriflen Schriften zur Philosophie Philosophie und Astheak). Asthetik). Jubilaumsausgabe (Band 1: Stuttgart: Friedrich Friedrieh Frommann Verlag (Gunther Holzboog). (Original work published 755). published 11755). Mendelssohn, ( 1969). Jerusalem (A. Jospe, Trans. & Ed.). New York: Mendelssohn, M. (1969). Schocken. (Original work 783). work published published 11783). Mook, D. G. ((1996). 1 996). Motivation: The organization o /action of action (2nd ed.). New York: W. W. Norton. 1987). Towards a cognitive theory of Oatley, K., & Johnson-Laird, P P.. N. ((1987). of emotion. emotion. Cognition Cogniaon and Emotion, Emoaon, 11,, 29-50. 29-50. Pervin, L. A. ((1990). 1 990). A brief history of of modem personality theory. In L. A. Pervin (Ed.), Handbook 0/ of personality theory theory and research research (pp. 3-8). New New York: York: Guilford. Guilford. Plutchik, R. ((1984). 1984). Emotions: A general psychoevolutionary theory. In K. R. Scherer & P. Ekman (Eds.), Approaches to emotion. emotion. Hillsdale, NJ: Lawrence Erlbaum. Priester, 1. J. R., & Petty Petty,, R. E. (1996). Gradual threshold model model of of ambivalence: Relating the positive and negative bases of attitudes to / Personality and Social Psychology, subjective subjective ambivalence. Journal 0of
JD. J.D Mayer, H. H. Frasier Chabot and KM K.M Carlsmith
63 63
71, 43 431-449. 71, 1 -449. of brain: Experimental paradoxes Pribram, K. H. ((1971). 1 97 1 ). Languages o f the brain: and principles in neuropsychology. neuropsychology. Englewood Cliffs, NJ: Prentice Hall.. Hall Understanding motivation and emotion. emotion. Fort Worth, TX: 1. ((1992). 1 992). Understanding Reeve, J. Jovanovich. Harcourt, Brace, Jovanovich. P.,. , Mayer, J. D.,. , Goldman, SS.,. , Turvey, C, & Palfai, T. ((1995). 1. D 1 995). Salovey, P Emotional attention, clarity, and repair: Exploring emotional intelligence using the Trait Meta-Mood Scale. In 1J.. W. Pennebaker (Ed.), Emotion, disclosure, and health (pp. 125-1 125-154). disclosure, 54). Washington, DC: American Psychological Association. P.. & Mayer, 1. J. D D.. ((1990). Salovey, P 1 990). Emotional intelligence. Imagination, Cognition, and Personality, 9, 1185-211. 85-2 1 1 . of Psychology, 11,, 1105-118. 05- 1 1 8 . Sears, R. R. ((1950). 1950). Personality. Annual Review of J. ((1984). Taylor, G. 1. 1984). Alexithymia: Concept, measurement, and implications of 141,725-732. for treatment. American Journal o fPsychiatry, 141, 725-732. Affect, consciousness. Vol. Vol. 1: ect, imagery, imagery, consciousness. 1: The positive Tomkins, S. S. ((1962). 1 962). Aff affects. affe cts. New York: Springer. Author Notes Author
developing the graphics for the Paul Presson was instrumental in developing relational model of of personality; his patience during design sessions enabled us to develop a far clearer picture than we would have otherwise, otherwise, and we are grateful for his assistance.
This Page Intentionally Left Blank
Cognitive Science SciencePerspectives Perspectiveson on Personality Personalityand and Emotion Emotion - G. G. Matthews Matthews(Editor) (Editor) Cognitive -
©9 1997 1997Elsevier ElsevierScience Science B.V. B.V.All All rights rightsreserved. reserved.
CHAPTER 33 CHAPTER
Introduction to to the the Bidirectional Bidirectional Associative Associative Memory Memory Model: Model: Introduction Implications for for Psychopathology, Psychopathology, Treatment, Treatment, and and Research Research Implications Warren W W. Tryon Tryon Warren
Learning and and memory memory are are arguably arguably the the two two most most fundamental fundamental Learning psychological psychological processes. Without learning, infants infants would not not acquire the the skills that make them children and adults. Without memory, cumulative learning could not occur; we would continuously relearn everything. All connectionistic neural networks (CNNs) both learn and remember; they entail a learning and memory mechanism. It is therefore impossible to discuss of memory or memory in the absence of learning. learning in the absence of An important advantage of CNNs is that they are also compatible with An explanations.. The possibility that the synaptic network biological and genetic explanations comes preset at birth with sensitivities to, and biases for, processing certain ways was addressed by Seligman ((1970) information in certain 1 970) and Seligrnan Seligman and Hager Hager (1972) and (1 972) in terms of of biological biological preparedness. It is also possible that not all aspects of of the CNN are equally modifiable by experience. It may be that that certain networks function essentially unchanged throughout the subject's lifetime. These possibilities the fact that many organisms, possibilities do not not detract detract from from the fact that especially especially humans, humans, learn aa great deal during during their their lifetime and and that that some of of what what is learned learned plays plays an an important important role role in developmental developmental changes. changes. Personality Personality is heavily heavily dependent dependent upon upon memory. Persons with with Alzheimer's Disease provide empirical support support for for this this assertion. assertion. Their Their personalities personalities gradually gradually dissolve dissolve as as empirical they they forget forget their their life experiences experiences including including where where they they have have been, been, what what they they have have done, done, and and who who their their children children and and parents parents are are or or were. were. Psychopathology and and psychotherapy psychotherapy are are also also highly highly dependent dependent upon upon Psychopathology memory. memory. A A phobic phobic person person is is afraid afraid only only because because they they have have anxious anxious memories memories about about certain certain stimuli. stimuli. If Ifthe the anxious anxious memories memories of ofaa car car phobic phobic can can be be replaced replaced with with memories memories of of positive positive experiences, experiences, then then the the person person will will no no longer longer fear fear automobiles. automobiles. Other Other feelings feelings not not generated generated by by immediate immediate environmental environmental stimuli stimuli are also also memories. memories. This This includes includes feelings feelings of of depression, depression, insecurity, insecurity, and and low low are self-image. self-image. Lotius Loftus (1980, ( 1 980, p. p. xiv) xiv) described described aa hypothetical hypothetical future future memory memory doctor doctor as as being being able able to to cure cure psychological psychological disorders disorders by by modifying modifying the the memories memories giving giving rise rise to to the the associated associated feelings. feelings. Schafer's Schafer's (1978) ( 1 978) hermeneutic hermeneutic psychotherapy psychotherapy seeks seeks relief relief in in just just such such aa way; way; by by recalling recalling and and altering altering
66
Chapter 3
past events through reinterpreting them so that they are memory for past experienced more positively and in an integrated fashion. Wachtel's ((1977) 1977) description of of psychoanalysis includes recapturing disassociated memories through catharsis. Psychoanalysis can be and diffusing their emotional impact through accurately summarized as a theory of of conscious and unconscious memory f ormation and formation and recall especially under stressfol stressful conditions. conditions. The Bidirectional Associative Memory (BAM), one variety of of CNN, enables psychologists to address many of of the same topics as psychoanalysts but with much more testable models since CNNs, including the BAM, can be implemented on a computer and are are therefore fully open to analysis and experimentation. An of CNNs is their fundamental compatibility with neuroscience added benefit of and biological psychiatry. Contemporary behavior therapy is dominated by cognitive and cognitive cognitivebehavioral models. Reference to emotion or affect is conspicuously absent; Ellis excepted ((1962, 1962, 11980). 980). Hollon and Beck's ((1994) 1 994) description of of cognitive and cognitive-behavioral therapies discusses thinking, beliefs, and interpretations as important elements but does not include emotional emotionalmotivational variables. Blatt and Bers ((1993, 1 993, p. 1165) 65) observe that "The role of of affect is not only ignored in most cognitive behavioral considerations considerations of of self-schemas, but it is often considered an impediment to the assessment of of them. Rather Rather than than viewing the self-schema as a cognitive-affective structure, research from a cognitive-behavioral orientation often attempts to eliminate or control current mood as possibly confounding the assessment of of schemas". The authors subsequently noted that cognitive-behavioral theorists are generally reluctant to explore motivational, affective, and developmental issues. Cognitive and information processing models of of normal and abnormal behavior stress intellectual control. Contemporary Contemporary behavior therapies for children and adults emphasize corrective thinking for emotional as well as behavioral disorders. Put otherwise, psychologists have over intellectualized emotional emotional disorders. Any comprehensive explanation of of normal and abnormal behavior must address emotion as well as cognition and behavior. The main purpose of of the purpose of of this chapter is to augment interest in modeling mechanisms underlying normal and pathological phenomena using connectionistic neural networks networks by applying one particular CNN, the BAM, to several areas of interest. The areas of The fact that CNNs in general and the BAM in particular ychologists means that little empirical work has particular are are new to many ps psychologists been conducted to date. Hence, this chapter cannot review and evaluate the BAM in terms of of quantitative empirical data. The scope of of this chapter is
W. W W. Tryon Tryon W
67 67
therefore therefore restricted restricted to to discussing discussing the the heuristic heuristic value value of of the the BAM BAM for for understanding aa wide wide range range of of phenomena phenomena related related to to cognition, cognition, emotion, emotion, and and understanding psychopathology. psychopathology. This chapter chapter extends extends the the BAM BAM so so that that it it learns learns emotions emotions of of varying varying This intensities in in specific specific contexts contexts thereby thereby forming forming affective affective memories. memories. It It is is intensities important to to note note that that the the same same learning learning mechanism mechanism used used to to form form intellectual intellectual important memories is is also also capable capable of of forming forming affective affective memories. memories. In In order order to to fully fully memories understand this approach approach to "hot cognition", we must must review how the BAM understand stores stores and and recalls recalls memories. memories.
Bidirectional Associative Associative Memory Memory (BAM) (BAM) Bidirectional The BAM BAM model model was was selected selected for for the the following following reasons. masons. First, First, it it is is The designed to form form and and recall recall memories. memories. Second, Second, because because the the BAM BAM is is equally equally designed able to to associate associate among among stimuli, stimuli, emotions, emotions, and and behaviors, behaviors, it it is is applicable applicable to to able the full spectrum the full spectrum of of psychological psychological and and behavior behavior disorder. disorder. Third, Third, the the concepts concepts of well and basin of BAM provide provide new of memory memory well and basin of attraction attraction associated associated with with the the BAM new ways ways to to conceptualize conceptualize psychopathology psychopathology and and treatment; treatment; both both psychological psychological and biological. biological. Fourth, the BAM BAM is is aa relatively relatively simple simple system system and and and Fourth, the consequently is aa good point point of of departure. departure. The The BAM BAM entails entails symmetric symmetric consequently interconnections that that the the brain does does not not have have and and is is therefore therefore less biologically biologically interconnections plausible than than some some other other neural neural networks. networks. However, However, the the BAM BAM is is not not plausible intended to be be an brain structure structure but rather to to intended to an exact exact brain brain copy copy of of an an actual actual brain but rather simulate selected brain as simulate memory formation formation and recall using selected brain functions such such as parallel processing at at each each node. The present parallel distributed distributed processing processing and and local processing present discussion derives mainly from Kosko (1987a, 1987b, 1988) and Wasserman discussion ( 1 987a, 1 988) Wasserman regarding how the BAM works. ((1989). 1989). The The Appendix Appendix provides details regarding Because it is is not not entirely necessary to understand understand every detail detail of of how the the Because it necessary to BAM functions to appreciate appreciate its heuristic heuristic value in understanding understanding psychopathology, a a succinct overview of of the most important important elements is given next. The stimuli and responses responses that that the the BAM learns to are The stimuli and BAM learns to associate associate are represented and O's represented as as vectors, vectors, aa sequence sequence of of numbers, numbers, of of l's l 's and O's defining defining the the presence presence or or absence absence of of aa set of of characteristics. characteristics. The The attributes attributes coded for can be cognitive, affective, and/or and/or behavioral behavioral which makes makes the BAM aa highly general detail can general model model of of memory memory formation. formation. Any level of of detail can be modeled. At aa very low level of can represent of abstraction, abstraction, vector vector entries entries can represent the state state of of individual sensory neurons neurons and and motor motor fibers. At At aa high level of of abstraction, abstraction, vector networks dedicated dedicated to vector entries entries can can represent represent the the results results of of other other neural neural networks to
68
Chapter 3
recognizing perceptual perceptual features features (red hair), hair), affeetive affective states (see below), below), and/or and/or behavioral dispositions dispositions (response (response vs. no no response response or behavioral or flight vs. fight). If If the the elements of of one vector (A) (A) index rows and and the dements elements of of the the second second vector vector columns, the the resulting resulting square matrix matrix (outer product) constitutes aa (B) index columns, matrix (M) for the AB association. association. For For example, if if vector vector A A = = memory matrix the AB 1, I , -3, 5, 7 and and vector vector B = = 2, 4, -6, 8, then memory matrix matrix M M is defined as follows:
11 Vector A A Vector
-3 5 7
.
2 2 -6 110 0 14 14
Vector B 4 -6
4
.
-12 20 28
-6 118 8 -3O -30 -42 .,
8
....... 8 -24 40 56 ,,,,
The numerical values of of the matrix elements dements simulate functional synaptic properties of of excitation (positive values) and inhibition (negative values). Multiple memories, up to a computable limit, can be accurately encoded into a single memory matrix by ssumming umming corresponding cells over all individual memory matrices. Memory recall is accomplished by multiplying a stimulus vector by the composite memory matrix. If the result of of multiplying vector A A times memory matrix M is not exactly vector B (correct recall), then the obtained result is fed back through the memory matrix by multiplying the obtained result by the transpose of of M. The result of of this calculation is used as a modified stimulus and therefore multiplied by M, as was vector A. A. The result will either be vector B or something closer to it This active reverberating and reconstructive process of memory recall, continues until vector B is fully recalled or no further improvement can be obtained in which case the memory recalled is, as with people, the best approximation that can can be be generated. This process enables pattern completion where a whole whole memory can often be reconstructed from a partial stimulus. Neural networks are are good at at Gestalt psychology. This pattern completion property will be emphasized in our discussions of psychopathology. Because of parallels with physics, an an "energy" "energy" value can be calculated for each memory. This calculation provides the two dimensional memory matrix with a third dimension; height in this case, that enables one to visualize memory fonnation formation as the creation of memory wells wells in an otherwise
W. w. W. W. Tryon Tryon
69 69
fiat minimum energy energy isis the the state state of of best best flat surface. surface. This This isis because because the the state state of ofminimum fit between between the the AB AB vector vector pair. pair. Memory Memory recall recall occurs occurs whenever whenever this this energy energy fit state recurs. recurs. state Imagine Imagine aa fiat flat rubber rubber sheet sheet upon upon which which aa ball ball bearing bearing has has been been placed placed causing causing aa vertical vertical indentation. indentation. Since Since the the ball ball bearing bearing comes comes to to rest rest at at aa point point below the the surface, surface, itit isis associated associated with with aa negative, negative, and and therefore therefore minimum, minimum, below energy energy state state (see (see Figure Figure 1). 1). That That memory memory formation formation isis associated associated with with aa minimum energy energy state state can can be be understood understood as as similar similar to to how how "best "best fit" fit" occurs occurs minimum when minimized. Both when the the deviation deviation of ofdata data points points about about aa regression regression line line isis minimized. Both are measures measures of offit. fit. are
I• II Iglll •• IBm
6•0•
LU
X
ILl
"I . 4}1t z. 2il ,. . IOlD
IBm ....J<J aim
Y
IzI I.
f.<J"
"I .
\
•
• II
2o
aoe
6e
X
Figure Figure 1. 1 . Example Example memory memory field field containing containing five five memories; memories; one one near near each each comer plus one in the middle. corner plus one in the middle.
The The process process of of memory memory recall recall can can be be visualized visualized as as placing placing aa small small frictionless frictionless ball ball on on the the memory memory surface surface and and letting letting itit roll roll down down into into aa memory memory well. well. The The memory memory is is recalled recalled when when this this locus locus of of memory memory recall recall reaches reaches the the
70
Chapter 3
bottom of of the the well well because because that that is is the point point associated associated with with the the best best fi fitt bottom between the the AB vector vector pair. pair. It It is to note note that that although although vectors vectors A A is important important to between and B have and B have been been represented represented in in a a distributed distributed way way across across several several or or many many elements and energy elements and entail entail many many synapses, synapses, the the entire entire memory memory has has aa single single energy value. Each Each memory memory formation formation creates creates aa local minimum in in the the memory memory fi field value. local minimum eld as by Figure Figure 1. I. as illustrated illustrated schematically schematically by All points points on on the the memory memory surface surface leading leading downward downward to to the the memory memory well well All are called called the the basin of attraction attraction for for that that memory. memory. Memory Memory wells wells vary vary in in their their are basin of depth (intensity) depth (intensity) and and breadth breadth of of their their basins basins of of attraction. attraction. Broader Broader and and deeper deeper memory wells wells are are more more potent potent organizers organizers of of cognition, cognition, affect, affect, and and behavior behavior memory than are narrower narrower and and shallower shallower ones. ones. The The possibility possibility that that trauma trauma creates creates a a than superbasin superbasin in in the the midst midst of of existing existing memories memories is is discussed discussed below. below. One One effect effect of an event event is to incorporate incorporate prior prior basins basins of of attraction attraction within within aa larger larger of such such an is to one, that the ow of one, tilting tilting those those basins basins so so that the fl flow of memory memory recall recall might might pass pass by by the the normal normal memories memories and and recall recall the the traumatic traumatic one. one. It should that memory memory wells It should be be noted noted that wells are are not not pure pure metaphor metaphor but but the the geometric geometric consequences consequences of of the the mathematics mathematics associated associated with with memory memory formation. Memory Memory wells wells are are visual visual representations representations of of the the mathematics mathematics of of formation. memory BAM memory formation formation and and therefore therefore are are explicit explicit consequences consequences of of the the BAM model. The BAM BAM generalizes across the traditional distinction model. The generalizes across the traditional distinction of of semantic semantic and memory. The memory mechanism and episodic episodic memory. The same same memory mechanism is is postulated postulated for for both both types types of of memories. memories. Encoding Encoding Emotion Emotion
This section section draws draws heavily heavily from from Tryon Tryon ((1996a). The first first section section briefl briefly This 1996a). The y reviews reviews previous previous efforts efforts to to encode encode emotions emotions into into CNNs. CNNs. Subsequent Subsequent sections sections recommend recommend more more direct direct solutions. solutions.
orts Previous eff efforts That the mediates emotion emotion makes neural networks That the brain brain mediates makes brain-inspired brain-inspired neural networks logical logical candidates candidates for for incorporating incorporating emotional emotional factors factors into into and and integrating integrating them cognitive processes. Levine and 1992) discuss them with with cognitive processes. Levine and Leven Leven ((1992) discuss "motivation, "motivation, emotion, and goal direction in neural networks". Part II of their book emotion, and goal direction in neural networks". Part II of their book contains contains articles articles on on "Top-down "Top-down processes, processes, attention, attention, and and motivation motivation in in cognitive cognitive tasks" tasks" Banquet, Smith Smith and and Giinther "A neural neural network network theory theory of of manic manicGunther (1992), ( 1992), "A by Banquet, depressive illness" 1 992), "Learned depressive illness" by by Hestenes Hestenes ((1992), "Learned helplessness, helplessness, memory, memory, and and the 1992), "Integration, the dynamics dynamics of of hope" hope" by by Leven Leven ((1992), "Integration, disintegration, disintegration, and and the the
7711
W. W W. Tryon W
frontal lobes" by Levine, Leven, and Prueitt ((1992) 1 992) plus "Familiarity and novelty: The The contributions contributions of of the the limbic limbic forebrain forebrain to to valuation valuation and and the the novelty: processing of relevance" by Pribram ((1992). Generalized affect, drive, has 1 992). already been been incorporated incorporated into into some some neural neural networks networks capable capable of of classical classical already conditioning (e.g., Grossberg & Levine, 11987; 987; Grossberg & Schmajuk, 1987). However, However, the the ability ability to to generally generally represent represent specific specific emotions emotions in in aa 1987). CNN has not yet been accomplished. Because CNNs can learn anything they can encode, encode, our our problem problem reduces reduces to to the the question question of of how how to to represent represent emotions emotions can as aa vector, vector, the the required required BAM BAM inputs. inputs. as
Color and emotion following sections sections draw draw heavily heavily upon upon Plutchik Plutchik ((1980, in press) press) and and 1 980, in The following Tryon ((1996a). 1 996a). Plutchik (in press) recognizes McDougall ((1921) 1 92 1 ) as the first author to to comment comment on on the the parallel parallel between emotions emotions and and colors. colors. Schlosberg Schlosberg author analyzed emotions emotions in response response to to the the 72 72 Frois-Wittman Frois-Wittman pictures pictures of of ((1941) 1 94 1 ) analyzed that they could be arranged in a two dimensional facial expression and found that 1 954) created a cone shaped model by adding an circumplex. Schlosberg ((1954) intensity dimension. The fundamental idea being that some emotions are primary, like primary colors, while all others derive from combinations of of basic emotions. Plutchik ((1994, 53-64) reports reports complete complete agreement agreement across across Plutchik 1 994, pp. pp. 53-64) investigators and no than 11 primary emotions emotions exist exist and and investigators that that at at least least 33 and no more more than 1 1 primary that other emotions emotions are are combinations combinations of these primary primary ones. Most theorists theorists that all all other of these ones. Most identify between between 5 and 9 primary emotions. Plutchik (1958, ( 1958, 1980) 1 980) proposed 88 basic based on Conte (1975) who asked asked subjects subjects to to rate 146 basic emotions based ( 1 975) who 1 46 emotional words on an 1I l-point from -5 = opposite, I-point bipolar scale ranging from through relation, to +5 = the the same same relative relative to three three reference reference words: through 0 = no relation, accepting, angry, angry, and and sad. sad. The The correlations correlations among ratings over subjects were accepting, among ratings over subjects were Nunnally (1967, ( 1 967, p. p. 299) 299) discusses discusses how correlation between calculated. NunnaUy how the the correlation between calculated. two variables can the cosine cosine of of an an angle angle between between two two unit can be expressed as the vectors vectors originating from the the same point. Using each of of three three words as aa referent, words were angular displacements referent, all all other other words were plotted plotted on on aa circle circle using using angular displacements calculated calculated from obtained obtained correlations. correlations. The The final angular placement placement was the average of the three methods; ofthe methods; each each using aa separate separate referent. The has the following structure: Items with with high high The resulting reSUlting circumplex has positive to one one another. Items Items that positive correlations correlations are are placed placed close to that are uncorrelated uncorrelated with one another another are are placed at at right angles. Items that that are negatively correlated, correlated, polar opposites, opposites, are placed opposite opposite one other. =
=
72
Chapter 3
at the top top and moving clockwise, the eight Basic Emotions Emotions are: Beginning at Joy. Acceptance, Fear, Surprise, Sadness, Disgust, Anger, Anticipation, and Joy. They form four bipolar pairs: Acceptance - Disgust, Fear - Anger, Surprise Joy. Beginning with any emotion on the circumplex Anticipation, Sadness -- Joy. and skipping three consecutive emotions identifies the polar opposite emotion. Using the color analogy, Plutchik ((1980) 1 980) defined Primary Emotional of adjacent pairs of of Basic Emotions which resulted Dyads as equal mixtures of of primary emotional dyads: Love = - Joy Joy + + in the following 8 pairs of Acceptance, Submission = = Acceptance + Fear, Awe = = Fear + Surprise, Embarrassment = = Surprise + Sadness, Misery = = Sadness + + Disgust, Scorn = = Disgust + Anger, Aggression = = Anger + Anticipation, Optimism = = Anticipation + Joy. These emotions also form four bipolar pairs: Love Remorse, Submission - Contempt, Awe - Aggressiveness, Disappointment Optimism. Secondary Emotional Dyads are formed from equal mixtures of of two Basic Emotions once removed, separated by one circumplex sector. Tertiary Emotional Dyads are formed from equal mixtures of of two Basic Emotions twice removed, separated by two circumplex eircumplex sectors. Plutchik (1994) noted that different words represent the same emotion at various intensities. For example, annoyance, irritation, anger, rage, and fury differ primarily in intensity. Adding an intensity dimension orthogonal to the circumplex represents these related emotions. Hence, every emotion requires a circumplex and an intensity code.
Proposed emotional codes The following five coding schemes for representing the three dimensional extension of of the emotional circumplex are offered. First, 8 vector elements -of-N code lI .. are sufficient to to represent one Basic Emotion as an 8-position l1-of-N Another 8 vector elements are required to to represent its intensity using a 2 thermometer code2.. This approach is economical in that only 116 6 vector elements are required but limited in that only a single emotion and its intensity are represented.
I1 A ll-of-N -of-N code selects from among N N = 8 choices as follows: 10000000 10000000 selects the first item, 00010000 selects the fourth item, and 00000001 selects the eighth item. 00010000 00000001 2 2 A thermometer code represents intensity by the number of of elements, from left to right, that are in the ·on" "on" position. If a thermometer code contains 8 elements, then the code 00000 00 indicates none, 111110000 1 1 10000 indicates half, and 111111111 1 1 1 1 1 1 1 indicates the 000000000 maximum maximum amoWlt. amount.
W W. W W.
73 73
Tryon
The second approach approach encodes one Primary, Secondary, or Tertiary Emotional Dyad by using 116 6 vector elements to encode two 8-position l1-of-N -of-N representing the appropriate appropriate pair of of Basic Emotions. Eight additional codes representing vector elements are required required to represent represent the intensity of of the composite thermometer code. This is limited in that that it assumes that that emotion using a thermometer mixtures mixtures of of Basic Emotions are achieved using components of of equal intensity. The third approach approach allows varying amounts of of Basic Emotions to be The encoded. Gradations Gradations of of pairs of of Basic Emotions can be accomplished accomplished by using using 32 vector elements to represent represent each of of the Basic Emotions using using an 8position ll-of-N -of-N code and their intensities using using an an 8-position therrnometer thermometer code. The The fourth method allows allows for the simultaneous simultaneous representation representation of of multiple Primary, Secondary, and Tertiary Tertiary Emotional Dyads by encoding multiple pairs of of Basic Emotions by doubling the number number of of vector elements described above in the second and third methods. This would require 32 or 64 described vector elements respectively. The The two two variants variants of of the the fifth fitth method are are general and and allow one to store from I1 to 8 Basic Emotions and their intensities. The first variant stores emotions in the vector array in any order by specifically encoding each emotion using 116 6 vector elements. The first eight elements use a l1-of-N -of-N code to select the Basic Emotion and the next eight elements to represent the corresponding therrnometer code. All 8 Basic corresponding intensity using an 8-position thermometer Emotions at 6 = 128 at 8 different intensities intensities can can be represented represented using 8 x• 116 vector elements. Emotions not represented represented in this method are encoded 00000000 as are their intensities. 00000000 The second variant variant of of the fifth method requires only 64 vector elements to store all 8 Basic Basic Emotions by presuming presuming that that these emotions are represented represented in a fixed order beginning with a fixed referent emotion. An An 8position thermometer code is used to indicate the intensity of each of thermometer to of of the the88 Basic Emotions. Emotions not present are coded 00000000. not present 00000000. Emotions occur occur in a context. Current Current behaviors, other persons, consequences, consequences, sights, sounds, smell, taste, taste, and touch touch provide important contextual contextual inforrnation. information. This context context information can be encoded using additional vector elements. =
74
Chapter 3
Application Emotion can be encoded 'using using the same vector approach used to by 1 983) to encode cognitive components regarding three Greek Anderson ((1983) mortals and three Greek gods using 50-element -16 50-clement vectors where elements 11-16 coded name information (Socrates, Alcibiades, Plato, Zeus, Apollo, Diana), elements 117-32 7-32 coded supernatural status (man or god), and elements 33-48 coded life span (immortal or mortal). Elements 49 and 50 were not used and were set to zero. All elements were 1 . Multiple elements were wore either + +1I or --1. used to represent discrete categorical information to implement distributed representation� representation; i.e., distributing information across multiple nodes or neurons. This approach can also be used to learn about emotions and the contexts in which they occur. Anderson's (1983) ( 1 983) approach could be used to associate emotions and contexts thereby forming emotional memories. This would result in purely emotional content like Anderson's purely cognitive content. Because emotion and cognition are highly interdependent, I (cf. Tryon, 1996a) propose extending cognitive vectors, like those used by Anderson ((1983), 1983), to include emotional and contextual information. The memories formed formexi by these vectors will be a cognitive-affective composite� composite; they will entail "hot cognition" cognition".. Prior research on the pattern completion, content addressable, addressablr properties of of these networks (i.e., Levine, 1 99 1 � Wasserman, 11989) 989) indicates that this approach Lcvine, 1991; will integrate cognitive and affective affectivr information. Presenting cognitive stimuli will recall affective affoetive memories. Presenting affective affectivr stimuli will recall cognitive memories. If context stimuli are included in the vectors, then presenting context stimuli will recall both cognitive and affective affectivr memories.
Additional Additional considerations The approach just just take� taken represented representeA emotion and cognitIon cognition using a common vector. A consequence of of this choice is that affective affectivr and intellectual content are highly integrated, fused, into a single memory and therefore jointly influence behavior and the retrieval of of associated fused memories. This approach assumes that emotions have no special status in connectionistic systems and consequently do not need to be treated separately in terms of of how they are stored and/or processed. This approach is parsimonious in that no second memory system is required to store emotions. Nor is a second emotion processor required. Nor is any method required to
W. W IV. W
Tryon
75
integrate emotion emotion and cognition. These theoretical advantages strongly argue for coding emotion and cognition in the same vector. An alternative approach approach is to to represent emotions using a separate vector. This This approach assumes that using a single vector to represent both emotion separate vectors and cognition is inappropriate. One possible reason for using separate is that emotion and cognition may be considered to be different experiential domains. This approach approach does not solve the problem of of how emotion and cognition are fused into unified memories. It does not address how emotion is integrated with cognition. Is a second network required to integrate these two aspects of of experience? At least, some additional processing at some level is required but the nature nature of of such processing remains unspecified. It should should be noted that all of of the above mentioned CNN options presume the existence of of other networks functioning as feature detectors detectors and that that they selectively turn tum individual vector elements on or off in both the cognitive and affective fields. These feature detection networks are responsible for perceiving cognitive and affective elements from sensory experience.
Implications for DSM-IV DSM-IV Disorders Disorders Implications The purpose of of this section is to show that the memory concepts described above are generally applicable to a broad range of of psychopathology found in DSM-IV (APA, 11994). 994). Of Of special relevance are the concepts of of energy well and basin of 995a). General reference to of attraction (Tryon, 11995a). neural networks will occasionally be made when appropriate. Hypotheses about 994) disorders are presented, implications for about DSM-IV (APA, 11994) treatment are are considered, and and recommendations are are made made for future research. What What follows are are suggestions regarding new ways to think about clinical disorders based on a few neural network principles rather than definitive resolutions and proofs supported by empirical research. Treatment implications and directions for future research follow are subsequently considered. One approach to this section is to select one or two specific DSM-IV diagnoses within a single DSM-IV category to illustrate the theoretical relevance of of neural networks. Since the underlying logical structure of of this section is induction, critics could rightly question whether the BAM applies to any other other disorders. Hence, I chose to address the question of of generality by discussing multiple diagnostic entries within each of of the three major DSM-IV categories: Dissociative, Anxiety, and Mood Disorders.
76 76
Chapter Chapter 3
Dissociative disorders disorders Dissociative Carson, Carson, Butcher, Butcher, and and Coleman Coleman (1988) ( 1 988) define define dissociation dissociation as as the the "separation "separation or or 'isolation' 'isolation' of of mental mental processes processes in in such such aa way way that that they they become become split off off from the the main personality or or lose their their normal thought-affect relationship" (p. G-5). Hence, dissociation is aa disruption disruption of of normal associative processes. processes. Such Such aa dysfunction dysfunction could could be be due due to to problems problems with with associative memory memory formation formation and/or and/or recall. recall. It It is is important important to to determine determine if if aa person person with with thought-affect isolation: isolation: (1) ( 1) ever ever had had appropriate appropriate emotions, emotions, (2) (2) had had aa thought-affect appropriate emotions emotions but but lost lost them them at at some some point, point, or or (3) (3) retains retains appropriate appropriate appropriate thought-affect associations for some topics topics but but not others. A A neural network explanation of of the the first first case case implicates implicates abnormal abnormal learning learning experience, experience, explanation abnormal neural neural architecture, architecture, and/or abnormal memory formation process. process. abnormal and/or abnormal memory formation Excessive axonal pruning such as Hoffman and Hoffinan Hoffman and Excessive axonal pruning such as Hoffinan (1987, ( 1 987, 11992) 992) and and Dobscha (1989) ( 1 989) have reported in in adolescent adolescent schizophrenics schizophrenics can can explain explain why why Dobscha have reported normal cognition as children children but "loose" associations adults. they had had normal cognition as but form form "loose" associations as as adults. Traumatic or toxic toxic insult insult to brain of why normal normal Traumatic or to the the brain of adults adults can can explain explain why thought-affect lost at thought-affect associations associations are are lost at some some point. point. Having Having appropriate appropriate thought-affect associations associations for for some topics but not others specific thought-affect some topics but not others implicates implicates specific learning learning experiences. experiences. Spiegel (1990) ( 1 990) discusses discusses three three theoretical Spiegel theoretical advantages advantages of of parallel parallel distributed processing processing (PDP) First, the the distributed (PDP) models models of of dissociation. dissociation. First, autoassociative property of autoassociative pattern pattern completion completion property of neural neural networks networks causes causes them them to complete memory given partial This means to recall recall aa complete memory given partial information. information. This means that that one one need all aspects aspects of need not not be be fully fully conscious conscious of of all of aa stimulus stimulus situation situation before before reacting Second, neural reacting to to the the situation. situation. Second, neural networks networks entail entail local local learning learning without governance governance from from aa central central processing processing unit. Learning can can therefore therefore without unit. Learning take place at take place at different different levels levels with with varying varying degrees degrees of of consciousness. consciousness. Third, Third, "The concept concept of of dissociation dissociation implies implies some some kind kind of of parallel parallel access access to to "The awareness" (p. 1123); 23); hence, PDP models awareness" (p. hence, PDP models intrinsically intrinsically reflect reflect aa fundamental fundamental property 1977) explained property of of dissociation. dissociation. Hilgard Hilgard ((1977) explained hypnosis hypnosis using using aa horizontal 1 920), versus horizontal view view of of conscious conscious states, states, as as did did Janet Janet ((1920), versus Freud's Freud's vertical vertical model. model. Conscious Conscious states states are are seen seen as as existing existing side side by by side side like like rooms rooms in in aa one one story story ranch ranch house. house. Dissociation Dissociation entails entails access access to to some some rooms rooms but but not not others. others. Hilgard Hilgard hypothesizes hypothesizes that that hypnosis hypnosis activates activates two two or or more more of of these these distributed distributed conscious conscious states. states. An extraordinary extraordinary Dissociative Identity Disorder (multiple personality). An consequence massive and consequence of of massive and systematic systematic dissociation dissociation can can be be the the formation formation of of two or or more more personalities personalities each each with with their their own own distinct distinct set set of of memories memories and and two
W. W W. Tryon W
77 77
associations. One One neural neural network network explanation explanation of of two two personalities personalities entails entails the the associations. creation of of two two large large crater-like crater-like basins basins of of attraction attraction in in an an otherwise otherwise flat fiat creation memory surface. surface. While While the the locus locus of of memory memory formation/recall formation/recall remains remains in in the the memory first crater, crater, normal normal size size memory memory wells wells form form in in the the craters' craters' floor. floor. The The crater crater first walls usually prevent prevent the the locus of memory memory recall recall from from flowing flowing into into the the other other walls usually locus of crater, super super basin basin of of attraction. attraction. However, However, aa sufficient sufficient temporary temporary energy energy crater, increase might might jump jump the the locus of memory memory formation/recall formation/recall out out of of the the first first increase locus of crater and and into into the the second second one one thereby thereby causing causing aa personality shift. An indirect indirect crater personality shift. transfer could could occur occur by by first boosting the the locus locus of of memory memory formation/recall formation/recall transfer first boosting onto the the plateau plateau dividing dividing the the sunken sunken areas areas and and "rolling", "rolling", converging, converging, into into the the onto other basjn basin thereby thereby accomplishing accomplishing aa personality personality shift. shift. This This argument argument assumes assumes other that momentum momentum from from the the ejection ejection is is sufficient sufficient to to traverse traverse the the flat fiat middle middle that section. While While in in the the second second crater, crater, normal normal size size memory memory wells wells would would form form in in section. its floor floor until until the the locus locus of of memory memory formation/recall formation/recall was was again again ejected ejected and and its returned to the the first first associative associative area. area. returned to alternative means means of of isolating isolating memory memory formation formation sites sites into into two two An alternative functionally separate separate regions regions would would be the formation formation of of a a wall or mountain mountain functionally be the wall or range like structure such such that that the the locus locus of of memory memory formation/recall formation/recall would would range like structure normally be contained on one one side; side; in in one one valley or the the other. other. This This normally be contained on valley or formulation differs differs from from the the one one above above in in that that aa thinner thinner barrier up from from formulation barrier rises rises up a flat fiat plane rather than than requiring requiring two two large large depressed areas to to be be formed. formed. As As a plane rather depressed areas before, before, unusual unusual circumstances circumstances may may push push the the locus locus of of memory memory formation/retrieval formation/retrieval over over the the top top of of this this elevated elevated structure structure allowing allowing it it to to roll roll (flow) into the the other the dividing (flow) into other side, side, valley. valley. Alternatively, Alternatively, it it is is possible possible that that the dividing structure completely from one end end of the memory memory field structure does does not not extend extend completely from one of the field to to the the other. 5% region other. Perhaps Perhaps it it occupies occupies the the middle middle 90 90 percent percent leaving leaving aa 5% region at at each each end end where where it it is is possible possible to to move move from from one one side, side, valley, valley, to to the the other other without without crossing the the barrier. barrier. Or Or perhaps perhaps the the dividing dividing structure structure begins one end of crossing begins at at one end of associative memory memory and and extends extends 90 90 percent percent of the way way across across the the memory memory associative of the field one end. that the the field leaving leaving aa 10% 1 0% transfer transfer section section at at one end. A A third third possibility possibility is is that barrier the memory barrier extends extends from from one one end end of of the memory field field to to the the other other but but with with one one or or more along the the way way providing more paths more significant significant breaks breaks along providing one one or or more paths to to the the valley. A A variant variant of of all all three three options options is is that that the the barrier barrier might might extend extend other other valley. from from one one end end of of the the memory memory field field to to the the other other but but have have variable variable height height such such that that extraordinary extraordinary energy that at at one one or or more more points points it it becomes becomes low low enough enough that energy increases increases are are not not needed needed to to cross cross the the barrier barrier at at these these places; places; like like crossing crossing aa mountain the barrier mountain pass. pass. A A fourth fourth possibility possibility is is that that the barrier might might have have variable variable thickness thickness such such that that the the locus locus of of memory memory formation/recall formation/recall might might tunnel tunnel through, through, penetrate, penetrate, the the barrier barrier at at its its thinnest thinnest points points analogous analogous to to tunneling tunneling in in
78
Chapter 3
quantum mechanics. Additional alternatives can be created using different combinations of of these possibilities. The The nature of of memories/associations occupying all transfer regions is especially important because because activity in these areas areas positions the locus of of memory recall/formation where the probability of of crossing to the other side is substantially greater than with other topics. Future investigators should determine the topography of of associative structures structures to comment upon the above hypotheses. Dissociative Amnesia. DSM-IV (APA, 11994) 994) cites the ""... . . . inability to important personal personal information, usually of of a traumatic or stressful recall important nature, that is too extensive to be explained by ordinary forgetfulness" as the primary inclusion criterion. To explain Dissociative Amnesia via the BAM model is to explain how to prevent memory retrieval from occurring. We consider two possibilities. Energy hills. The formation of of energy wells in one area of of the memory of energy hills in one or more other areas of of the surface entail the formation of same surface. Memory encoding for every association also encodes its of the vector 11,, 0, 11,, °0 is the vector 0, 0, 11.. 0, 11,, 0, complement. The complement of of opposite sign. The energy values for both memories are equal but of Memories have negative energy values and and their complements have positive energy values resulting in energy hills. Memory hills can can extend the basin of of of a memory well if if they are adjacent to the well. A locus of of attraction of memory retrieval located anywhere on the side of of the energy hill adjoining the memory well will descend into the memory well. Since memory recall entails seeking energy minima, an energy hill prevents memory recall. If a memory hill were to be created in a path normally taken toward a memory well, then the associative process would be blocked in direct proportion to the diameter of of the base of of the memory hill. Perhaps a series of of adjacent energy hills could wall off a memory well thereby isolating it. off EEG. If If the polarity of of the BAM remained constant, then the contents of of memory hills would not ordinarily be accessed. Additional energy would be required to move the locus of of memory recall up the hill. This problem is solved by temporarily reversing the sign of of all BAM vectors which temporarily converts memory hills into memory wells thereby allowing retrieval of 1 ) dialectical of memory opposites consistent with Rychlak's Ryehlak's (198 (1981) emphasis. This process is especially efficient when memory hills border directly on memory wells such that the surface of of the well wall and the hillside are contiguous or nearly so.
IV. W IV.. Tryon Tryon W
79
Memory access takes time. If the memory well is deeper than the product of memory retrieval speed and repolarization time, then the memory well will be repolarized into a memory hill before the locus of recall reaches the bottom and the memory is retrieved. This could account for the inability to retrieve traumatic memories since they are hypothesized to be associated with deep memory wells. Time is reciprocally related to frequency. Faster cycling results in less time per cycle. More time per memory cycle would be available in a relaxed state where polarity reversals occur more slowly thereby providing more time to access deeper wells and consequently allowing access to information not available during a more aroused state. Martin ((1991) 1 99 1 ) labels the 0.5 - 4 Hz EEG band as delta, the 4 - 7 Hz band as theta, the 8 - 113 EEG 3 Hz band as alpha, Hz band as beta. Martin ((1991) that "Beta waves and the 113 3 - 30 Hz 1991) indicates that are normally seen over the frontal regions and over other regions during intense mental activity" (pp. 778-779). Slower "alpha waves are generally of relaxed wakefulness" (p. 778). Martin further associated with a state of that delta and theta theta waves are associated with sleep and have the indicates that largest amplitudes. These results results are are in exact agreement with BAM model expectations. These Rapid Rapid polarity changes would quickly give alternate access to memory hills and and wells but would prevent memory access to deep memory wells. This rapid beta beta waves being associated with intense hypothesis is consistent with rapid mental activity. It also explains why anxiety impairs memory (e.g., test anxiety). The The rapid rapid EEG EEG oscillations associated with hyper arousal provide little time for memory access. A slower rate rate of of oscillation provides more time to reach reach deeper memory wells. This hypothesis is consistent with better where memory for traumatic traumatic events under under conditions of of relaxed wakefulness where alpha waves predominate EEG slower alpha predominate such as during hypnosis. The slowest EEG waves are associated associated with sleep and and give the longest time to probe deep memory wells and consequently access more traumatic traumatic memories, perhaps perhaps in dreams since we are not not conscious when when these memories are being the form of of dreams retrieved. Such memory locations locations may Such slow slow alternate alternate access access to to memory may further impair our ability to to effectively process process the the recalled information. Still deeper deeper our memory wells may may not not be be accessed accessed at at all and and thereby thereby are are fully dissociated dissociated from consciousness. consciousness. Perhaps Perhaps these these deepest deepest recesses could be be probed probed if if ways ways from were were found found to to further further slow the the EEG EEG or or to to increase the the proportion of of slower EEG waves. A A corollary to to the the above above argument that all stimulants defend EEG argument is that against against memory retrieval by by limiting repolarization repolarization time. This This could explain explain why why some people people chronically seek stimulation. -
80
Chapter 3
The possibility that the the energy values associated with memory wells could oscillate is entirely a theoretical conjecture at at this point. While the impression may have been given that memories are stable in all respects it may be that while their structure, shape, remains constant that that their polarity oscillates. Just as there there is no empirical evidence that memory wells can oscillate with regard regard to polarity, no empirical evidence indicates that they cannot do so. The possibility of of oscillation is theoretically attractive in the ways described above and deserves further consideration. These hypothetical memory polarity changes are not necessarily explained by the same mechanisms that govern EEG signals; they could have an entirely different physical basis but covary c,ovary with EEG. Dissociative Fugue. The same dissociative mechanisms discussed in Dissocmtive Fugue. may also operate here. connection with Dissociative Amnesia above may Avoidance Avoidance sometimes entails removing ones' self to a different location. Elements of of Dissociative Identity Disorder (multiple personality) are are present to the extent extent that that the person assumes a new identity elsewhere. Depersonalization Disorder. This disorder appears to involve a mild form of of the dissociative mechanisms discussed above in that memory is not lost for specific events.
Anxiety disorders Posttraumatic Stress Disorder (PTSD). (PTSD). Although not explicitly connectionistic 1 988), Creamer, Burgess, and connectionistic in nature, Chemtob et r al. ((1988), Pattison ((1992), 1992), Foa and Kozak ((1986), 1 986), Foa and Riggs ((1993), 1 993), Foa, Steketee, and Rothbaum 1989), Foa, Zinbarg, and Rothbaum ((1992), 1992), and Rothbaum ((1989), Lang (1979, ( 1979, 11985) 985) theoretically implicate an emotion-memory network in the of PTSD. Their Their work is important important here because it emphasizes the etiology of concept of of network, and by implication, the parallel distributed processing approach to PTSD 1 990) and Litz PTSD advocated below. Both Jones and Barlow ((1990) ((1992) 1992) cite Lang's (1985) ( 1 985) emotion-memory network (cf. Lang, 11979). 979). Leventhal's 1 984) perceptual-motor Leventhars ((1984) perceptual-motor theory of of emotional response entails associative memory structures which neural networks clearly are. Li and Spiegel ((1992) 1 992) explain both PTSD PTSD and Multiple Personality Disorder in terms of of traumatic traumatic constraints constraints placed on a neural network which alter the topology of of its "goodness-of-fit surface" which conforms to the memory energy field discussed above. characteristics of of a Jones and Barlow ((1990) 1990) require the following characteristics comprehensive PTSD PTSD theory: 11)) symptom constellation of of the disorder, 2)
W W Tryon W. W. Tryon
8811
differential symptom severity, 3) 3) reexperiencing reexperiencing the trauma through memory, 4) why only some people develop PTSD given a traumatic experience. 4) includes: 11)) persistently The DSM-IV (APA, 11994) 994) symptom constellation includes: reexperieneing the traumatic event, i) 2) persistent avoidance of of stimuli reexperiencing associated with the trauma and a general emotional numbing, and 3) 3) 1996) persistent symptoms of of arousal. Brewin, Dalgleish, and Joseph ((1996) specified five points that a complete PTSD theory must address. Two of these issues are the same as Jones and Barlow's criteria and three are different. The two that are the same are: a) to account for the clinical characteristics of PTSD PTSD and b) to explain individual differences in symptom severity. The three additional criteria are: a) to indicate whether PTSD symptoms are indicative of an abnormal process and if if not how they differ from normal processes, b) of why comorbidity is found with depression, generalized anxiety, substance abuse, and somatization disorder, and c) doing a better explanatory job than other theories plus making novel predictions. This third requirement contains two distinct parts. The first concerns comprehensiveness of explanation and 1 996b) provides a the second entails new and unique predictions. predictions. Tryon ((1996b) BAM explanation of of all required aspects of of PTSD. 994, pp. Because the DSM-IV (APA, 11994, pp. 428-429) 428-429) diagnosis of of Posttraumatic Stress Disorder requires symptoms to exist for at least 11 994, pp. month, and because the diagnosis of Acute Stress Disorder (APA, 11994, 1 -432) requires symptoms to last between 22 days and 4 43 431-432) 4 weeks, the following comments pertain to both disorders. disorders. Combat. War experiences experiences are hypothesized to warp existing memory energy fields by creating new deep memory wells with broad basins of of attraction. Stimuli which previously flowed to normal memories and associations now reside within the basin of of attraction of a war related memory and therefore retrieve war memories. The energy minima seeking nature of cognition and the steepness of of the war-related energy well walls causes the locus of of memory retrieval to pass through or by a previous deeper, more compelling, war-related terminus toward the memory/association. Unable to prevent this associative process from reaching energy minima, the PTSD veteran copes by minimizing the frequency of of war related associations through avoidance of all stimuli associated with the broad basin of attraction resulting in emotional numbing in direct proportion to the scope of of the attractor basin. The clinical appearance is that many of of their memories work in unison. To reduce the frequency of of war-related feelings, PTSD patients find it helpful to inhibit all feelings resulting in emotional numbing. numbing. and
82
Chapter 3
Since a separate memory well is associated with every memory, many deep memory wells may exist, each with its own broad basin of attraction. PTSD severity would therefore be a function of the depth and breadth of all attractor basins plus their number and distribution across the memory field. Perhaps Perhaps these memory wells are arc separated by broad relatively flat fiat areas or they may be sufficiently close to one another that their basins of attraction intersect yielding a complex super basin of attraction. Either of these cases eases could explain the broadly generalized nature of EN. Future research should attempt to map the basins of attraction to determine the frequency with which each of these possibilities exists. The present view of EN is consistent with Keane et al.'s (1985) view of emotional numbing as avoidance motivated and Litz ((1992) 1992) who argued that PTSD veterans retain the ability to feel fed normally. Neural networks can be cascaded to form a series of associations thereby introducing the possibility of connecting connecting associations. If an association is connected to a war-related memory, then any stimulus associated with the connecting memory will evoke the war-related memory. Future investigators should examine the topography of memory fields containing connecting associations to learn more about how such systems work. Rape/incest. Rape~incest. Emotional numbing is also characteristic of rape and incest victims. In the case ease of adult rape, sexual behavior that may have been previously associated with love and affection is traumatically traumatieaUy associated with fear, anger, and other strong negative emotions. Synaptic Synaptic weights undergo important changes resulting in observed symptoms. Because it is generally much easier to destroy than to build up, and because learning about life threatening events require one-trial learning to minimize fatalities, larger changes in synaptic synaptie weights may result from aversive than positive experience; a point for future research to clarify. For example, a woman develops into a normal adult over say 25 years as the result of many constructive experiences. Yet a single rape experience can compromise so much of what took so long to develop. One possible neural network mechanism is that memory formation is a relatively rapid process and that traumatic memories produce broad deep memory wells much as an intense explosion rapidly creates a large crater. This makes evolutionary sense in that memory for nearly fatal behaviors must be formed quickly and retained across the organisms life span so that this behavior is not repeated. Efforts to avoid recalling memories associated with rape, and the associated negative emotions, require avoidance of all relevant cues including normal sexual
w. W. W. W.
Tryon
83
situations with husband/lover which unfortunately strains an important of social support sometimes to the point of of separation or divorce and source of that occasions the onset of of more problems thereby further complicating her readjustment. of child sex abuse abuse and and incest, the young person, usually In the case of female, is much much smaller and weaker than the full grown, usually male, offending adult. When When the adult adult is also the child's parent, then the victim is dependent upon upon the the abuser abuser for emotional support as well as food, clothing, and shelter. Further trauma trauma is inflicted when a normally protective protective mother ignores her husband's child abuse. Parents can enlist police and other agencies to find and return their child should she or he run away. Lack of of education and financial resources further curtail the child's running away. These These conditions probably probably create large deep aversive memory wells with broad basins of of attraction attraction that that produce enduring psychopathology which warps the memory surface associating thoughts, feelings, and actions. that physical escape from ongoing abuse is not possible and that Given that the related aversive associations are unavoidable, dissociation becomes likely. might try to forget during traumatic memory forming experiences. One One might stimulus such as a ceiling light to method is to concentrate on an external stimulus reduce awareness of of the the traumatic traumatic events in progress thereby minimizing their present impact and and impairing memory formation regarding these events. More complete dissociation affords greater psychological protection. However, the reconstructive aspect aspect of of associative processes can create the full memory from a portion of of it. Partial cues, can sometimes evoke the entire traumatic memory. Hence, dissociative strategies are only partially effective unless extreme. DSM-IV (APA, 11994) 994) treats treats anxiety disorders as entirely separate separate from dissociative disorders. The above considerations indicate an important overlap between the two classes of of psychological disorder. Obsessive-Compulsive Disorders. Disorders. In the absence of of a DSM-IV (APA, 11994) 994) definition, we return to DSM-III-R (APA, 11987) 987) which defines obsessions as ""... . . . persistent ideas, thoughts, impulses or images ...." . . " (p. 245). All of of these symptoms are associative in nature and can therefore be addressed from the the neural network perspective. A prominent feature of of obsession is that that certain associations are highly repetitive. Neural networks can be autoassociative which means that Stimulus A evokes Stimulus B which can elicit Stimulus A and and repeat the the cycle or may recall Stimuli C, D D.. E E,, etc. before recalling Stimulus A and repeating once again.
84
Chapter 3
( 1 988) indicates that that the the continuous BAM BAM digs Kosko (1988) digs its own own energy perhaps initially reality based, might (memory) wells. Excessive worry, while perhaps of memory surface by virtue create an association which dominates a region of of an an excessively broad basin of of attraction and deep energy well. The broad of broad basin of attraction basin of attraction will cause many stimuli to evoke the same memory, association. Each obsession may have its own memory well and attractor basin. These memory wells may be widely spread across the memory surface of each other or they may reside sufficiently and therefore be independent of of attraction intersect forming a close to one another another that that their basins of complex super super basin of of attraction. Future investigators should attempt to determine the topography of of these attractor basins. An obsessive episode refers to a period of time during which associative period of processes operate operate within an abnormal abnormal attractor basin. The trigger stimuli locate the associative process within the relevant attractor basin. How one of the attractor basin is less obvious. One escapes from the influence of possibility is to suspend the associative process. This could be done through meditation aimed at clearing one's mind of of all thoughts. Or it could be done by initiating behaviors incompatible with thought such as reading aloud or singing. Another possibility is to remove oneself from the cliciting eliciting stimulus by leaving the situation. A third possibility is to initiate another line of of association such as engaging a cross word puzzle or mathematical or logic proof. This is the primary rational behind thought stopping techniques. Compulsions are . . . repetitive, purposeful, and intentional arc defined as ""... behaviors that are . . " (APA, 1987, p. arc performed in response to an obsession ...." 245). Obsessions set the occasion for compulsions. Treatments directed at compulsions are arc hypothesized hypothcsize~ to work because of their effect on associative processes. Compulsive disorders can be explained similarly to obsessive disorders by substituting Behavior A for Stimulus B such that Stimulus A sets the occasion for Behavior A which elicits Stimulus A which again sets the occasion for Behavior A, etc.. Hence, the autoassociative autoassociativc nature of the BAM can account for compUlsions compulsions in addition to obsessions. The Hebbian Hcbbian nature of of the BAM BAlM allows memory processes to to create ever deeper, and perhaps broader basins of attraction, through repeated association. This would allow normal worry to escalate into obsession through through excessive excessive reassociation rcassociation that that could set the occasion occasion for acting consistently with the obsession. Panic 3 Pamc Disorder. Panic attacks entail the the sudden sudden onset onset of at at least least 4 of 113 somatic or cognitive symptoms (APA, 1994, 1994, p. 395). Jones and Barlow (1990) report that that "Almost all patients presenting with with Panic Disorder have have a
IV..W IV. W
To, on Tryon
85
of negative life events preceding the first panic attack" (p. high incidence of 309). "Situationally bound (cued) Panic Attacks" and "situationally can be explained on the predisposed Panic Attacks" (APA, 11994, 994, p. 395) can basis that elements of the present situation recall memories and emotions associated with the initial trauma. of neural network memory systems is that that they can An important feature of retrieve a complete memory from a partial stimulus. While this property facilitates normal perception given degraded stimuli, i.e., partial views, it also provides a basis for psychopathology. The immune system sometimes misperceives allergens as disease and mounts an inappropriate attack. Hence, memory systems may misrecall certain memory content especially when given only partial information. An abnormal variant of this process might include of a few stimulus inappropriately recalling fearful material on the basis of elements. Perhaps this is what sometimes occurs in "situationally bound (cued) Panic Attacks" and in "situationally predisposed Panic Attacks" (APA 11994, 994, p. 395). Maybe one or a few elements erroneously elicits fearful of the BAM in combination with emotion. The pattern completion property of its well documented ability to associate responses with stimuli provides for of the the possibility that stimuli with only a partial similarity to those of traumatic incident may mistakenly elicit unprovoked aggressive behaviors in traumatized combat veterans. Panic disorder may be explained by a theoretical connection with obsessions. We know that that Hebbian learning mechanisms, and their BAM equivalent, generate, synthesize, increasingly deep memory wells and associated basins of of attraction as the associative process repeats. This allows an obsessive associative process to dig a large energy well which then functions as a memory. Obsessions about traumatic events may therefore create the functional equivalent of of traumatic memories. Such abnormal memories may elicit strong anxiety when accessed. Alternatively, there may be nothing unpleasant about the synthetic memory basin but its development might intersect with one or more basins of of attraction associated with legitimately fearful memories thereby providing passage for for the locus of of memory recall from the synthetic synthetic basin to one or or more another basins associated with with legitimately anxiety provoking memories. Specific Phobia. Phobia. Formerly called Simple Phobia, Specific Phobias entail Specific fear cued situation (APA, 1994). cued by by aa specific object or or situation 1 994). If If their phobia phobia was was created by aa traumatic event, event, then then memory for this event occupies the the energy minimum, bottom, of of the the memory well. If If no such event can be recalled, then memory for it may have been dissociated.
86
Chapter 3
The The breadth of of the basin of of attraction determines the generality of of the disorder. A single highly circumscribed simple phobia is modeled by a single of attraction. Multiple deep memory well with a relatively small basin of phobias involve multiple wells which may either be in separate areas of of be sufficiently close to one another that their basins of of memory or they may be attraction intersect to create a complex super basin. Social Phobia. Phobia. The same comments apply here as to Specific Phobia that the feared object is a social situation. with the qualification that Generalized Anxiety Disorder. Disorder. The DSM-IV inclusion criteria for (GAD) include excessive worry for at least six Generalized Anxiety Disorder (GAO) months over concerns that are not limited to a particular object, setting, event or medical condition that the person finds difficult to control. It seems unlikely that that such diffuse anxiety can can be explained on the basis of of multiple traumatizations. Partial pattern completion is one possible explanation of of GAD. Perhaps persons with GAD have been traumatizexl traumatized but the memory completion process is both partial and limited to affective affcctive components. This of ordinary stimuli would elicit anxiety without would explain why a variety of of cognitive the ability to describe the basis for feeling anxious. The nonrecall of components would preclude being able to say why they felt anxious. The conditions under which partial pattern completion can occur, if it can occur at all, are presently unclear. Another explanation is based on the fact that CNNs can form spurious memories for events that were never experienced. These are composite memories derived from two or more nearby memory wells. If If a spontaneous memory develops in an area popUlated populated by unpleasant and fearful memories, then a spurious nonspecific fearful memory might result. Hoffinan 987, Hoffman (1 (1987, 1992), Hoffinan and Dobscha ((1989) 1989) and Hopfield, Feinstein, and Palmer Ho~ ((1983) 1 983) describe the formation of of parasitic, spurious, memories as a consequence of of normal memory formation. Uncertainty remains regarding the clarity such memories may have. These memories may be clear or they may er of be diffuse. The haphazard mann manner of their formation suggests that it is more likely that memory formation is fuzzy than clear. If these memories entail anxious content, then stimuli able to recall such memories can explain the presence of of diffuse anxiety. It is entirely possible that the so-called False Memory Syndrome is based on spurious memories meaning that the memories derive from the interaction of of memory wells and not from actual experIence. experience. Hopfield et al. ((1983) 1 983) have demonstrated that unlearning spurious memories improves the normal memory function of of Hopfield CNNs. Crick
W. W. W. w.
Tryon
87
and Mitchison ((1983) 1 983) hypothesize that dreams entail unlearning. If a problem then spurious memories would be retained exists with this purgative process then despite repeated dreams about their content thereby explaining the persistence of diffuse diffuse anxiety anxiety over over time. time. of of Anxiety Disorders Disordersl: Current cognitive-behavioral Cognitive-behavioral Neuropsychology of theories maintain that emotions, such as anxiety, are a consequence of thought (frontal lobes). Emotions are believed to occur because of attributions 962, 11980). 980). Cognitive attributions we we make; make; what what we we tell tell ourselves ourselves (Ellis, (Ellis, 11962, Cognitive psychotherapy of emotional disorders largely entails altering irrational attributions and beliefs. Segal and Blatt's ((1993) 1 993) contributors consistently recommended that affect should be better integrated into cognitive/models of psychopathology.. psychopathology Neuropsychological evidence regarding the formation of emotional memories, like those associated with anxiety, phobias, panic attack, and cognitiveposttraumatic stress disorder, indicate that major changes to cognitive behavioral theories need to be made. Subcortical pathways play an important role in the formation of emotional memories (LeDoux, 1994) 1994) that are not considered by by contemporary contemporary cognitive cognitive theories. theories. Neural Neural network models, models, such such considered as the BAM, readily lend themselves to accounting for both cortical and subcortical associations. LeDoux ((1994) reviews neuropsychological neuropsychological research, research, spanning spanning at at least least LeDoux 1 994) reviews the the last last decade, decade, that that clarifies the the physiological phYsiological basis basis of of conditioned conditioned fear. fear. Aversive stimuli stimuli inform inform the thalamus thalamus which jointly jointly informs informs the the lateral lateral nucleus nucleus Aversive of the the amygdala and the the cortex. The The lateral nucleus nucleus of of the the amygdala amygdala directly directly of of the amygdala which initiates communicates with the central nucleus of physiological and behavioral changes via via the the brain stem. stem. The The lateral lateral nucleus nucleus physiological communicates indirectly with the central nucleus through the accessory basal nucleus and the basolateral nucleus of the amygdala. This subcortical system a more more immediate immediate response. response. is shorter than the cortical route and provides a of the amygdala as does the The cortex informs the lateral nucleus of thalamus but with with higher higher resolution, resolution, more The thalamus but more fully fully processed, processed, information. information. The cortex also informs the hippocampus which communicates with the lateral nucleus of of the the amygdala. amygdala. The The only only site site where where lesions lesions can can be be made made without nucleus interfering with the ability to learn conditioned conditioned emotional responses is the If subjects subjects are are conditioned conditioned while while intact intact and and the the cortex cortex lesioned lesioned cortex. If conditioned emotional response is disrupted. disrupted. However, this subsequently, the conditioned effect appears to be because the lesions interfere with with long-term memory retrieval since the conditioned conditioned emotional emotional response partly recurs when reminder reminder cues are presented. presented. The same brain pathways appear to be involved
88 88
Chapter 3
in all all m mammals and possibly possibly in in all all vertebrates. vertebrates. Many Many of of these these species species are are not not ammals and in
known ffor cognitive processes processes and and all of them them do do not not have have known or their their higher higher cognitive all of
language nor do they they make attributions. language nor do make attributions.
leDoux 1 994) reports LeDoux ((1994) reports that that emotional emotional memories memories are are long long lasting lasting and and the the
result result
of of long-term long-term potentiation potentiation
neurotransmitter neurotransmitter
glutamate. glutamate.
This This
(L TP) (LTP)
of of NMDA NMDA
synaptic synaptic
receptors receptors
alteration alteration
causes causes
to to
the the
larger larger
postsynaptic responses neural signals. postsynaptie responses to to the the same same neural signals. The above above comments comments are are consistent consistent with with the the BAM BAM in in two two important important The ways. First, the BAM, like all connectionist neural networks, learns ways. First, the BAM, like all connectionist neural networks, learns by by changing changing synaptic synaptie weights weights which which is is what what LTP LTP entails. entails. Second, Second, the the BAM BAM is is equally equally capable capable of of associating associating emotions emotions with with stimuli stimuli and and emotions emotions with with emotions as is in emotions as it it is in associating associating stimuli stimuli with with responses responses and and stimuli stimuli with with other other stimuli; the the BAM a general associative mechanism. mechanism. Hence, it is is possible possible stimuli; BAM is is a general associative Hence, it for stimuli to directly directly cue cue emotions. emotions. Because Because neural neural network network models models derive derive f or stimuli their properties by simple neuron-like their functional functional properties by interconnecting interconnecting simple neuron-like elements elements and and changing connection connection (synaptic) weights as as aa result result of of experience, they can can be be changing (synaptic) weights experience, they directly informed by by advances advances in in the neurosciences. Traditional Traditional cognitive cognitivedirectly informed the neurosciences. behavioral theories theories have have largely largely ignored ignored neuroscientific neuroscientific evidence evidence because because they they behavioral f ocus exclusively focus exclusively on on psychological psychological processes; processes; they they do do not not have have the the bridging bridging properties inherent in neural properties inherent in neural network network models. models.
Mood disorders Depression Depression
involves involves
repetItIve repetitive
self-deprecating self-deprecating
associations associations
of of
hopelessness, hopelessness, uncontrollability, uncontrollability, and and worthlessness worthlessness (excessive (excessive or or inappropriate inappropriate
guilt), among among other including psychomotor psychomotor retardation retardation or or agitation, agitation, guilt), other symptoms symptoms including
insomnia insomnia
or or
hypersomnia, hypersomnia,
difficulty difficulty
concentrating, concentrating,
thinking, thinking,
making making
decisions, social events. decisions, and and reduced reAucexi participation participation in in occupational occupational and and social events. At At
least three causal hypotheses exist regarding least three causal hypotheses exist regarding the the relationship relationship between between
cognition first possibility possibility is is that cognition and and depression. depression. The The first that cognitive cognitive style style causes causes
depression. depression. A A second second possibility possibility is is that that cognitive cognitive change change is is part part of of the the
depressive depressive disorder disorder and and emerges emerges simultaneously simultaneously with with other other depressive depressive
f eatures. A features. A third third possibility possibility is is that that cognitive cognitive changes changes occur occur as as aa consequence consequence of depressed. I argue for the first, and of being being depressed. argue for the first, and especially especially the the second, second, and and
against against the the third third possibility possibility in in the the remainder remainder of of this this section. section.
The The first first etiological etiological possibility possibility is is represented represented by by Abramson, Abramson, Seligman Seligman
and 1 978) and and Teasdale Tcasdale ((1978) and Seligman, Seligman, Abramson, Abramson, Semmel, Scmmcl, and and von von Baeyer Bacycr
who describe a dcpressogcnir attributional style style (DAS) (DAS) where where persons ((1979) 1 979) who describe a depressogenic attributional persons
invoke invoke internal, internal, stable, stable, and and global global explanations explanations of of negative negative events, events, and and to to aa
W. W W. W
Tryon
89
of positive events lesser extent, external, specific, and unstable explanations of as causally relevant to the onset of depression. Causal relevance does not factor imply either a necessary or sufficient condition but only that a risk fact()r exists that can range from small to large. Beck ((1967, 1 967, 11976) 976) and Beck, Rush, Shaw, and Emery ((1979) 1979) argue that psychosocial stress precipitates depression in persons who already emphasize negative outcomes, over generalize, magnify the importance of negative events, and think in absolute "all-or-none" terms. of this hypothesis is that the pattern of of synaptic weights A CNN view of that instantiates the depressive attributional style is consistent with, and that partially implements, the pattern associated with depression but is insufficient to give rise to a diagnosable mood disorder. However, the more intense the attributional style, the closer the synaptic weight pattern is to a depressed state. That depressive attributional attributional style is a risk risk factor for and not a That of depression depression is demonstrated demonstrated by the fact that that physical necessary condition of of a loved one, marital illness, alcohol and cocaine dependencies, death of separation and/or divorce, and childbirth can elicit a Major Depressive separation 221). broader range of of premorbid Episode (APA, 1987, p. 22 1). Hence, a much broader personalities than those characterized by a DAS can become depressed. may also change synaptic weights so that they converge towards These events may a depressive state. The second etiological hypothesis is that The that cognitive and emotional changes that consequence that characterize characterize depression depression emerge simultaneously simultaneously as a consequence of synaptic weights. I base this hypothesis on the of aa particular particular pattern pattern of of synaptic theoretical jointly theoretical position position that that the functional functional attributes attributes of of a CNN are jointly dependent upon its architecture of synaptic states (connection architecture and pattern pattern of weights) across the network. I conclude that architecture is not critical to that architecture depression depression because because virtually virtually all depressed persons have aa history of of normal emotion and cognition as children and perhaps as adolescents and some portion of of adulthood adulthood prior to the onset of of depression. Moreover, depression depression remits either either naturally consequence of naturally or or as as aa consequence of treatments. treatments. Several investigators investigators (Dobson (Dobson & & Shaw, 1986; 1 986; Eaves & Rush, 1984; 1 984; Hollon, Kendall, & Lumry, Lumry, 1986) 1 986) have have reported reported an increase increase in negative thoughts during during depression depression plus aa return return to to normal normal when when depression depression remits. Changes in neural neural architecture probably do not architecture not occur occur these cases. Hence, depression depression probably probably results synaptic weights. It is these results from from the the pattern pattern of of synaptic these connection connection weights that that determine determine the the graphs graphs showing the structure structure of of the basins basins of of attraction attraction associated associated with with memories.
90
Chapter 3
The present CNN hypothesis about the etiology of of depression assumes that that cognition and emotion are encoded as a single vector thereby integrating intellectual and affective content as described above. It is hypothesized here that depression warps the memory field such that basins of of attraction associated with normal memories that entail both positive and negative affect now lie within a super basin that leads to a global minimum associated with negative affect. This condition causes stimuli associated with normal memories to consistently return negative affect. Put otherwise, the part-whole pattern completion CNN property now returns negative emotion most of of the time. This memory effect has the following hypothesized causal consequences. That many different stimuli (situations) return negative affect leads to a global attribution. That the pattern completion process is highly stable attribution. The absence of replicable leads to a stable of precipitating external events, or the over reaction to events such as personal loss, leads to an internal attribution. The uncontrollable nature of of the pattern completion hopelessness. Hence, the cognitive distortions associated process leads to hopelessness. with depression are a consequence of of a distorted memory process associated with a broad and deep basin of of attraction. A consistent preexisting DAS means that the network is already partially trained toward a depressive configuration and therefore requires less change to reach a state sufficient to affeetive characteristics associated with a clinical give rise to cognitive and affective of Major Depression. DAS is therefore a risk factor as diagnosis of acknowledged above. However, these changes can occur in anyone following a depressogenic event therefore explaining why DAS is not a necessary condition. Storing emotion and cognition in the same vector fuses these two aspects of experience into the same memory. The part-whole pattern completion of of CNNs means that that an emotional stimulus, or partial stimulus, can property of recall other emotional elements and all cognitive elements associated with that memory thereby explaining state-dependent (mood-congruent) learning and & Campbell, 11992). recall (Matt, Vazquez, & 992). Depressive mood may also steepen memory well walls thereby reducing the time taken to recall such (el. Blaney, 11986; & Cohen, 11982; 1984; Teasdale & & material (cf. 986; Bower & 982; Isen, 1984; 979; Williams, Watts, Macleod, Fogarty, 11979; MacLeod, & & Mathews, 1988). The third etiological assumption is that negative affect is a consequence of, comes after, afLer, depression. Lewinsohn, Steinmetz, Larson, and Franklin data showing that cognitive distortions are a ((1981) 1 98 1) provide longitudinal data of depression. The absence of of accurate data consequence, rather than a cause, of of depression and onset of of cognitive regarding the delay between the onset of
145.W. w.
Tryon
9911
distortions allows for the possibility of simultaneous emergence and therefore compatibility with the second alternative discussed above. A major problem two separate processes with the sequential hypothesis is that it presumes two the first (emotion, depression) influences the second second (cognition) (cognition).. This where the of how one process influences the other. This problem is raises the problem of of the mind-body problem where the distinction between mind reminiscent of of how they interacted. Mind as emergent from and brain created the problem of brain "solves" the problem by not making the initial distinction. I suggest that the same lesson applies to the cognitive and affective aspects of depression. All treatments are hypothesized here to exert their effects by altering synaptic weights. Pharmacological treatments attempt to directly alter synaptic function. Kandel ((1991) 199 1 ) has shown that learning entails long term synaptic change due to the synthesis of new proteins under genetic control. Hence, psychological treatments entail biological changes. That both treatments have a common effect makes it understandable why combined intervention often works best. This view should make psychologists more understanding of of and respectful toward drug treatments and psychiatrists understand of of and respectful toward toward psychological treatments. For more understand of Beck's cognitive-behavioral treatment can be example, the effectiveness of explained on the basis that that experience alters synaptic weights. Therapeutic experiences, both both in in the the office office and and during homework, homework, may may alter alter the the synaptic synaptic experiences, weights in in ways ways that that normalize memory processes. processes. As As another another example, example, weights Bellack 1 985) reported that pharmacologic therapy normalizes depressive Bellack ((1985) depressive cognitions as effectively as cognitive therapy. therapy. Long Long term term success in in either cognitions case depends upon upon the durability of of synaptic change. Segal and and Blatt's (1993) the integration of of affect Segal ( 1 993) contributors called for the into cognitive/models of of psychopathology. Encoding emotion and cognition in of human the same vector seamlessly integrates these two important aspects of experience. It was previously observed that traumatic events, such as rape, exert aa large negative effect in aa small amount of affective of time. Perhaps affective intensity exerts strong influence over the formation of of memory wells thereby creating aa dominant, controlling, influence for aa long time. It makes evolutionary sense that that memory formation would be much more potent potent for potentially lethal events than positive ones. Forgetting aa potentially lethal than for positive experience could result in death death if if repeated repeated even once whereas forgetting aa positive would not not have this effect. positive encounter encounter would
92
Chapter 3 Treatment Treatment
Learning therapy Psychotherapy and behavior therapy share aa common assumption assumption that that therapeutic change entails new learning; theorists differ mainly in what what they learned. Some therapists therapists emphasize the role of of reinforcement believe is learned. new attributional attributional styles or contingencies while others accentuate learning now other information processing strategies, but but all agree that that therapeutic changes are learned; learned; otherwise all but but purely biological treatments would be ineffective and pointless. Neural Neural networks are learning learning mechanisms and causal explanations for how infrahuman infrahuman and consequently provide proximal causal human human learning can can occur (Tryon, 1995b). This scientific base provides a a unified perspective capable of the disparate treatment techniques unified of organizing the associated with behavior therapy. I have previously termed this general connectionist approach to learning as Neural Network Learning Theory 1 993b). (NNLT: Tryon, 1993b). Computer models of of memory formation and alteration, such as the BAM, can serve the same heuristic function as animal models of of psychopathology; they provide aa well controlled context in which to study psychopathology, therapeutic principles. This is not to say that that animal research will be replaced by computer simulation but rather that that computer simulation can be used as a productive new tool for evaluating hypotheses. Eysenck ((1964) 1964) maintained that "Behaviour "Bchaviour therapy may be defined as the attempt to alter human human behaviour bchaviour and emotion in a beneficial manner modem learning theory" (p. 1). Wolpe and Lazarus according to the laws of of modern ((1966) 1 966) agreed that . . . the application of that behavior therapy entailed ""... of experimentally established principles of of learning" (p. 1). l). However, Kazdin ((1978) 1 978) noted that that "The definition of of behavior therapy has been broadened, and the role of of learning theory has been reduced substantially to the point that the precise role of of learning theory in actual practice of of behavior therapy has been questioned" (p. 1195). 95). Kazdin ((1979) 1979) criticized the claim that behavior therapy was based on modem learning theory was a fiction. Spiegler and Guevremont ((1993) 1 993) argued that the statement, "Behavior therapy is the application of -established laws of learning," is currently "predominantly of well well-established false" (pp. 4-5). Connectionism in the form of NNLT provides a learning learningmemory base for all psychological therapies, including behavior therapy. The full spectrum of behavior therapies can can once again be said to rest on CNNs as modem learning theory. Whereas traditional animal models of
W. W W. Tryon Tryon W
93
as too limited to to form a general framework for psychopathology are seen as behavior therapy, neural networks have the required broad relevance to both animal and human perception, cognition, memory, and behavior. Three corollaries of learning can be distinguished from from our neural Three network perspective. First is that new learning, information, is linked with previous knowledge. Developmental processes, such as cumulative (of. Staats, 11986), 986), are good examples. The BAM hierarchical learning (cf. mechanism discussed above illustrates one mechanism for how these changes can occur. Connectionist systems blend knowledge sources because multiple associations are encoded by modifying the same set of synaptic weights. This of therapy entails new learning to provide missing or inefficient aspect of interpersonal and other skills required for "normal" performance. A second corollary of of learning therapy is unlearning; attempting to undo of prior learning that has caused the person problems. Systematic the effects of desensitization is a good example. A new, and more normal, response is learned to stimuli and events that that previously produced fear and avoidance. The BAM model provides a possible mechanism for unlearning (forgetting). complement. For example, Any specific memory can be deleted by storing its complement. if stimulus complex 11,1,0,0,1,1 if , 1 ,0,0, 1 , 1 has been stored then all memory for it can be 0,0,1,1,0,0. 1 , 1 ,0,0. Put more informally, learning that erased by storing stimulus 0,0, dogs are kind affectionate house pets removes the association that they are cruel vicious wild animals. The relevant substitutions are kind-cruel, experience, verbal affectionate-vicious, house pets-wild animals. Personal experience, association, and/or observation all function to store inverse memories thereby ameliorating pathological memories/associations. Alternatively, unlearning can be accomplished by storing a new memory formed by associating the opposite characteristics. It follows that therapeutic efficacy is directly proportional to the extent extent to which all elements of of a stimulus complex have been addressed and is directly proportional to to the degree to which each element can be fully inverted. It may only rarely be possible to completely but may well suffice to bring erase clinically relevant memories in this way but much appreciated appreciated relief. Since memory formation entails generating generating aa well of of some depth depth with aa basin basin of of attraction of of some breadth, it follows that that memory removal by storing the the exact exact opposite association reduces both the the depth depth and and breadth breadth of of to zero. It also follows that that storing intermediate associations the memory well to should should reduce both both the the depth and and breadth breadth of of memory wells. Perhaps the reversal effect is aa nonlinear function of similarity to an of an exact complement. Normal forgetting could be the result of be seen seen as the of incidental complementarity
94
Chapter 3
as a result of of new memory formation. This makes forgetting an active rather than passive process. The third corollary of of learning theory is that the pattern completion property of of the BAM allows one to facilitate memory recall in at least two ways. The first method is based on the fact that content addressable connectionist neural neural networks, like the BAM, are capable of of recalling a entire memory given only a memory fragment. Hence, by systematically focusing on partial sights, sounds, tactile and olfactory sensations, and kinesthetic cues, more complete memories often form. The second procedure proeexture for facilitating memory entails mood-congruent (state-dependent) recall (Matt et aI., al., 1992). This method is based on the finding that affective state influences memory recall. Happy memories are more available when one feels cheerful and sad memories are more available when one feels depressed. Affect appears to be stored with content and consequently, content content which may not be readily available for recall may be remembered in the presence of of affective cues. Both Both methods can can be combined by first first inducing an emotional state, or capitalizing on a naturally produced one, and presenting specific visual, auditory, olfactory, tactile, and/or kinesthetic cues. Old photographs taken during childhood and/or other memorabilia may be used. Unconscious processes. Experimental evidence has established that unconscious (implicit) learning (Seger, 11994) 994) occurs, though in a much more 992; Jacoby, Lindsay, & limited way that Freud suggested (Greenwald, 11992; Toth, 1992; Kihlstrom, 1987; Kihlstrom, Barnhardt, & Tataryn, 1992; Toth, Schacter, 1987). For For example, implicit memory is demonstrated through the "savings" technique where material once learned, but forgotten, can be relearned in fewer trials than than it initially took to learn. Savings result when the connection (synaptic) weights associated with the hidden middle units retain values close to what they were when the behavior was fully learned. leamed. Fewer weight adjustments are required to satisfy the performance criterion used to demonstrate learning thereby resulting in savings. Serial information processing models have never satisfactorily explained how semantic analysis is possible for information only partially perceived. Kihlstrom ((1987) 1987) indicates that parallel distributed neural networks solve this problem because: 1) unconscious processes can occur within the "hidden" network layer (cf. (el. Greenwald, 1992), 2); no central processing unit is involved, 3) there are no rules to be aware of, 4) partial effects can be exerted, and 5) neural networks networks merge perception and cognition thereby allowing unconscious processes to exert their influence (cf. (of. Greenwald, 1992). Learning occurs because the connection (synaptic) weights change.
W IV..W IV.. Tryon Tryon
95 95
The same mechanism accounts accounts for unconscious learning. The absence of of a The rule-governed ruleogovcrned central processing pro~ssing unit unit (self) (scl0 makes it unnecessary to explain unconscious learning differently from from normal learning. That That learning takes unconscious regardless of of whether whether the subject subject is aware aware or not place in the same way regardless makes the distinction distinction pointless thereby rendering rendering the conscious conscious vs. unconscious unconscious schism a nonissue. nonissuc. This is another example of of how how neural networks networks produce produce theoretical theoretical synthesis synthesis by by making making problematic problematic schisms schisms like like Tryon, 11993a, b), previous previous problematic problematic schisms become become nonissues nonissucs 33.. (cf. Tryon, 993a, b); Greenwald 1988) explains Grccnwald ((1988) explains self-deception as avoidance avoidance based based on partial perception. Perceiving Perceiving a subset of of cues may be bc sufficient to occasion perception. avoidance behavior behavior even in the absence absence of of full awareness awareness of of the underlying avoidance associative process. process. Greenwald Grccnwald ((1988) repression as self-deception associative 1 988) explains repression based on memory versus current current perception; avoidance avoidance behavior behavior in response response based to a subset subset of of recalled recalled associations. associations. Neural Neural networks networks can implement implement both both of of functions. these functions. The memory memory energy function function (cf. equation equation 5 in the the Appendix) Appendix) suggests The another view of of repression. repression. Positive Positive energy another energy states states result in energy energy hills with basins of of repulsion repulsion just just as negative negative energy energy wells have have basins basins of of attraction. attraction. basins Repulsion Repulsion results results because because memory memory recall recall entails entails seeking seeking an an energy energy minimum. minimum. Just as does not not flow flow up up hill on its own, so also also do associative associative Just as water water does
33 Tryon Tryon (1993a,b) ( 1 993a,b) discusses other instances instances where problematic schisms schisms become body issue. Terms like psychosomatic psychosomatic and somatopsychic somatopsychic nonissues. One is the mind vs. body issue. Terms attempt to explain how the mind can effect the body body and how the body body can can effect effect the mind. CNNs provide existence proofs that mind emerges emerges from from bodily networks, that two separate they interact. entities do not exist, and therefore it is pointless to talk about how they interact. The behavioral vs. cognitive debate is also resolved by CNNs. CNNs. CNNs CNNs are cognitive cognitive models developed to study study the the microstructure of cognition. cognition. Donahoe Donahoe and Palmer (1989) ( 1 989) and Tryon I 996c) have shown shown that connectionism connectionism is completely ( 1 993b, 1995c, 1 995c, 1996c) (1993b, completely consistent with operant behaviorism. Consequently, connectionism of the cognitive connectionism is theoretically synthetic of behavioral debate in the Hegelian sense of combining thesis and antithesis into It debate into synthesis. synthesis. It is a single perspective consistent with two seemingly contradictory perspectives. Consequently this debate is now moot. third schism is between human and animal moot. A third research. These areas of research employ different different vocabulary, vocabulary, concepts, and do not cite each others work. Tryon (1995b) work. Tryon ( 1 995b) shows that connectionism applies equally well to animals humans. One vocabulary and set of concepts is applied across the phylogenetic phylogenetic scale. and humans. This list of theoretical syntheses is not exhaustive.
96
Chapter 3
flow up up hill. Recalling Recalling such such memories memories would require require an energy processes not flow increase e ~ of increase to to reach reach the the p peak ofthis inverted inverted memory well. Unremembered Unremembered traumatic traumatic events can can influence both memory and behavior. Janet ( 1 893, 1904) was perhaps the first first to to describe describe cases cases of of behavior. Janet (1893, 1904) was perhaps the hysterical (traumatic) amnesia in which frightening experiences were both unremembered and influential. He explain~ explained these disorders disorders as dissociation dissociation of of memories memories by emotion (see above comments on dissociative dissociative disorders). disorders). learning. Unlike FFreudian extraordinary New learning. rcudian theory which gives extraordinary experience, neural networks suggest that new learning learning influence to early experience, continues to change change synaptic continues synaptir weights throughout the life span. However, the accumulation of of prior prior learning learning and the normal developmental decrease in neurotransmitters probably reduce plasticity over time. Psychotherapy neurotransmitters equates to methods for producing new learning to correct and/or compensate of prior learning. for the effects effects of therapy. Learning Individual therapy. Learning can occur through experience with the both the informational of what the therapist therapist. This includes both informational content of says, the therapist's calm reassuring reassuring voice, and the therapist's relaxed relaxed posture posture plus other accepting/reassuring nonverbal cues. Thought stopping procedures enlargement can be prescribed on the basis that that they inhibit the memory well enlargement described in connection with generalized anxiety and obsessive compulsive disorders above. New New associative strategies strategies may be found found to shrink memory wells. Specialized therapeutics may be necessary f or clients who have for dissociated traumatic experiences. The phenomenon of of mood-congruent recall (Matt et al., 1992), 1992), also known as state-dependent state-dependent memory, suggests a possible approach. Whereas normal memory is biased toward recalling positive events, depression depression alters this bias toward negative events, in direct proportion to the severity of of depression. Because experimentally experimentally induced dysphoria also negatively biases memory recall, the same or similar procedures might be used to temporarily augment depression acilitate depression to ffacilitate recollection recollection of of traumatic experiences. Care should be taken to avoid recalling too much during a single session and thereby further traumatizing the person. This would include include terminating a memory search search after one or two items were recalled in order not to overwhelm the subject. Each recollection would need to be examined in the session during which it was retrieved. rctrieve~. therapy. Learning ~ r n i n g can occur occur in group settings where clients Group therapy. interact with other people experiencing similar problems. The redundant credible information provided by group members should facilitate facilitate the learning learning process. process.
W. W. W. w.
Tryon
97
Token economies. Bellack ((1986) 1 986) documented the effectiveness of of token economies for schizophrenic patients. The therapeutic efficacy for this population can be explained by Hoffinan's 1 987, 11992) 992) observation that Hoffman's ((1987, Hamming memory energy flow over a significant H amming distance (see Appendix) is vulnerable to attraction by large deep deep energy wells associated with the disorder. The clarity and consistency of contingent stimulus-consequence stimulus-consequence sequences associated with token economies might help patients with sufficient reality focus to function more normally. This schizophrenia retain sufficient formulation explains why they remain dependent upon highly structured environments; a behavioral prosthesis. Life experience Life experience.. Learning can occur though personal experience outside of therapy settings. This experience can be in the form of carefully crafted exposure therapy guided by prior hierarchy construction or it can occur spontaneously. spontaneously. Pharmacotherapy
Basins of of attraction are calculated on the basis of synaptic weights and consequently anything that modifies synaptic weights also modifies basins of attraction. Three arguments implicate synaptic change as the basis of therapeutic improvement. First, cognitive change is a form of learning and learning has been shown to entail synaptic change (Donahoe & 994; & Palmer, 11994; Thompson, 11986, 986, 11990). 990). Second, pharmacotherapy influences neuro neurotransmitters such as dopamine which are known to influence synaptic function. Third, Bellack ((1985) 1 985) reported that pharmacologic therapy alone normalizes, or changes, depressive cognition as much as does cognitive therapy. Because the two primary variables affecting the BAM network are depth of the memory well and breadth of of its basin of attraction, it follows that the psychological and behavioral effects of pharmacotherapy result from altered synaptic function causing changes in either the depth of the memory well and/or breadth of its basin of of attraction. Because shallower and narrower are associated with normal, I hypothesize that pharmacotherapy reduces the of BAM memory wells and diminishes their basins of of attraction. depth of The primary theoretical means of of decreasing the energy values associated with stored memories is to work backwards, using Equation (5) from the Appendix (E = -SMRT T)) from energy (E) to memory matrix elements (M). The S and R vectors remain as they are because they describe external will realize this stimuli and responses. After reading the Appendix, the reader will =
98
Chapter 3
entails entails minimizing the the SM SM multiplication which is done by by minimizing the the elements in the the M M matrix. matrix. This This partly partly depends depends upon upon how how many of of the the component component products products are are of of the the same same sign. sign. More More same signed products products will create aa larger larger sum sum whereas whereas products products of of different signs will tend to cancel cancel out out create tend to leaving aa smaller sum.
Electroconvulsive therapy (ECT) (ECT) therapeutic effects of of ECT EeT can understood from aa neural network The therapeutic can be understood perspective in one of of two ways. The The first explanation entails the assumption assumption that depression results from aa deep memory well with aa large large basin of that of attraction and that ECT attraction and that EeT produces produces aa temporary energy increase like that that utilized when training training Boltzman Boltzman neural neural networks using "simulated annealing". Annealing is a metallurgic process used to remove internal stresses and The process begins by heating, energizing, the toughen metal and glass. The substance to a point above its melting temperature such that the atoms are in substance violent random motion. As the temperature is slowly lowered, the atoms gradually form aa crystalline structure which forms aa collective energy minimum. Simulated annealing is aa method of of escaping unproductive local energy minima to promote convergence on the global energy minimum corresponding to problem solution or optimal performance (cf. Wasserman, 1989, 1 989, pp. 7783). Neural networks trained by annealing contain a "temperature" parameter of the network being in their learning function which controls the probability of in a particular particular energy state. In the beginning, all energy states, including high ones, are essentially equally probable. A preference develops for lower energy states as temperature temperature decreases but the possibility of of temporarily jumping to state remains. That That the system sometimes increases its energy a higher energy state state before continuing its gradient descent toward an energy minimum often enables it to escape from aa local energy minimum and move toward a global energy minimum. ECT may provide the temporary energy increase needed to escape a EeT ECT may local energy minimum associated with a depressive memory well. EeT of memory recalVformation recall/formation to another area of the memory propel the locus of field. The number of sessions required for clinical effectiveness may depend upon the depth of of the memory well and the diameter of its basin of of attraction. ECT sessions may be due to the nondirective nature the The need for repeated EeT treatment. Increasing energy state does not necessarily project the locus of recall/formation in a predictable direction; its trajectory is likely to memory recalVformation
W. W W. W
n Tryon Tryo
99
the locus of memory be random. The energy jump may relocate the recall/formation to a point still within the current basin of attraction but through repeated efforts may escape this basin. Perhaps a more focal and/or phased application of energy would produce more specific and controllable effects. effects. of this hypothesis is that it assumes that the energy well A limitation of remains just as deep and wide as before and could subsequently reattract memory given prior stimulus conditions. This problem can be solved by postulating a second therapeutic effect. Perhaps ECT directly modifies the structure of the memory well; making it more shallow and narrow. Memory wells are deep by virtue of large negative energy values. ECT may make all memory energy levels more positive, in which case the well retains its relative depth. This process would effect all memories equally. A more therapeutic assumption is that energy is absorbed in direct proportion to the pre-existing negative energy level. Consequently, the most negative regions, deepest wells, of energy thereby reducing them the most. The absorb the greatest amount of of the attractor basin thereby same process may also reduce the diameter of of a relatively permanent cure. explaining the possibility of ECT produces amnesia for events just just prior to treatment. This ECT understood from the neural network perspective as phenomenon can be understood disrupting the outer product matrix multiplication and subsequent matrix addition involved in creating long term memory storage and representation across the network (see Appendix). In so far as the patient is experiencing depressive symptoms symptoms at at the time of the underlying associative of treatment, the process will be disorganized. Repeated disorganizations of of a depressive complex may normalize the the associative associative process. It may therefore be beneficial to have the the person focus on on their most depressing associations associations immediately prior to administering ECT.
Research Strategies Strategies Research Assessment Assessment I am any current mapping the the breadth am unaware unaware of of any current methods for mapping breadth and and depth depth of of memory wells. Such aa technology needs to be developed because these two theoretical theoretical constructs constructs are critically important important and empirical tests of of this model are dependent upon are dependent upon such such assessments. Perhaps Perhaps the the first step step is to to exhaustively survey all existing existing methods of of memory assessment assessment to to determine their suitability for mapping mapping basins basins of of attraction attraction and and depth depth of of memory wells.
100 1O0
Chapter 3
All methods of of measuring associative strength are potentially relevant to assessing the basin of of attraction and possibly depth of of memory wells. wells. Clinical methods for ascertaining which events trigger which memories should also be considered.
Treatment Treatment
The hypothesis that phobics have large basins of of attraction, compared compared to normals, can be evaluated, in principle, by applying the assessment techniques to be developed to phobic and control sUbjects. subjects. The hypothesis that successful successful treatment shrinks the basin of of attraction can be evaluated by comparing phobic subjects before and after treatment with empirically prc and post assessments with validated procedures and comparing the pre untreated control subjects. Consideration was given above under the heading of of pharmacotherapy pharmacothcrapy of memory wells could be reduced by altering the as to how the depth of of this memory matrix so that smaller energy values resulted. The brevity of discussion partly reflects a lack of of study of of this issue. No theoretical conjecture or empirical evidence could be found relating learning based or any other psychological intervention to changes in the BAM or any other CNN. Perhaps readers of of this chapter will have, and publish, additional ideas on this topic. The most that can be said at this juncture is that the mechanism of of therapeutic therapeutic action for learning based therapies will very likely be the same as for pharmacotherapy. This conclusion has an important implication for psychologists and psychiatrists (pharmacologists) especially when working together in medical settings. Connectionism emphasizes the compatibility of of learning and pharmacological therapies; both of them are directed at synaptic change. Pharmacological agents rapidly change synaptic function but these changes may be temporary in which case relapse occurs when medication is removed. Learning based approaches to synaptic change often take longer but can produce more lasting change. The value of using both therapeutic methods concurrently is obvious. This view should make psychologists more understanding and supportive of of psychopharmacology. Neuroleptic medications are making similar psychological changes to those produced by learning based therapies. I cite the changes in cognitive style that accompany remission (cf. 986; Eaves & Rush, 1984; (of. Dobson & Shaw, 11986; 1984; Hollon et aI., al., 1986) or successful antidepressive pharmacotherapy pharmaeotherapy (cf. (el. Bellack, 1985). 1985). Likewise, connectionism connectionism should make physicians more accepting of of learning based therapies since they are altering the same synaptic functions as
W W. W W.
Tryon
1101 01
pharmacologists target. Hence, the theoretical synthesis provided by connectionism also provides a basis for more cooperative and respectful collegial relationships between psychology and psychiatry. are creating brain changes through learning therapies That psychologists are has positive implications for reimbursement by third parties and inclusion within health care legislation. legislation. Connectionism makes it considerably more difficult for medical organizations to exclude psychological interventions. Conclusions
The bidirectional associative memory and its resulting memory field and funnel-shaped memory wells has many heuristic properties for understanding both normal and abnormal psychological processes and the behaviors they mediate. mediate. Like all neural networks, the BAM can be implemented on a digital computer computer and its functional properties studied in detail. The combination of of compelling theoretical hypotheses and openness to complete experimental investigation set the occasion for many research opportunities. References
Abramson, L. Y., Seligman, M. E. P., & Teasdale, J. 1. D. ((1978). 1 978). Learned of helplessness in humans: Critique and reformulation. Journal of Abnormal Psychology, 87, 49-74. staastical manual American Psychiatric Association ((1987). 1 987). Diagnostic and statistical of mental disorders (3rd ed. - Revised). Washington, DC: American of Psychiatric Association. American Psychiatric Association ((1994). 1994). Diagnostic and statistical manual of o f mental disorders (4th ed.). Washington, DC: American Psychiatric Association. Anderson, J. 1. A. ((1983). 1 983). Cognitive and psychological computation with neural models. IEEE IEEE Transactions Transactmns on S Systems, ystems, Man, and Cybernetics, SMC799-815. 113, 3, 799-8 15. M.,. , & Gunther, Giinther, W W.. ((1992). Banquet, 1J.. P., Smith, M 1 992). Top-down processes, S.. Levine & SS.. J. attention, and motivation in cognitive tasks. In D. S Leven (Eds.), Motivation, Moavation, emotion, and goal direction in neural networks (pp. 1169-208). 69-208). Hillsdale, NJ: Lawrence Erlbaum. and treatment. Philadelphia, P PA: 1 967). Depression: Causes and A: Beck, A. T. ((1967). of Pennsylvania Press. University of
1 02 102
Chapter 3
Beck, A. T. (1976). and depressive disorders. New York: ( 1976). Cognitive therapy and International Universities Press. International Beck, A. T., Rush, Rush, A. J., 1., Shaw, Shaw, B., & Emery, G. (1979). ( 1 979). Cogniave Cognitive therapy Guilford Press. ofdepression: A treatment manual. New York: Guilford of Bellack, A. S. (1985). ( 1 985). Psychotherapy research in depression: An overview. In (Eds.), Handbook of of depression: Treatment, E. Beckham & W. Leber (Eds.), assessment assessment and research (pp. 204-219). 204-2 1 9). Homewood, IL: Dorsey Press. ( 1 986). Schizophrenia: Schizophrenia: Behavior therapy's forgotten child. Bellack, A. S. (1986). Behavior Therapy, Therapy, 17, 1 7, 199-214. 1 99-2 14. Behavior Blaney, P. H. (1986). ( 1 986). Affect and memory: A review. Psychological Bulletin, 229-246. 99, 229-246. Blatt, Blatt, S. J., 1., & Bers, S. A. (1993). ( 1 993). Commentary on "A cognitive perspective perspective on self-representation self-representation in depression". depression". In Z. V. Segal & S. J. 1. Blatt, (Eds.), The self Cognitive and psychodynamic self in emotional distress: distress: Cognitive 1 3 1 - 1 70). New York: Guilford Press. perspectives (pp. 131-170). perspectives Bower, G. H., & Cohen, influences in memory and Cohen, P. R. (1982). ( 1 982). Emotional influences thinking: Data and theory. In M. SS.. Clark & SS.. T. Fiske (Eds.), Affect and cognition (pp. 241-331). 24 1 -33 1). Hillsdale, NJ: Lawrence Erlbaum. Abnormal Carson, R. C., Butcher, Butcher, J. 1. N., & Coleman, J. 1. C. ((1988). 1 988). Abnormal life (Sth (8th ed.). Glenview, Glenview, IL: Scott, Scott, Foresman & psychology and modern life Co. c . , Roitblat, Roitblat, H. L., Hamada, R. S., Carlson, J. 1. G., & Twentyman, Chemtob, C., C. T. ((1988). 1 988). A cognitive action theory of post-traumatic stress disorder. Journal oof fAnxiety Disorder, 2, 253-275 253-275.. Conte, H. R. ((1975). 1 975). A circumplex model f or personality for personality traits. Unpublished doctoral dissertation, New York University. University. (Dissertation (Dissertation Abstracts International: 1 73 1 ). lnternaaonal: University Microfilms No. 260 2601731). Conte, H. R., & Plutchik, R. ((1981). 1 98 1). A circumplex model model for interpersonal personality traits. Journal o f Personality and Social Psychology, of Psychology, 40, 70 1-7 1 1 . 701-711. 1992). Reaction to trauma: A Creamer, M., Burgess, P., & Pattison, P. ((1992). cognitive processing model. f Abnormal Psychology, 01, model. Journal oof Psychology, 1101, 452-459. Crick, F., & Mitchison, G. ((1983). 1 983). The function of Nature, 304, 304, of dream sleep. sleep. Nature, 1111-114. 1 1 - 1 14. S.,. , & Shaw, B. FF.. ((1986). Dobson, K. S 1986). Cognitive assessment with major depressive disorders. Cognitive Cognitive Therapy Therapy and Research, Research, 10, 10, 13-29. 13-29. J. W., & Palmer, D. C. ((1989). Donahoe, 1. 1 989). The interpretation of complex Some reactions to Parallel Distributed Processing, human behavior: Some
W. W W. Tryon Tryon W
103 103
by 1. J. L. McClelland, D. E. Rumelhart, and the PDP research edited by 399-4 1 6 . group. Journal of of the Experimental Analysis of of Behavior, Behavior, 51, 51,399-416. J. W W.,. , & Palmer, D D.. C. ((1994). Donahoe, 1. 1 994). Learning and complex behavior (pp. 66-67, Note 70). New York: Allyn and Bacon. Eaves, G., & Rush, A. 1. 1 984). Cognitive patterns in symptomatic and J. ((1984). of Psychology, f Abnormal Psychology, remitted unipolar major depression. Journal o 93, 331-40. 1 -40. emotion in psychotherapy. New York: Lyle Ellis, A. ((1962). 1 962). Reason and emotion Stuart. Ellis, A. ((1980). 1 980). Rational-emotive therapy and cognitive behavior therapy: Cognitive Therapy Therapy and Research, Research, 4, 4, 325Similarities and differences. Cognitive 340. therapy: Readings in Eysenck, H. J. ((1964). 1 964) . Experiments in behaviour therapy: modern methods of modern methods of treatment of of mental disorders derived from theory. New York: Macmillan. learning theory. J. ((1986). of fear: Exposure 1986). Emotional processing of Foa, E. B., & Kozak, M. 1. 20-35. information. Psychological Bulletin, 99, 20-3 to corrective information. 5. Foa, E. B B.,. , & Riggs, D D.. SS.. ((1993). 1 993). Post-traumatic stress disorder in rape J. Oldham, M M.. B B.. Riba, & A. Tasman (Eds (Eds.), .), American victims. In 1. (Vol. 112, of psychiatry (Vol. Psychiatric Press review review of 2, pp. 273-303). Washington, DC: American Psychiatric Press. B.. O O.. ((1989). 1 989). Behavioral/cognitive Foa, E. B., Steketee, G., & Rothbaum, B conceptualization of of post-traumatic stress disorder. Behavior Therapy, Therapy, 20, 155-176. 1 5 5 - 1 76. Foa, E. B., Zinbarg, R., & Rothbaum, B. UncontroUability and B . O. ((1992). 1 992). Uncontrollability unpredictability in posttraumatic stress disorder: disorder: An An animal model. Psychological Bulletin, Bulletin, 1112, 1 2, 218-238. 2 1 8-23 8 . Gilgen, A A.,. , & Abraham, F. (1995). ( 1 995). Chaos Chaos theory in psychology. Westport, CT: Greenwood Greenwood Publishing. Publishing. ( 1 988). Self-knowledge Self-knowledge and self-deception. Greenwald, A. G. (1988). self-deception. In J. S. Self-decepaon: An adaptive adaptive Lockard & D. D . L. Paulhus (Eds.), Self-deception: mechanism? (pp. 113-131). 1 1 3- 1 3 1 ) . Englewood Englewood Cliffs, NJ: Prentice Hall. ( 1 992). New look 3' 3 : Unconscious cognition cognition reclaimed. Greenwald, A. G. (1992). American American Psychologist, 47, 766-779. 766-779. Grossberg, of attentionaUy Grossberg, S., & Levine, D. S. (1987). ( 1 987). Neural dynamics of attentionally modulated Pavlovian conditioning: blocking, interstimulus interval, and secondary reinforcement. reinforcement. Applied Optics, 26, 5015-5030. 5 0 1 5 -5030.
1 04 104
Chapter 3
Grossberg, S., & Schmajuk, N. A. (1987). ( 1 987). Neural dynamics of of attentionaUy attentionally modulated modulated Pavlovian conditioning: conditioning: Conditioned Conditioned reinforcement, inhibition, and opponent opponent processing. Psychobiology, 15, 195-240. 1 95-240. ( 1992). A neural network network theory of of manic-depressive manic-depressive illness. In Hestenes, D. (1992). D. S. Levine & S. J. and goal 1. Leven (Eds.), Motivation, Motivation, emotion, emotion, and direction in neural networks (pp. 209-257). Hillsdale, NJ: Lawrence networks (pp. Erlbaum. Erlbaum. Hilgard, Hilgard, E. R. (1977). ( 1 977). Divided DiVided consciousness: Multiple Multiple controls in human thought thought and action. New York: Wiley. Hoffman, R. E. (1987). ( 1 987). Computer simulations of of neural information infonnation processing and the schizophrenia-mania schizophrenia-mania dichotomy. Archives of of General 1 78-1 88. Psychiatry, 44, 178-188. Hoffman, Hoffinan, R. E. (1992). ( 1 992). Attractor neural networks and psychotic disorders. 1 19-1 24 . Psychiatric Annals, 22, 119-124. Hoffman, R.. E E.,. , & Dobscha, S.. K K.. ((1989). Hoffinan, R Dobscha, S 1989). Cortical pruning and the of schizophrenia: A computer model. Schizophrenia development of computer model. Schizophrenia Bulletin, 15, 477-490. Hollon, S. D., & Beck, Beck, A. T. (1994). ( 1994). Cognitive and cognitive-behavioral therapies. In A. E. Bergin & S. L. Garfield (Eds.), Handbook of of therapies. psychotherapy and behavior change. psychotherapy change. (4th ed., pp. 428-466). New York: Wiley. Hollon, S. D., Kendall, P. C., & Lumry, A. ((1986). 1 986). Specificity of of depressotypic f Abnormal depressotypie cognitions in clinical clinical depression. Journal oof Psychology, 95, 29-52. Hopfield, J. 1. J. ((1982). 1982). Neural networks and physical systems with emergent f the National Academy collective computational abilities. Proceedings o of 79, 2554-2558. of Sciences, of Hopfield, 1. 1 983). 'Unlearning' has a J. 1., J., Feinstein, D D.. I., & Palmer, R. G. ((1983). stabilizing effect in collective memories. Nature, 58-1 59. Nature, 304, 304, 1158-159. Hopfield, 1. 1 987). Computing with neural circuits: A J. J., & Tank, D D.. W. ((1987). model. Science, SCience, 233, 233, 625-632. cognition, and social behavior. In R. S. S. Wyer & Isen, A. M. ((1984). 1984). Affect, cognition, T. R. Srull (Eds.), 79-236). (Eds.), Handbook of of social cognition cognition (Vol. (Vol. 3, pp. 1179-236). Hillsdale, NJ: Lawrence Erlbaum. Jacoby, L. L., Lindsay, D. S., & Toth, 1. 1992). Unconscious influences J. P. ((1992). revealed: Attention, awareness, and control. control. American Psychologist, Psychologist, 47, 47, 802-809. Janet, P. ((1893). 1 893). L'amnesie continue [Continuous amnesia]. amnesia]. Revue Revue General Des Sciences, 4, 167-1 79. Sciences, 4, 167-179.
W. W W. W
Tryon
1105 05
Janet, PP.. ((1904). et la dissociation des souvenirs par remotion l'emotion Janet, 1 904). L'amnesie et emotion].. Journal de [Amnesia and the dissociation of memories by emotion] 1, 4417-453. 1 7-453 . Psychologie Normale et Pathologique, 1, Janet, P 1 920). The major symptoms of P.. ((1920). of hysteria. New York: Macmillan. C.,. , & Barlow, D D.. H. ((1990). Jones, J. 1. C 1990). The etiology of post-traumatic stress disorder. Clinical Psychology Review, 10, 299-328. R. ((1991). of learning and the biological Kandel, E. R. 199 1). Cellular mechanisms of of individuality. In E. R. Kandel, J. basis of E. R. 1. H. Schwartz, & T. M. Jessell of neural science (pp. 11009-1031). (Eds.), Principles of 009- 1 03 1). Norwalk, CT: Appleton & Lange. Kazdin, A 1 978). History of A. E. ((1978). of behavior modification: Experimental foundations of contemporary research. Baltimore, MD: University Park f oundations of Press. Kazdin, A A. E. ((1979). of behavior therapy. 1 979). Fictions, factions, and functions of 629-654.. Behavior Therapy, 10, 629-654 M.,. , Fairbank, 1. J. A, A., Caddell, 1. J. M., Zimering, R. R. T T.,. , & Bender, M. Keane, T. M postE. ((1985). 1 985). A behavioral approach to assessing and treating post traumatic stress disorder in Vietnam veterans. In C. R. Figley (Ed.), Trauma and its wake (pp. 257-294). New York: Brunner/Mazel. Kihlstrom, 1. 1 987). The cognitive unconscious. SCience, 7, 11445-237. 445-237. J. F. ((1987). Science, 23 237, J. F., Barnhardt, T. M., & Tataryn, D. 1. J. ((1992). Kihlstrom, 1. 1 992). The psychological unconscious: unconscious: Found, lost, and regained. American Psychologist, 47, 4 7, 788-791. 788-79 1 . ( 1987a). Constructing an associative memory. Byte, Kosko, B B.. (1987a). Byte, 12, September, 137-144. 1 37-144 . Kosko, B. (1987b). ( 1 987b). Adaptive bidirectional associative memories. Applied Applied Optics, 26, 4947-4960. 4947-4960. ( 1 988). Bi-directional associative memories. IEEE IEEE Transactions Kosko, B. (1988). on Systems, Systems, Man Man and and Cybernetics, Cybernetics, 18, 49-60. of emotional imagery. Lang, P. J. 1. (1979). ( 1 979). A bio-information theory of Psychophysiology, 16, 1 6, 495-512. 495-5 12. Lang, P. J. of emotion: 1. (1985). ( 1 985). The cognitive psychophysiology of emotion: Fear and A H. Turner & J. 1. Maser (Eds.), Anxiety Anxiety and and the anxiety anxiety. In A. disorders (pp. 131-170). 1 3 1 - 1 70). Hillsdale, NJ: Lawrence Erlbaum. 1. E. (1994). ( 1 994). Emotion, memory and Scientific American, LeDoux, J. and the brain. Scientific
270, 50-57. 50-57. Leven, S. S . J. 1. (1992). ( 1 992). Learned helplessness, memory, and the dynamics of of S . Levine & S. S . J. 1. Leven (Eds.), Motivation, Motivation, emotion, and and hope. In D. S.
106
Chapter 3
(pp. 259-299). 259-299). Hillsdale, NJ: goal direction in neural networks (pp. Lawrence Erlbaum. Leventhal, H. ((1984). of emotion. emotion. In K. R. 1 984). A perceptual-motor theory of Scherer & P. Ekman (Eds .), Approaches to emotion (pp. 1 -29 1). (Eds.), (pp. 27 271-291). New York: Academic Press. 1 991). Introduction to neural and cognitive modeling (pp. Levine, D. SS.. ((1991). 41-92). 4 1-92). Hillsdale, NJ: Lawrence Erlbaum. 1992). Motivation, emotion, Levine, D. S., & Leven, S. 1. J. (Eds.) ((1992). emotion, and goal 69-365). Hillsdale, NJ: direction in neural networks (Part II, pp. 1169-365). Lawrence Erlbaum. 992). Integration, Levine, D. S S.,. , Leven, SS.. 1., J., & Prueitt, P. SS.. (1 (1992). disintegration, disintegration, and the frontal lobes. lobes. In D. SS.. Levine & SS.. J. Leven (Eds.), Motivation, emotion, emotion, and goal direction in neural networks (pp. (Eds.), 301-335). Hillsdale, NJ: Lawrence Erlbaum. 301-335). Lewinsohn, P. M., Steinmetz, J. 1. L., Larson, D. W., & Franklin, J. 1. ((1981). 198 1). Depression-related cognitions: Antecedent or consequences? Journal of of Abnormal Psychology, 90, 2213-219. 1 3-2 1 9. Spiegel, D D.. ((1992). model of of dissociative Li, D., & Spiegel, 1 992). A neural network model disorders. Psychiatric Annals, 22, 1144-147. disorders. 44-147. B.. T. ((1992). Litz, B 1992). Emotional numbing in combat-related post-traumatic stress disorder: A critical review and ref ormulation. Clinical Psychology reformulation. Review, 12, 4417-432. 1 7-432. Loftus, 1 980). Memory: Surprising new insights into how we remember LoRus, E. ((1980). and orget. Reading MA: Addison-Wesley. and why we f forget. 1. H. ((1991). 199 1). The collective electrical behavior of Martin, J. of cortical neurons: The electroencephalogram electroencephalogram and the mechanisms of of epilepsy. In E. R. Kandel, 1. J. H. Schwartz, & T. M. Jessell (Eds.), Principles of of neural science (3rd ed., pp. 777-79 1). New York: 777-791). York: Elsevier. 1 992). Mood-congruent Matt, G. E E.,. , Vazquez, Vazquez, C., & Campbell, W. K. ((1992). Mood-congruent recall of ectively ton� of aff affectively toned stimuli: A meta-analytic review. review. Clinical Psychology ReView, Review, 12, 227-255 227-255.. McDougall, W. ((1921). 1 92 1) . An introduction introduction to social psychology. Boston: Boston: Luce. Lute. 1. C. ((1967). 1 967). Psychometric theory. New York: McGraw-Hill. Nunnally, NunnaUy, J. McGraw-Hill. Plutchik, R. ((1958). 1 958). Outlines of of a new theory of of emotion. emotion. Transactions of of the New York Academy of Sciences, 20, 394-403 . of 394-403. 1980). Emotion: A psychoevolutionary synthesis. New York: Plutchik, R. ((1980). Harper & Row. 1994). The psychology and biology of Plutchik, R. ((1994). of emotion. emotion. New York: HarperCollins.
W.W. W. w.
Tryon Tryon
1107 07
Plutchik, R. (in press). The circumplex as a general model of the structure of Plutchik & H. R. Conte (Eds.), emotions and personality. In R. Plutchik models of of personality and emotions. emotions. Washington, DC: Circumplex models American Psychological Association. 1 992). Familiarity and novelty: The contributions of the limbic Pribram, K. ((1992). S.. Levine forebrain to valuation and the processing of relevance. In D. S S.. J. Leven (Eds.), Motivation, Motivation, emotion, emotion, and goal direction in neural &S networks (pp. 337-365). Hillsdale, NJ: Lawrence Erlbaum. J. F. ((1981). of science science f for 1 98 1). A philosophy of or personality theory (2nd Rychlak, 1. ed.). Malabar, FL: Robert E. Krieger Publishing Co. status.. Journal 1987). Implicit memory: History and current status Schacter, D. L. ((1987).
of Experimental Psychology: Psychology: Learning, Learning, Memory, Memory, and Cognition, Cognition, J13, of 3, 5501-518. 0 1 -5 1 8. insight. New Haven: Yale University Schafer, R. ((1978). 1 978). Language and insight. Press. of facial expressions. Schlosberg, H. ((1941). 1 94 1). A scale for the judgment of of Experimental Psychology, Psychology, 29, 29, 497-5 497-510. Journal of 10. of emotion. Psychological Review, Review, Schlosberg, H. ((1954). 1 954). Three dimensions of 61, 881-88. 1 -88. Segal, Z 1 993). The self Z.. V., & Blatt, SS.. J. (Eds.) ((1993). self in emotional distress: Cognitive andpsychodynamic perspectives. New York: Guilford Press. ical Bulletin, 1115, Seger, C. A. ((1994). 1 994). Implicit learning. Psycholog 1 5, 1163-196. 63- 196. Psychological Seligman, M. E. P. (1970). On the generality of of learning. ( 1970). On of the laws of Psychological Review, 77, 406-4 406-418. 18. Seligman, M. E E.. P., Abramson, L. Y Y.,. , Semmel, A A.,. , &von & von Baeyer, C. ((1979). 1 979). Jownal of of Abnormal Psychology, Psychology, 88, Depressive attributional style. Journal 242-247. 1. L. (Eds.) (1972). ( 1 972). Biological boundaries of Seligman, M. E. P., & Hager, J. of learning. Englewood Englewood Cliffs, NJ: Prentice-Hall. Spiegel, D. (1990). ( 1 990). Hypnosis, dissociation, and trauma: Hidden and overt 1. L. Singer (Ed.), Repression and dissociation: dissociation: observers. In J. for personality theory, psychopathology, and Implications for and health (pp. 121-142). Chicago, IL: The The University of 1 2 1 -1 42). Chicago, of Chicago Press. ( 1 993). Contemporary behavior Spiegler, M. D., & Guevremont, D. C. (1993). Spiegler, Grove, CA: Brooks/Cole Publishing Co. therapy (2nd ed.). Pacific Grove, ( 1 986). Behaviorism with a personality: The paradigmatic Staats, A. W. (1986). behavioral assessment approach. In R. O. Nelson & S. Hayes (Eds.), S . C. Hayes foundations of of behavioral assessment (pp. 242-296). New Conceptual foundations York: The Guilford Guilford Press.
1 08 108
Chapter 3
Teasdale, Teasdale, J. 1. D., D., & & Fogarty, Fogarty, S. S . J. (1979). ( 1 979). Differential Differential effects effects of of induced induced mood on on retrieval of of pleasant pleasant and and unpleasant unpleasant events events from from episodic episodic memory. ofAbnormal Abnormal Psychology, 88, 248-257. 248-257. Journal of Thompson, R. F. (1986). ( 1986). The The neurobiology neurobiology of of learning learning and and memory. Science, Thompson, 94 1 -947. 233, 941-947. Thompson, R. F. (1990). ( 1 990). Neural Neural mechanisms of of classical conditioning conditioning in Thompson, mammals. Philosophical Transactions Transactions of of the Royal Society, Society, 306, 306, London, 161-170. 1 6 1 - 1 70. London, Tryon, W. W. (1993a). ( 1993a). Neural Networks: Networks: I. Theoretical Unification Through Review, 13, 1 3, 341-352. 34 1-352. Connectionism. Clinical Psychology Review, ( 1 993b). Neural Networks: II. Unified Learning Theory and Tryon, W. W. (1993b). ReView, 13, 353-371. 353-37 1 . Behavioral Psychotherapy. Clinical Psychology Review, Tryon, W. W. (1995a). ( 1 995a). Neural networks networks for behavior therapists: What What they Therapy, 26, 295-318. 295-3 1 8 . are and why they are important. important. Behavior Therapy, Tryon, W W.. W W.. ((1995b). 1995b). Synthesizing animal and human research via neural Therapy and network learning theory. Journal of of Behavior Therapy 303-3 12. Psychiatry, 26, 303-312. Experimental Psychiatry, Tryon, W. W. ((1995e). 1 995c). Resolving the cognitive behavioral controversy. controversy, the Behavior Therapist, Therapist, 18, 18, 83-86. Tryon, W. W. ((1996a). 1996a). Encoding emotion in connectionistic eonneetionistic neural networks. Manuscript submitted for publication. model of of posttraumatic stress Tryon, W. W. ((1996b). 1996b). A neural network model disorder. Manuscript submitted for publication. Tryon, W. W. ((1996c). 1996c). Yes - Neural network learning theory can resolve the 9, 72-73. behavioral-cognitive behavioral-cognitive controversy. The Behavior Therapist, Therapist, 119, and behavior therapy: Wachtel, P. L. ((1977). 1977). Psychoanalysis Psychoanalysis behavior therapy: Toward Toward an integration. integration. New York: Basic Books. Wasse rman, P 1989). Neural computing: practice (pp. 1 3Wasserman, P.. D. D. ((1989). compuang: Theory Theoryand andpractice (pp. 11131125). 25). New York: Van Nostrand Reinhold. 1 988). Williams, 1. J. M. G., Watts, �. F.. N., MacLeod, c., C., & Mathews, A. ((1988). Cognitive Cognitive psychology and emotional emoaonal disorders. disorders. New York: York: Wiley. J., & & Lazarus, lazarus, A. A. ((1966). therapy techniques: techniques: A guide 1 966). Behavior therapy Wolpe, 1., to the treatment treatmentofneuroses. neuroses. New York: Pergamon. Author Author Notes Notes
II wish wish to to thank thank Scott Scott Badgett Badger for his careful careful reading reading and and helpful helpful comments comments during during an an earlier earlier stage stage of manuscript manuscript preparation preparation and and for for testing testing the the readability readability of the the Appendix material. material.
1109 09
IV. w. 14(. W. Tryon
Appendix: Bidirectional Associative Memory Appendix: Description of the Bidirectional Network architecture architecture Network two layer neural network which forms the Figure 2 illustrates a simple two of the BAM BAM mechanism we will explore. The S column of of stimulus basis of the R column of of response nodes. The nodes is completely interconnected with the connecting synaptic weights are drawn generically as lines but can also be represented as a memory matrix (M).
S S
() 0 0 C) 0 0 (3 S--� S
M M
;RR
�
0 0 0 0 0 0 0 (Z)
R R T M T M
Two layer layer neural network underlying the bidirectional associative Figure 2. Two (BAM) mechanism. memory (BAM)
Stimulus Stimulus and and response response definitions definitions Stimulus and and response vectors (number (number strings vs. matrices) can can be (0,1) defined as a binary (0, 1 ) pattern. Elements in the stimulus vectors can refer to of specific characteristics such as red hair, bald, and the presence vs. absence of
1110 10
Chapter 33 Chapter
the response vectors can reflect the presence vs. absence of of tall. Elements in the particular particular acts such as saying thank-you. Different levels of of specificity can be chosen. At a micro level, each stimulus node can refer to a afferent neuron and each response node can can refer to a efferent neuron. More typically, each stimulus node corresponds to a perceptual attribute and each response node to an entire behavioral response. Stimuli can include the physical consequences of of one's own behavior and that that of of others as well as inanimate stimuli. to the results of of memory retrieval, Responses, in this article, primarily refer to also represent actions taken including pertinent associations, but can can taken in response to stimuli which set the occasion for behaving. Stimulus Stimulus and response vectors can also be given bipolar codes of o f -- 11 and 11 instead of of 0 and 11 thereby directly representing polar opposite characteristics such as large vs. small, heavy vs. light, sharp vs. smooth or responses such as approach vs. avoid, dominant vs. submissive. Long stimulus and and response vectors can can be folded to form matrices. For example, the 00 elements of the 1100 of a stimulus vector can be folded to represent elements in a 110 0 by 110 0 matrix on which a visual pattern can be imposed by coloring the pixels (picture elements) according to 0 = = white and 11 = = black. For example, the letter A could be encoded as illustrated in Figure 33.. Vector elements 0 - 9 form the first row, elements 10 - 19 form the second row, through elements 90 - 99 which form the final row. Another folded matrix might represent the subject's perception of of the based on the response vector might letter A. Response vectors, and and folded matrices, can can represent memories of of stimuli which in turn can function as stimuli for other associations. We limit our discussion to a single S-R configuration but many sequences can be cascaded. Response vectors can can also represent a sequence of of actions to be taken or a code for complex behavior.
Memory (S-R) (S-R) encoding encoding The steps necessary to encode three S-R associations, memories, into our neural network are described in Table 11.. The architecture of of this network is fully interconnected meaning that all stimulus nodes are connected to every response node. The first step is to define stimulus-response pairs; items to be associated in memory. The three pairs are specified in binary form as S Il-R1, -R l , $2-R2, and S3-R3 $3-R3.. Notice that each stimulus is characterized by 8 digits and S2-R2, each response by 5 digits. An equal number could have been chosen, or more
~ W W. W
I111 II
Tryon Tryon
Ill //
Start
//
// // m
m
m
~
i
!
l..- ~.-.
End
i m
m
Figure 33.. Example Example of of encoding the letter A into a 100 element stimulus vector that has been folded into a a 110 that 0 by 110 0 matrix.
digits allocated The 8 stimulus stimulus digits digits could could represent represent the digits allocated to to responses. responses. The presence absence of of 8 physical physical attributes attributes of of the the stimulus or they could could presence vs. absence represent aa binary code referencing one of 256 situations. situations . Similarly, the represent binary code referencing one of 28 2 8 = 256 5 response response digits could could represent the presence presence vs. absence of of 5 physical characteristics the subject's could represent characteristics of of the subject's response response or it could represent aa binary binary code code referencing responses. The Hopfield variant of the the different responses. The Hopfield variant of referencing one one of of 25 2 5 = 32 different BAM R vector vector equal to the Tank, 1987) 1 987) sets the R to the BAM (Hopfield, (Hopfield, 1982; 1 982; Hopfield Hopfield & Tank, S vector. The second these pairs pairs in bipolar bipolar form to to avoid The second step is to to redefine redefine these subsequently subsequently introducing introducing asymmetrical asymmetrical effects effects when when the the threshold threshold function function is implemented. to -1 and This is done done by by converting converting every every 0 to and leaving the the + + l's l 's implemented. This unchanged. The distance between -1 and 0 is now equal to the distance now equal to the distance unchanged. The distance between - 1 and between 0 and 1. between and ++ 1. Next Next we we construct construct aa memory memory matrix matrix using using steps steps three three and and four. four. The The third step creates a I • J distributed memory matrix (M) in accordance with third step creates a x distributed memory matrix accordance with equation equation (1) ( 1 ) where where T T refers refers to to transposition transposition meaning meaning that that numbers numbers previously previously written row are are now now written written as as aa horizontal horizontal row written as as aa vertical vertical column column or or vice vice =
=
Chapter 3 Chapter
1112 12
Table 11.. Example Example Calculations Calculations for for Memory M e m o r y Encoding Encoding and and Decoding Decoding in in a a Table Bidirectional Associative Associative Memory M e m o r y (BAM). (BAM). Bidirectional
I" Def Define Binary Values for S, J Binary Binary Values for R Step I: me I Binary S I ==111 10 0 0000101 1 1 $ 2 == 0011000 0 00100 1 0 S2 $3== I1 0 0000000 0 10 1 S3
R RlI ==11001 01 10 1 R 2 = I1I 01 1011 1 R2 R 3 = 1I 0O1 I1 0I 0 R3
Step 2: 2: Transf Transform onn Binary Values for S and R into Bipolar Values S I == 11 11-1-1-1-1 -1 -1 -I -1 11 11 $2=-1 S2 = -I I1-1-1-1-1 -l -l -l -l l1-1 -l $3== 11 --11 --11 --11 --1 1 - -1 1 - 1-I I1 S3
R I ==I - 1 -1 11-1 Rl -1 11 R 2 = I1 11-1 R2 -1 11 I1 R3 = 11-1 -1 11 11-1 -1
3: Create a Memory Matrix Matrix for Each Each Stimulus-Response Pair Step 3: Second Pair
First Pair
SII T T S
-1 -1 -1 -1
1 -1 1 1 -1 -1 =1 -1 --11 1 1
R1 RI 1 -1
1
-1 -1 1 1 1 1 -1 -1 -1
1 1 -1 -1 -1 -1 --11 1 1
=1 -1 -1 1 1 1 1 -1 -1 --11
1 1 =1 -1 -1 -1 -1 -1 1 1
-1 -I -1 -1 -1 1 1 1 1 -1 -1 -1 -1
R3 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 1 1 1 1
-1 -1 -1 -1 1 1 1 1 -1 -1 -1 -1
Third Pair Pair T S3 T
-1 -1 -1 -1
1 1 1 -1 -1 =1 -1 -1 -1 -1 -1 1 1
$2 T T S2 -1 11 -1 -1 -1 -1 -1 -1 11 --11
1 -1 11 -1 =1 -1 -1 -1 =1 -1 -1 1 -1 -1
R2 1 -1 1
1
-1 =1 1 --11 --11 11 --11 1 1 -1 -1 1 -=11 -1 =1 -1 -1 1 -1 -1 -1 -1 -1 -1 1 -1 =1 -1 -1 -1 -1 1 -1 -1 -1 -1 11 -1 11 -1 11 -1 -1 -1 -1 1 -1 -1
W. W W. W Tryon
113 1 13
TTable a b l e 11 ccontinued. ontinued. 4: Create Create aa Composite Composite Memory Memory Matrix Matrix Through Through Adding Adding Corresponding Corresponding Elements Elements Step 4:
Across the the Above Above Three Three Matrices Matrices Across
J 1 1 -3 -3 -33 -3 -3 1 1 -
-3 -3 1 1 1 1 1 1 -3 -3
33 -1 -1 -I -I -1 -1 -1 -1 -1 -1 33
-I -1 -1 -I -1 -1 -1 -1 -1 -1 -1 -1 -1 -1
-I -1 33 -I -1 -1 -1 -I -1 -1 -1 33 -1
Step 5: New Stimulus 8timulus for Presentation Presentation 5: 8elect Select aa New We choose choose the the following variant variant of of S 8 1" 1: S 8 = (1 ( 1 11 0 11 11 0 11 11)) We 6: Vector Vector Multiply Multiply M M by by S 8 Step 6:
The The stimulus stimulus vector is written written horizontally horizontally as follows: 8 S ==( (111 01 101 10 1 10) 1 1 )
The memory matrix matrix is written written as follows: The
1 1 -3 -3 -3 -3 -3 -3 -3 1 1
J -3 -3 3 1 --11 1 --11 1 -1 1 -I -1 1 --11 I1 --11 -3 -3 3
-1 --1I --11 -1 --1I --1I --1I --1I
-1 3 --11 -1 -1 --1I 3 --11
Matrix multiplication requires combining the the one stimulus row row with with each each of of the columns in Matrix the M M matrix resulting in a single entry for each column of of the the M M matrix as follows. the
1114 14
Chapter 3 Chapter
Table 11 continued. continued. Table The The first tenn term of of each each nwnerical numerical pair refers to the S vector and the second tenn term refers to a column in the the M matrix: colwnn [( -3) + -3) + (O I XI ) + ( I X-3) + I X I )] = -2, I XI ) + I XI ) + [(IX1) + ((IX1) + (OX (ox-3) + (IX-3) + ((II X X-3) (ox-3) + ((IX1) + ((I)(I)1 X-3) + which is the the first entry in the first first memory response vector. The second entry in the the R vector colwnn of the M matrix. The third, is obtained by multiplying the S S1I vector by the second column fourth, and fifth entries in the R vector are obtained by combining the S vector with the fourth, and fifth colwnns columns of of the M matrix. third, fourth,
- 2 --2 2 22- 6-62 )2) R ==((-2 Multiplying Multiplying aa matrix matrix by by aa binary binary vector, vector, such such as S, S, equates equates to to adding adding the the colwnn column entries entries in in the matrix by the rows associated with Il's. 's. In our case, this means that we add the first, fourth, fifth, seventh, and eighth row entries in each M colwnn. column. second, fourth,
Step Step 7: 7: Apply the Threshold Function Function (Eq. 33)) The first entry ooff R R iinn Step 6 iiss -2 which iiss less than than 00,, therefore the threshold function function replaces replaces the the -2 -2 entry entry with with aa 00 as indicated indicated below. below. The The same same is is true true for for the the second second entry entry of of-2. -6 is 2. The The third third entry entry of of + +22 exceeds exceeds 0 0 and and is is therefore therefore replaced replaced by by + + I1.. The The fourth fourth entry entry of of-6 is less less than than 00 and and is is replaced replaced by by 00 whereas whereas the the fifth fifth entry entry of of 22 is is replaced replaced by by + +lI because because it it An value exceeds exceeds O. 0. An value of of 0 0 remains remains O0.. R R -=( 0( 00 100 1I) 0 I ) This ers by This response response diff differs by II bit bit from from the the target target memory memory of of II 0 0 Il 0 0 I1 and and therefore therefore is is said said to to have have aa Hamming Hamming distance distance of of I1 from from the the target target memory. memory.
Step 8: 8: Use Use R R to to Associate Associate to to S S The first response vector is written as follows: R ==((000 100 1I ) 0 I ) Since Since the transpose transpose of of aa matrix matrix involves involves interchanging interchanging rows rows and and colwnns, columns, we we can can operate operate on on the transpose transpose of matrix by the of the the M M matrix by treating treating its its rows rows as as colwnns. columns.
W. W. w.
1115 15
Tryon
Table Table 11 continued. continued. ROWS of the M matrix as follows: follows: The R vector is combined with the ROW8 [(OX 1 ) + (OX-3) l X3) + (OX- I )] = [(0)(1) (0)(-3) + ((1X3) (0)(-1) X-I)] - +2, which is the first entry in the modified 8 S I ) + ((1l X vector. Multiplying the the transpose of the memory matrix by a binary vector reduces reduces to adding the vector. Given the R vector (0 0 11 0 11), we add the row entries corresponding to a 11 in the R vector. third and fifth row entries in the M matrix. The second through eighth entries of the 8 S vector are obtained by combining the R vector with the second through eighth rows of the M matrix resulting in the following modified 8 S vector. 8 (2 22-2-2-2-2 -2 -2 -2 -2 2 22)) S= =(2
9: Redefine Redefine 8 S by Applying the Threshold Threshold Function (Eq. 3) Step 9: 8 S ==((I1 11 0 0 0 0 11 11))
Step I10: 0: Multiply M by 8 S and Apply the Threshold Function (Eq. 3) R ==((I1 0 1l 0 11))
R iiss the correctly recalled memory. If this response iiss used to associate to a new stimulus 8 ). The S = ((11 11 0 0 0 0 11 11)) will be obtained which will again produce R = ((11 0 11 0 11). recall recall process is said to have stabilized. ==
versa. 3, First First Pair Pair contains versa. Notice Notice that that Table Table 11,, Step Step 3, contains a a matrix matrix where where Rl R1 values values are are entered entered horizontally horizontally just just as as written written in in Step Step 2. 2. The The S S I1 values values have have been transposed to a vertical column. entries equal equal the product of been transposed to a vertical column. The The matrix matrix entries the product of the marginal is constructed the marginal (row (row x• column) column) values. values. Another Another such such matrix matrix is constructed for for each each of of the the remaining remaining S-R S-R pairs pairs as as illustrated illustrated under under Step Step 33 in in Table Table 11.. Stated Stated technically, the outer product of technically, we we have have formed formed the outer product of the the transpose transpose of of the the S S matrix times times the the R R matrix matrix for for each each associated associated pair. pair. matrix Eq. ((1) Eq. 1)
M =
ST R
The The fourth fourth step step creates creates a a composite composite memory memory matrix matrix b byy adding adding the the corresponding elements o f all memory matrices. matrices. Hence, the entry entry in in Row Row 11,, corresponding elements of all memory Hence, the
1116 16
Chapter Chapter 33
Column 11 of of the the first first pair pair matrix matrix (+ 1) is added to to the the entry entry in in Row Row 11,, Column Column Column is added 11 of of the the second second pair pair matrix matrix (-1) which which is is added added to to the the entry entry in in Row Row 11,, Column matrix (+ 11)) to Column 11 of of the the third third pair pair matrix to obtain obtain the the Row Row 11,, Column Column 11 entry entry 1 . The Row 11,, Column -3 in of M matrix of the the M matrix of of + +1. The Row Column 2 entry entry of of-3 in the the M M matrix matrix is is -1 in derived derived from from the the corresponding corresponding entries entries of of-1 in the the first, first, second, second, and and third third pair pair matrices. matrices. This This final fmal matrix matrix is is sometimes sometimes described described as as aa "correlation "correlation matrix" matrix" because it it associates associates all all stimuli stimuli and and responses. responses. because Learning Learning entails entails encoding/representation. encoding/representation. Neural Neural networks networks can can learn learn anything anything they they can can encode/represent. encode/represent. The The memory memory matrix matrix is is composed composed of of the the weights weights which which link link each each stimulus stimulus node node with with every every response response node. node. We We refer refer to to this as as the the M M rather rather than than W W matrix matrix to to emphasize emphasize the the long long term term memory memory this function these weights. is the M, others wish to function of of these weights. Since Since W W is the flip flip of of M, others may may wish to think think of of it it as as the the weight weight matrix. matrix. Two Two observations observations are are addressed addressed to to psychologists psychologists interested interested in in dialectical dialectical 1). First, First, encoding psychological processes psychological processes (cf. (el'. Rychlak, Rychlak, 198 1981). encoding memory memory for for the the memory for association association of of S S to R R simultaneously simultaneously encodes encodes memory for the the association association between "complement of" between SSCc to to RC R c where where cc represents represents "complement of" meaning meaning that that 1l's's and and O's have O's have been been exchanged exchanged (e.g., (e.g., the the complement complement of of 11 0 0 11 11 11 is is 0 11 11 0 0 0). Second, memory memory removal removal can can be be accomplished accomplished by by adding adding the the complement complement of of Second, the memory memory to be be deleted; deleted; for for example, example, adding adding S S RC R c or or SC S c R. R. Alternately, Alternately, the memories for for specific can be be deleted deleted from from memory memory by by subtracting subtracting memories specific S-R S-R pairs pairs can the corresponding "correlation matrix" or the the equivalent equivalent operation operation of of adding adding the corresponding "correlation matrix" or -XT T Y Y to to M. M. Forgetting Forgetting is hereby modeled modeled as as an an active active process process of of altering _X is hereby altering memory content content either to new new learning learning or or by by therapeutic design. memory either incidental incidental to therapeutic design. We have have associated associated one one stimulus stimulus with with one one response. response. Many Many stimuli stimuli can can We be associated associated with with aa given response and and many many responses responses can can be be associated associated be given response with a a single single stimulus. stimulus. Given stimuli and and J responses responses an an I x• J memory memory with Given I stimuli matrix matrix results. results.
Memory recall A stimulus stimulus is in Step Step 5 ffor presentation to the the BAM to to evaluate evaluate A is selected selected in or presentation its memory memory ability. ability. It It can can be be one one of of the the original three stimuli encoded into into the the its original three stimuli encoded memory or or it it can can be be aa new new stimulus stimulus which which can can be be seen seen as as a a corrupted corrupted version version memory of one one of of the the encoded encoded stimuli; stimuli; an an environmental environmental stimulus stimulus which which is is similar similar to, to, of but not not exactly exactly like, S 11.. Choosing Choosing one one of of the the stimuli stimuli selected selected for for memory memory but like, S encoding will readily yield the appropriate appropriate response associated with with it. it. We We encoding will readily yield the response associated have chosen a slight variation of of the the first first stimulus stimulus to to make make recall recall more more have chosen a slight variation demanding and and to to illustrate illustrate the the important important property property that that the the BAM BAM is is aa flexible flexible demanding
W.. l,g.. Tryon Tryon w. w.
1117 17
memory device device that that creatively creatively recalls the the best best fitting memory memory to to a a novel memory stimulus. stimulus. Because we we have have chosen chosen a a novel stimulus, stimulus, the BAM BAM will will not converge converge to to Because a stable stable response response in a a single cycle. It will be necessary necessary to generate generate a a first first memory response, response, feed it back back through through the the BAM BAM to modify modify the the stimulus stimulus memory representation, which which will then then be used used in a a second second successful successful attempt attempt at at representation, memory memory retrieval. retrieval. Step Step 6 involves applying applying a a stimulus stimulus to to the the memory memory matrix matrix and and calculating the the first first memory multiplying the memory by calculating memory response response by multiplying memory matrix matrix by the the selected selected input input vector vector as per per Equation Equation (2). Table Table I1 describes describes how how this matrix multiplication multiplication is accomplished. accomplished. matrix
Eq. (2)
R =SM R= M
Neurons only only fire when when their their input input exceeds exceeds a a threshold threshold value. Neurons Accordingly, in Step 7 we we apply apply a a threshold threshold function function (Eq. 3) to the the results Accordingly, obtained obtained from from Eq. 2.
Eq. (3)
If R R > 0 then then R R = 11 If If R R <= <= 0 then then R R=0 If
The The first first memory memory response response of of R R = (0 0 11 0 11)) differs by by I1 bit from the the target memory memory of of R R Il to to S 11 of of 11 0 11 0 I1 and and therefore therefore is said to have have a target Hamming distance distance 44 of of 11 from from the the target target memory. memory. Had Had we used used exactly exactly S 11 we we Hamming would would now now have have exactly exactly R R1.l . Since, we do not, not, memory memory retrieval retrieval continues continues ero) has convergence on the the smallest Hamming value (including zzero) has been until convergence smallest Hamming achieved resulting in an an energy minimum. continue with Step 8 achieved minimum. Hence, we continue the response response is used used to modify the the stimulus stimulus trace trace and and associate associate to a a where the to modify new stimulus by multiplying the response vector by the transpose of the new stimulus by multiplying the response vector the transpose of the memory matrix matrix as as per Equation The transpose transpose of of matrix matrix M M simply simply uses uses memory Equation (4). The the illustrated in Step the rows rows as columns columns as illustrated Step 8 of of Table Table 11.. The The resulting resulting modified trace (representation) (representation) is S = (2 2 -2 -2 -2 -2 2 2). modified stimulus stimulus trace Eq. (4) Eq.
S = R MT
44 Hamming Hamming distance refers refers to the nwnber number of unequal entries in two vectors vectors or matrices of equal size. [1 11 0 0 I1 1 1]] and vector [1 0 11 0 11 1 1]] differ in their size. Vector VectorA = [1 vector B = [1 therefore a Hamming Hamming distance distance of 2 apart. apart. second and third positions positions and are therefore
Chapter 3
1118 18
Table 2 2.. Illustration Illustration of o f Memory M e m o r y Energy E n e r g y Calculation C a l c u l a t i o n for for S S I-RI l - R 1 .. Table
S 1 from from Step Step 2, Table 1I is 81 S 1 = ((1l 11 -1 -1 -1 -1 -1 -1 -1 -1 11 11)) The stimulus vector Sl Negating S S results results in in SI S1= = (-1 (-1 -1 11 1 11 11 -1 -1 -1)) Negating The memory matrix is written as before.
Step 11
with each of of the the col columns M. The calculations for the the first first column ofM of M is as Combine SI 81 with umns of ofM. follows: -I X 1 ) + I X-3) + IXI) + [(-1Xl) + ((-lXl) + ((1X-3) + ((lX-3) + ((1X-3) + ((1X-3) + (-I (-1X1) + (-I (-1X1)] I X-3) + I X-3) + I X-3) + XI) + [( X 1 )] = 116 6
The calculations calculations for all 5 columns are: [-16 8 -8 0 -8] The
2 Step 2 transpose of of the response vector RI is written written as follows: The transpose RI R I = 11 -1 1 -1 1 It is necessary as aa column matrix multiplication multiplication requires that that when It is necessary to rewrite rewrite R1 RI as column because matrix multiplying that one move left to right in the X vector mUltiplying vector X times times vector Y (e.g., XY) X Y) that and from top in the the Y vector. Since -SM produced a horizontal vector, R must must be and top to bottom bottom in 8ince -8M transposed to create a column vector. transposed to The R as follows for The results results of of the the -SM -8M multiplication multiplication are are combined combined with with the the transpose transpose of ofR the the first first memory: + (-8X1) [(-16XI) [(-I6X l ) + (-8XI ) + + (ox-1) (OX- I ) ++ (-8)(1)] + (8X-l) (8X- I ) + (-8X I )] = -40.
This value on aa graph like Figure the decimal This value would would be be plotted plotted on graph like Figure 11 at at the decimal equivalent equivalent of of the the stimulus stimulus and and response vector.
W W. W Tryon W.
119 1 19
Step 9 (Table (Table 1) 1 ) we we reapply reapply the the threshold threshold function function to to the the modified modified In Step stimulus trace ( 1 11 00 0 00 0 11 1) 1 ) which which is trace obtained obtained from from Step Step 8 and and find find S == (1 stimulus exactly equal to S 1 . exactly equal to 1. We We now now apply apply the the modified modified stimulus stimulus vector vector as as we we did the the original Step 10 1 0 multiplies the the memory memory matrix by the the new new stimulus, stimulus. Step matrix by reapplies the the threshold threshold function function and and computes computes aa new new response response which which is now now reapplies to R1 R l resulting resulting in aa correct correct memory recall despite aa slightly corrupted corrupted equal to stimulus cue. stimulus Had the the Hamming the first distance associated associated with with the first association association been been Had Hamming distance larger then additional memory cycles may larger than than 1, 1 , perhaps perhaps 5, then may have have ensued ensued until either obtained or either aa completely correct correct memory, Hamming Hamming = 0, was was obtained or until no reductions in Hamming no further further reductions Hamming distance distance were were achieved. Convergence Convergence is especially rapid demonstrated here. rapid in small networks networks such as as the the one demonstrated
BAM BAM energy jUnction function Neural are described If one imagines Neural networks are described by an energy function. If the fiat rubber rubber sheet, then then memory the energy field prior to to memory creation as aa flat encoding is like placing that sheet. Memory encoding placing aa ball bearing on that a creates aa memory well whose depth is given by Equation Equation 5. 5 . S represents a T stimulus vector as the memory matrix as before. M M is the matrix as before. R RT is the transpose are arranged transpose of of the the response vector which means that that its entries are vertically down the page page rather than horizontally across the page. Because memory recall involves finding energy minima, memory energy is taken taken as the memory negative of of SMR SMR T T.. However, positive energy memory hills form when negative SMR SMR T T is negative. Table Table 2 illustrates the energy calculation for the first memory. Eq. (5)
SMR T T E = --SMR
Figure 11 depicts depicts what what five memories might look like; one near near each Figure of a memory well are found by comer plus a middle entry. The coordinates of X translating the binary stimulus and response patterns into decimal X and Y (response) coordinates and and plotting the energy value (stimulus) and vertical Z axis labeled energy. For associated with that memory along a vertical example, S = 0 11 0 0 0 0 11 11 is a binary number whose decimal equivalent is, working right to left:
120
Chapter Chapter 3 4) + 0(2 7) 3 ) + 0(2 5 ) + 11(26) 2) + 0(2 0(22) 0(23) 0(24) 0(25) 0(27) (26) + 0(2 11(20) (2 0) + 11(21) (2 1 ) + 0(2 + 604++64 0 +0 = 11 + 22 + 00++0 0+ 0++00+ = 67 =67 Background energy values are are the energy values associated associated with
stimulus and and response codes not not previously associated. They are represented by all remaining pennutations permutations of of the binary digits within the stimulus and
response vectors. Background Background energy values are arc obtained obtained by substituting substituting all possible possible remaining remaining binary binary stimulus and response response patterns into Equation Equation 5 and and the result result at at the the decimal decimal equivalent coordinate of of the S-R S-R pair. For For plotting the example, S = 11 11 I1 I1 11 11 11 11 and R = 0 0 0 0 0, two binary patterns not part and R two binary patterns not part of the the three pairs pairs we used in Step 11 of of Table 11,, constitute aa background background pair pair of because because neither neither this S vector nor nor this R R vector were purposely stored in M; yet when memory memory matrix matrix M; when applied applied to Equation Equation 5 will yield an energy value that can be plotted plotted at at the the decimal equivalent equivalent coordinates of of S and and R. that Each memory memory well is a a local energy minimum called an attractor attractor 55.. Its Each field of of influence is specified by the shape of of its walls; steeper equals stronger. The of a well's attraction The extent of a memory wcU's attraction is tenned termed its basin of of attraction represented by the mouth of attraction and is graphically rcprcscntexi of its funnel-like structure (see Figure 1). If If this were a physical structure, structure, then a ball bearing structure placed at at all points points defining the basin of of attraction attraction would roll toward toward it. Like placed water water running down hill, memory and and associative processes processes located anywhere the rim of of aa basin basin of a local within the of attraction proceed (flow) downhill until a encountered where where upon the corresponding corresponding memory or energy minimum is encountered association is retrieved. Cognition entails seeking energy minima. The The shape, association or topography, of of the energy energy field fidd completely detennines determines all aspects aspects of of the or associative process process including its direction at all points in time and its final associative end state. Figure 4 was was created created ffrom Figure I1 by increasing the the basin basin of of attraction attraction rom Figure of the central central memory well. The The memory associated associated with its local minimum is of what will be recalled for many more starting points than than was previously the what case. Stimuli that that may previously have flowed into one of of the corner comer drawn into into the the abnonnally abnormally large central well.Why memories may now be drawn should energy minima be associated associated with memories, associations, associations, and neural should to aa wide variety of of problems problems such as perception and and network solutions to
55 The attractor attractor concept chaos theory which neural neural networks networksare concept is also also found found in in chaos theory to which
related. Interested Interestedreaders readers may may wish wish to consult Abraham((1995). related. consult Gilgen Oilgen and Abraham 1995).
W W W. W.
1121 21
Tryon
D• I •I I•
>0 Z W
I10 • • 6 0. Ii 4 0. "I z . 211 II l•l •
• • O0
Y
, 6 0.
I011 , . 410
z.
2a
II
•
21
X
Figure 4. Figure 11 modified such that that the central memory has a much much larger basin of of attraction.
language skills? For the the same reason that a regression line is the best fitting of squared deviations between line; because it minimizes the sum of observed and predicted points. The correct or desired solution to many problems involves minimizing a discrepancy quantity.
Summary We have formed a simple three We three memory BAM BAM connecting three pairs of of 8-bit stimuli and 5-bit response patterns. The three separate associations have been combined into a single memory matrix whose elements can be interpreted as synaptic weights between a set of of stimulus "neurons" and a set of of response "neurons". It is crucial to note that that all elements of of the memory matrix participate in storing each of of the three S-R associations (memories). This This is the essence of of a parallel distributed memory system and differs
1122 22
Chapter Chapter 3
fundamentally from the serial computer analogy of storing information in discrete places. An energy value is associated with every memory resulting in a funnel shaped memory well with a basin of of attraction. Memory retrieval entails descending to the bottom of of the memory well through a process known as gradient descent. Our example took two cycles to reach an energy minimum and correctly recall R l . We now apply these fundamental R1l when given S 1. concepts to psychopathology.
Perspectives on Personality and Emotion Cognitive Science Perspectives
- G. Matthews (Editor)
-
1997 Elsevier Science B.V. 1997 CHAPTER 4 CHAPTER
Space-Time, Order, and Hierarchy in Fronto-Hippocampal Fronto-Hippocampai System: A Neural Basis of Personality
Gaussier, Jean Claude Dreher, Cedric Cddric Joulain, Jean P. Banquet, Philippe Gaussier, ried Gunther Arnaud Revel and Wilf Wilfried G~nther
characteristic of of human human primates, in the Personality is to a large extent a characteristic language is. It results indeed from a unique combination of of same way that language higher order brain brain functions including an explicit historical memory of of the and at the same time, the ability to make plans, to change change goals, to adapt adapt past, and strategies and therefore therefore to project into the future. Both functions functions are based on strategies perception of of a a temporal temporal dimension, or at least of of an ordering of of the perception sequences of of past or future events, in the field of of consciousness. consciousness. This field of of sequences consciousness itself, or its extension extension in working memory (WM), (WM), can be consciousness considered, from a a neurophysiological neurophysiological point of of view, as the embodiment embodiment of of considered, present. the present. the past past decades, decades, the the anatomical anatomical and and structural structural bases of of these In the functions have have been been more and and more precisely delineated. In particular, particular, functions hippocampal system system (Hs) been established established as the the support support for hippocampal (Hs) has has been for the the acquisition acquisition of of explicit explicit or or declarative declarative historical historical memory. Prefrontal Prefrontal cortex cortex (Pc) could function one one step step further. Indeed, it is not not so much could bring bring this this temporal temporal function much the of events, a a function the specific repository repository of of learned learned past past sequences sequences of function it could share temporal and and parietal parietal cortices. share with other other cortical cortical structures structures such such as temporal More specifically, specifically, it emerges the elective site site for planification, planification, More emerges as as the intentionality, and goal and therefore therefore the the structure structure intentionality, motivation motivation and goal propagation, propagation, and where the the hierarchical hierarchical ordering ordering of of the the sequences sequences of of cognitive cognitive or where or motor motor events events takes not limited to extemaUytakes place. place. More More importantly, importantly, this planification planification is not to externally elicited elicited behaviors, behaviors, but but includes includes self-initiated self-initiated endeavours endeavours founded founded on on internal internal motivations motivations or or "willed" actions. actions. These These higher higher order order brain brain functions functions can can be be considered as as the the highest highest expression expression of of creativity creativity and and personality, in so so far far as as considered they they truly truly express express field independence independence in behavior, behavior, the the neurophysiological neurophysiological counterpart counterpart of of the the controversial controversial philosophical philosophical construct construct of of freedom. freedom. In "physical" "physical" time, time, present present is an an elusive elusive reality reality caught caught between between time-gone time-gone and and time-to-be. Yet, Yet, in the the time time experienced experienced by by the the mind, mind, this this present present supports supports the the time-to-be. field of consciousness, and forms the building blocks of our personality. The of consciousness, and forms the building blocks of our personality. The more more elaborated elaborated the the building building blocks, blocks, the the more more sophisticated sophisticated will be be the the whole whole
1 24 124
Chapter 4
construction. The The brain brain has has evolved aa solution to to this this elusiveness of of the present, that that we could name name working memory (WM). WM WM can can be defined as present, an an unique unique capacity to to maintain maintain information infonnation for for an an extended extended duration, and and at at the the same time to to process process this this information infonnation at at any order of of complexity during task performance. perfonnance. As such, it can be considered as an extension of of the task present. This WM WM is the brain function which makes the junction junction between past and future. Indeed, by by its capacity to to extend the duration of of of events together. Most instantaneous events, it operates to link sequences of investigations have have emphasized the controlled aspect of of WM. Not surprisingly, the volitional support located in the the support of of this function has been located prefrontal and temporal cortices. In this paper and temporal paper we provide ncuropsychological neuropsychological and physiological evidence for aa twin aspect of of this function, which has been of WM, WM presented as one overlooked, namely an automatic uncontrolled face of of the slave systems of of this of the of the frontal executive. The main support of WM could be the hippocampal system. The automatic automatic WM The integration and novelty detection capacities of (Dcnham & Boitano, 11996) arc twin 996) are of the hippocampus (Denham functions which help both to create a stable state and to to delineate transitions between states. They They allow the construction of a stable representation of of the between of a of sequences of environment and also the linking together of of events. Symmetrically, by its capacity to to reactivate and and reenact memories, Hs provides the cortex with the capacity to maintain together in the field of of of logically related events, and consciousness, clusters of and therefore, gives to Pc the possibility for an hierarchical ordering and linking of of full in particular the sequences of of motor or cognitive actions to corne. come. The importance of of prefrontal cortex as a support of of personality is The of patients with a prefrontal syndrome. It is evident from the symptomatology of largely a consequence of of perturbation of of most of of the higher integrative cortical functions such as judgement, reasoning and motivation, and TM stores and the controlled aspect of degradation of of some part of of the L LTM of working memory. These perturbations amilial, prof essional and perturbations together put ffamilial, professional social lif lifee in jeopardy. Conversely, it is not obvious on the basis of of superficial ocampal lesions induce a deep disturbance observation, that that hipp hippocampal disturbance of of personality. In particular, there is no dramatic extended retrograde amnesia. Nevertheless, three decades of of neuropsychological ncuropsychological observation have made loss than after a ffrontal patent that personality disturbance is hardly less rontal lesion, at Antcrogradc amnesia least in cases presenting a total loss of Hs functions. Anterograde (the incapacity · to acquire new declarative or explicit memories) was of these patients. From these these results was emphasized in the exploration of TM inferred the fundamental function of of the hippocampus in the the process of of L LTM ,
JP. d.P. Banquet Banquet et al.
1125 25
consolidation, but not in L TM storage. These last functions were instead LTM devoted to the cortices. From there, a logical step was to explore the hippocampus-cortical relations. Hippocampus receives information from and sends it back to the totality of of the cortical mantle. Nevertheless, a specially close relationship is established with the prefrontal cortex. The cooperation of of the two systems in such diverse functions as working memory, rule formation, temporo-spatial tcmporo-spatial processing is that it is often difficult to disentangle the specific contributions of of so intricate that the two systems. In this article we make the assumption that some of of the most elaborated functions classically attributed to the prefrontal cortex, such as working memory, sequence ordering, timing, and novelty detection, are in fact of the based on a strong memory and the computational support of claboratod in the prefrontal cortex. From this hippocampus, and then further elaborated perspective, we do not consider the relations between the two systems as symmetrical. We rather think that the prefrontal cortex function is more dependent upon the integrity of of the hippocampal system rather than the reverse. Accordingly, we first explore the hippocampal contribution to WM. We exclude exclude the function of of LTM LTM consolidation. We consider this WM WM function of of the hippocampus, on the the basis of of neuropsychological ncuropsychological evidence, as the foundation of of some of of the higher order functions attributed to the prefrontal cortex. In both both Pc and and Hs WM WM capacity is supposed supposed to be responsible for the of order and time, and some aspects of of spatial processing such as processing of of maps and plans. In spite of of this apparent communality of of the building of function, in particular for time and space processing, we emphasize the functional specialization and complementarity of of the two structures: 1 ) Transient intermediate memory registers in Hs compute auto- and 1) of data on an intermediate range hetero-correlation and are suited for fusion of of minutes or more), and therefore allow for match-mismatch (of the order of of magnitude. Further, fast comparison on a window of of time of of this order of learning takes place in the hippocampal synapses, at the limit in one trial. These capacities together make make the the Hs ideal for the fast and and transient learning of 992a,b; 993a,b; 994) of temporal sequences (Banquet & Contreras-Vidal, 11992a, b; 11993a, b; 11994) or the multiple views of of a global scene (Gaussier & Zrehen, 11994a, 994a, 11995; 995; et aI., al., 1997a,b). Gaussier, Joulain, Revel & Banquet, 1996; 1 996; Gaussier et I 997a,b). the activation and the the propagation of of goals at at Pc 2) Complementarily, the of sub-cortical structures like the amygdala, in level, under the influence of the control power to design goal-oriented, particular, endows Pc with the of actions (Zrehen & Gaussier, 11997). hierarchically ordered sequences of 997). The
126
Chapter 4
of the Pc structures with different subcortical subcortieal structures give Pc the links of possibility for both changing goals as a function of of the context (link to Hs) and and selecting the appropriate programs for these goal implementations (links to the basal ganglia). Learning there is slower than in Hs, I-Is, and therefore needs of Hs. Symmetrically, Synunetrieally, just as Hs input inputeither repetition, or the mediation of output is in direct relation with perception and cognition, cognition, the Pc input-output output is directly connected to cognition and action. Due to this orientation to action, and also to its L TM capacity for storing sequences, Pc is the most plausible LTM site for recombination of of novel sequences of of events on the basis of of previous learning and present context, and, therefore, the seat of of creativity. Function: An Extended View Hippocampal Function:
Combined ncuropsyehological neuropsychological and biological evidence suggests a fundamental but selective role of of the hippocampal system (Hs) in some forms of learning. Hs is necessary for rapidly (one exposure at the limit) forming of declarative, explicit long-term memories (Banquet et ct aI., al., 11997), 997), but it appears to be unnecessary for the progressive acquisition of of procedural, implicit memories. Recall of of previously acquired declarative memories becomes gradually independent of of Hs itself, suggesting a graded process of of consolidation of of traces which would be be stored in another structure, plausibly cerebral and/or cerebellar eerebellar cortices. In the study of of declarative explicit memory and WM, and particularly of of Hs in this type of memory, human neuropsychology leads animal the role of of type of human neurophysiology. This unusual state of of affairs raises the problem of of the transposition to the animal domain of transposition:to of concepts specifically coined for human cognitive functions. Integrating in one single model concepts from neuropsychology neuropsyehology and animal neurophysiology implies the implicit assumption of of a continuity or even similarity (but certainly not identity) between memory processes taking place both in human and nonhuman primates or even lower order species. Anterograde/retrograde Anterograde/retrograde amnesia. amnesia. With this caveat, most authors agree that a major milestone in memory research was the report by Scoville and Milner ((1957) 1 957) of of a dramatic but selective impairment in memory consecutive to a bilateral ablation of of hippocampus and related structures in medial temporal lobe for the sake of of an otherwise untractable untraetable epilepsy. This type of of memory impairment in the patient HM contrasted with the preservation of of skill learning and priming effects, and more broadly all the capacities labelled as procedural, implicit memory (Schacter, Chu & Ochsner, 11993). 993). Most
J.P. Banquet Banquet et et a/. al. JP.
1127 27
authors draw draw two conclusions conclusions from from these those results. results. First, First, some some (declarative) (declarative) authors of memory are are initially initially stored stored in in Hs, and and then then gradually transferred transferred forms of from there to to other other more more permanent permanent sites sites of of storage, storage, as as neocortex. neocortex. Second, Second, from other (procedural) (procedural) forms forms of of memory memory are are directly directly stored stored in in cortical cortical areas. areas. other of neuropsychological neuropsychological results. results. But But in in These theories grossly capture the gist of some sense sense they are are partially wrong. Progress mostly mostly in in neurophysiology, neurophysiology, but but some also in in neural imaging imaging and modelling, allows allows for for more more precise precise statements statements locus and and the nature of of stored stored information, in in spite spite of of the fact fact that about the locus still remain aa the precise modalities of consolidation of these types of memory still mystery. store and a multirange buff buffer. Hippocampus: A link-operator store er. In particular, despite the contention of many neurophysiologists (Horel, 11994; 994; O'Keefr & Nadel, Nadr 1978; Burgess, Reece Rr162 & O'Keefe, O'Kccfr 11994) and also some some O'Keefe 994) and modellers, we will argue that Hs does not not"" play a central role in the primitive events...". storage and recall of the content of specific episodes and events ...". Only the "link operators" between compressed hippocampal connectivity or "link conical activation patterns are transiently stored stored in Hs, and representation of cortical secondarily transferred and developed in the cortex under Hs control. Place cells in the rat and view cells in the monkey are a typical illustration of these processes of temporo-spatial correlation-integration that take place in Hs and lead to a highly symbolic representation of the surrounding world. Yet, the full-fleAged full-fledged memory traces are initiated and finally stored at the cortical level. Cortex is the alpha and the omega of our souvenirs. There is no need and no capacity for transfer and storage of the full traces in Hs. Preliminary evidence also suggests that even if procedural memory does not rely on Hs for its consolidation, consolidation, it nevertheless implies complex cortico-subcortical conico-subcortical circuits involving in particular basal ganglia whenever motor responses are implicated in the learned processes. processes. Whatever the cortico-hippocampal mechanisms involved in long-term consolidation of declarative memories, experimental results suggest that the role of the Hs extends far beyond that of a transient LTM LTM store, for the time required for cortical cortical trace trace consolidation. consolidation. It seems to be implicated even during the very input. very first wave of cortical cortical processing triggered by by stimulus input. Even in the simple role of witholding and buffeting buffering information there is neuropsychological neuropsychological evidence from amnesic patients patients that Hs is already necessary necessary in the short-term range, as far as information excee.As exceeds the STM STM span. span. This This should not not be be surprising since the activation activation lag lag of the the Hs Hs compared compared to to primary primary cortical cortical areas areas in response to an external external input does does not exceed one or or aa few hundred hundred milliseconds. milliseconds. A A strong conclusion, conclusion, in
128
Chapter 4
and interactively. consequence, is that both systems process in parallel and Similar type of of evidence argues in favor of of a crucial role of of Hs in the intermediate intermexiiate range as a WM WM buffer.
A unitary mechanism supporting M consolidation. supporang WM and LT LTM consolidation. Its multirange temporal capacity makes the Hs act as a buffer that can reenact activation patterns of of information at cortical levels, not only during the of L LTM lengthy process of TM consolidation, but also when functioning as an WM, during the moment intermediate (minutes time-scale) register supporting WM to moment operation of of current tasks. This last hypothetical function is an automatic working memory. These two conceptualised in the model as an proposed functions are: - first, complementary, the process of TM consolidation being engaged of L LTM only if WM WM processing has transformed a short-lived trace to a transient LTM LTM trace; - second, they are based on identical, or at least similar physiological processes, namely reinstatement or reenactment of of electrical patterns of of cortical activity either spontaneously or in reaction to a cortical cue. This unique process allows the network activation, according to its locus of of initiation, either to trigger buff ering and rapid learning of buffeting of information in hippocampal subsystems (if the focus of of activation is first cortical) or to reactivate cortical patterns, and theref ore reinstate recent memories (if the therefore initial focus of of activation is endogenous in the hippocampus). of reactivation of of cortical activity patterns, or These processes of reenactment of of recent memories, directly derive from the capacity of of the Hs to function according to two two distinct, complementary modes: - first in a read mode, when it registers and processes external, cortical information; - then then in a print print mode, when it "endogeneously" "endogeneously" or reactively reinstates the corresponding patterns of of activation either locally in the hippocampal subsystems, or in both the Hs and the cortex. These two modes correspond to two clearly defined electrophysiological patterns in some animal species, theta 989). In primates only sharp waves have been and sharp waves (Buzsacki, 11989). consistently individualised. individualised. It is claimed that that this peculiarity corresponds in fact to an extension of of the read mode which takes place not only as a consequence of of the the physical exploration of of the environment (theta phase), phase), but also as a follow up to any endogenously generated "cognitive processing" processing".. The The reactivation function is related in the model to the the internal "bursting" capacity of of the CA3 pyramidal neurons, and and their collective capacity to synchronize under the modulation of of septal inputs. ,
-
-
-
J.P. Banquet et af. al. JP.
129
Hippocampus Hippocampus as a temporo-spatiaf temporo-spatial processor. processor In In relation relation plausibly plausibly with with this transient buffer function, behavioral and neurophysiological results point of the Hs in learning the temporal order of of serial to a more cognitive role of function could be sequences, at least at a low level of processing. This function illustrated in the process of of place recognition in lower species. Yet, spatial illustrated of this learning is only one example (possibly the best in some species) of multidimensional correlational correlational processing. Further, conditioning conditioning literature multidimensional of some parts of of the hippocampus hippocampus in learning points to an important role of temporal intervals intervals between between events, and more generally durations durations or timing. temporal of these these functions seem to be based based on very basic processes of of Some of differential synaptic plasticity (Granger, Whitson, Larson & Lynch, 11994) 994) as of the correlational learning capacity attributed attributed to most of of the a result of hippocampal subsystems, applied to successive events. Some others, such as hippocampal population coding by cells endowed with different different timing, could result from population that we locate in the dentate dentate gyrus in our model. In this view of of the dynamics that hippocampal function as combining combining and contrasting contrasting fusion-integration fusion-integration and hippocampal match-discrimination, place place recognition recognition appears appears in many respects as a a match-discrimination, degraded capacity to degraded by-product by-product of of the the Hs Hs capacity to register register sequences sequences of of events. events. declarative memory consolidation consolidation function attributed attributed to The long-term declarative LTP current focus of in the present present paper, we L TP is the current of much investigation. But, in compare and and contrast contrast the functions, functions, in particular particular WM, of of Hs and Pc. By compare neuropsychological and and physiological physiological arguments, arguments, we make a case for the neuropsychological automatic working memory memory function function supported supported by ITM automatic working ITM (intermediate-term (intermediate-term memory), operating operating in parallel parallel to to and and in relation relation with the cortical cortical WM, WM, as aa particular slave system of the frontal particular slave of the frontal executive. We We sketch sketch aa comparison comparison and and aa unification unification of of the experimental experimental results in both human human and and nonhuman nonhuman primates. primates. We We contrast contrast and and relate relate this WM WM function with the the more classical classical LTM LTM consolidation consolidation function function attributed attributed to to Hs. We We also provide provide physiological and behavioral evidence evidence for the the implication temporal order order and behavioral implication of of the Hs Hs in temporal sequence sequence processing the basis neural network processing and and timing, which are are the basis for our our neural model implementation. implementation. Working Working Memory Memory as Both Both a Cortical and a Hippocampal Hippocampal System System The existence The existence of of aa graded graded retrograde retrograde amnesia amnesia is aa strong strong argument argument in favor favor of of aa transient transient LTM, LTM, probably probably based based on hippocampal hippocampal LTP LTP and and involved in constant type in LTM LTM consolidation. consolidation. We We make make here here aa case case for aa shorter shorter time time constant type of WM being being based based on both both aa cortical system system or or systems, and and aa of memory, WM hippocampal and brain hippocampal ITM. Neuropsychological Neuropsychological and brain imagery imagery arguments arguments in favor favor
1130 30
Chapter 4
of STM and and ITM functions of of hippocampus are important for our model, of which is illustrated in Figure 11.. Since there is a general agreement on the role of the associative cortex, either prefrontal or temporal, in WM WM function, we of will mainly emphasize the arguments in favor of an hippocampal contribution to the WM WM function.
Psychological arguments f or a slave amodal automatic Working Memory for Baddeley ((1986) 1986) proposed a multicomponent model of of WM WM as a the shortcomings of of the the modal model of of Atkinson and Shiffrin substitute for the ((1968), 1 968), which implied a unique STM store as a necessary passage to L TM. A LTM. could not simultaneously function as an adequate WM, WM, and single STM store could was therefore evolved into a multicomponent model. Still, WM WM was viewed by Baddeley as a single common resource, with a limited capacity. Working Worlang Memory as control and slave systems. systems. The definition of of WM WM storage of of information in connection with performing other, as a "temporary storage more complex tasks" is vague enough to allow for any possible extensions or modifications of of the model. Baddeley assumes a limited-capacity attentional controller, the central executive, that supervises two slave modality-specific systems, the visuo-spatial sketchpad to hold and manipulate visual and spatial images, and the articulatory loop, to rehearse speech-based information. The articulatory loop manipulates memory for sounds. It comprises a memory of one or two seconds, store for holding phonological information for a period of process (Baddeley, 1986). Overt or coupled with an articulatory control process articulation allows both refreshing the auditory memory covert subvocal articulation traces and also feeding the phonological store with phonologically translated visual information. The temporary storage of of visual information in a visuo visuospatial sketchpad, would imply an occipital system involved in the visual spatial aspects and a parietal system involved in spatial coding, and also possibly a aspects frontal lobe participation (Goldman-Rakic, 11988). 988). A related but more of active memory. comprehensive construct proposed by Fuster (1995) is that of This is a state rather than than a system of of memory, and and includes a widely This distributed and changing representational network in the awake organism. not presuppose any mental Active memory therefore includes WM, but does not or cognitive operation. Until recently, only vague reference was made by Baddeley to the underlying brain structures supporting either executive or slave systems, but WM function was under the control of of conscious awareness and implicitly WM therefore had plausibly a cortical location. The same is also assumed for
Motor Motor Areas Areas
Prefrontal
Actions Actions
-
"Where" " Where"
J.P. Banquet et al.
Parietal Parietal Areas Areas
Primary Primary Visual Visual Areas Areas
w .-.
131
A global view of of cortico-hppocampal cortico-hippocampal relations. relations. In the model, the hippocampus is used as a plastic fusion Figure 11 .. A operator and Workmg Working Memory (WM). (WM). Its connection with the limbic system allows regulation of of vigilance and levels. learning levels.
Chapter 4
132
1988). Yet, a fundamental ambiguity animal WM (Goldman-Rakic, 1988). definition of WM WM,, and the explanation explanation of of the results concerning the precise defi nition of derived from the paradigms designed to probe WM theory, remains unresolved if one does not refer to the brain structures that support WM: "There is clearly a danger that a concept such as the central executive may reflect nothing more than a convenient homunculus . . . " (Baddeley, 995). reflect homunculus..." (Baddeley, 11995). Furthermore, Furthermore, the sole reference to cortical structures is not enough to account acx,ount for all the experimental data. More recently, results of of brain imagery and animal neurophysiology 995). Unfortunately, most of incorporated into the theory (Baddeley, (Baddeley, 11995). were incorporated to date, either with metabolic or electrical brain imaging with the experiments to cally designed event-related potentials (ERPs), use paradigms that are specifi specifically for the study of 1986), rehearsal of the components of of the model model of of Baddeley ((1986), 993). Furthermore, the systems in particular (Paulescu, Frith & Frackowiak, 11993). 99 1 ; Ruchkin et aI., ERP approach (Ruchkin, Johnson, Canoune & Ritter, 11991; al., of the cortical mantle, than of of the 11992) 992) is more adapted to the exploration of deep structures. structures. Therefore, these studies confirm the involvement involvement of of cortical structures in different WM paradigms. Yet, more classical recall paradigms, not formally requiring rehearsal, show joint cortical and hippocampal activation (Squire et aI., 992). al., 11992). Working Memory as a hippocampal automatic slave system. Neuropsychology and imagery suggest suggest an automatic component of of WM WM,, supported by an intermediate register located in Hs, along with the controlled, supposedly cortical, cortical, component of of the WM system explored both in humans 97 1 ; Olton, Becker & (Baddeley, 11986) 986) and animals (Fuster & Alexander, 11971; (Baddeley, Handelmann, 11979). 979). Like the cortical controlled WM WM,, this system is endowed with both storing and processing capacity. It is supposedly based upon several systems: of intrahippocampal or hippocampo-cortical hippocampo-cortical - First, a complementary set of loops; Second, a battery of of memory registers covering a large temporal - Second, spectrum; facilitation - Finally, rapidly instantiated but transient synaptic facilitation demonstrated in different hippocampal subsystems (Buzsaki, 11988; 988; Jones, 11993). 993). Moreover, the operation mode of of this intermediate system supporting supporting WM is supposed to be just a restriction restriction to the intermediate range (minutes) (minutes) of the more comprehensive process of L TM consolidation generally attributed to LTM Hs. This general process consists of maintenance and/or reenactment of -
-
JP. d.P. Banquet et al.
1133 33
cortical patterns of activation by reverberant activity between reciprocally that connected systems. This hypothesis extends to WM the dichotomy that already exists in
LTM L TM between declarative explicit memory and procedural
of this hippocampal implicit memory. Psychological arguments in favor of automatic component ofWM of WM are provided below. automatic Refreshing either the visuo-spatial sketchpad or the phonological store
process.. So too is performance by rehearsal every second or so is a controlled process of the concurrent cognitive task, usually verbal, required from subjects in of WM paradigms. Supposedly, the central executive responsible for planning, WM coordination of of inf information of these strategy selection, and coordination ormation is monitoring both of tasks. Thus, several controlled controlled processes work in parallel. Shiffrin and
((1977) 1 977) demonstrated the very limited capacity for controlled human brain. They made made a good case that only several processing in the human automatic processes, or at at best one controlled and one automatic automatic process, automatic performed could be perf ormed in parallel. Yet, surprisingly, the usual paradigms testing WM are tractable WM tractable without without overwhelming difficulty even by patients or aging (Br6bion, 11994). There could be several reasons for that, that, not subjects (Brebion, 994). There and more or less implicated according to tasks and necessarily exclusive, and subjects: The so-called complex cognitive task task may be largely automatized in - The spite of particular when it involves verbal comprehension. spite of its complexity, in particular would be relatively free to ffocus of the to toThus, attention would ocus on active rehearsal of be-memorized material. to be memorized may may be more or or less related - Alternatively, the material to to the task not really competition and interference task to be performed, performed, so there is not interference between withholding information a nice interleaving information and processing it, but but a and integration integration between the two tasks. tasks . This type of of paradigm (where the information to be remembered is related to the cognitive process process in progress) is certainly close close to actual WM natural conditions. WM operation in natural Certainly, one cannot deny the the reality of of the rehearsal process. But we contend that of limited that this low level and rote strategy is expensive in terms of controlled everyday situations. controlled processing processing capacity, and and unnecessary in most everyday Whenever Whenever possible, possible, subjects subjects resort resort to to cognitive strategies aiming to to create create supraordinate supraordinate chunks of of items or events, in order to to increase the limited capacity capacity of of STM. STM. However, this this is still a a controlled controlled process. process. We We therefore therefore suppose suppose the the existence of of an an automatic automatic support support to to WM, WM, an ITM, which does not need register is in such with the not need rehearsal. rehearsal. This This register such a a functional relationship relationship with various various cortical cortical areas, areas, that that it may may readily refresh refresh recent memories relevant for the task the task in progress, progress, by by simple maintenance and/or and/or reenactment reenactment of of the the Schneider
-
134
Chapter 4
corresponding
activation
patterns
in
the the
cortical
populations.
This
reactivation does not not preclude aa state of of priming or subliminal activation of of the cortical areas areas to-be-reactivated due to their recent activity. This function onns, in the different processes of is exactly what what Hs I-Is supposedly perf performs, of long long-
term consolidation of of the declarative episodic and factual semantic memory tenn traces. Both
WM TM refer in some sense to an episodic, WM and transient transient L LTM
context dependent, or a factual type of onnation to be memorized. of inf information
the only important distinction between the two is that the trace Nevertheless, the LTM, L TM, while trace (and eventually the activity pattern) pattern) in WM WM is just a a candidate for the trace L TP and undergoes a nnanent L TM store. It is LTP a test of eligibility to pe permanent LTM TM consolidation theref ore further hypothesized that the two processes of therefore of L LTM of WM WM refreshing on the other, are roughly similar and on one hand, and of of memory complementary. Only the duration during which the two types of or the subject diff ers. This is of traces keep relevance ffor differs. of the order of of minutes for for working memories, but weeks or months f or memories that must be "pennanently" "permanently" consolidated in the long-tenn long-term stores. They are also transformation LTM complementary because transf onnation into transient L TM can only be considered if the long-tenn onnation to be stored has been long-term relevance of of the inf information confinned confirmed by the processing perfonned performed either in the intennediate intermediate hippocampal store or in the cortical system of of WM, or in both. The neurophysiological counterpart of ormation selection f or long-tenn of this inf information for long-term storage would correspond to the transition from loop iterative and punctual activation or even more more important short-tenn short-term synaptic facilitation (based on short-tenn short-term potentiation: STP) to a a transient long-tenn long-term synaptic synaptie facilitation (based on L TP). Both types of erent levels of LTP). of learning are documented at diff different of hippocampal subsystems (Jones, 11993; 993; Buzsaki, 11988). 988). "long-term potentiated" will certainly be transf transferred already "long-tenn erred to
f Working Memory and transient Personality correlates correlates oof transient LTM One of of the the main main implications of of our model is the augmentation of of WM WM beyond the classical, cortical, rehearsal-expandable STM range, by a rehearsal-independent register mostly based based on ITM in the hippocampus. This
TM based TP. It is noteworthy that ITM ITM itself is distinct from transient L LTM based on L LTP. personality traits
seem to to
correlate diff erentially with these memory differentially
components. Both extraversion and neuroticism relate, in experimental studies, to perf onnance on retention tasks. Howarth performance Howarth and Eysenck
(1968)
showed that, in retention of of paired-associates, extraverts show superior recall over retention retention intervals up to
5 minutes or so (in the ITM range), but
J.P. J P. Banquet et al.
1135 35
thereat~er introverts show reminiscence and increasingly better performance thereafter than extraverts. These results are difficult for current WM theory to explain, the typical durations because the advantage of extraverts persists beyond the associated with the short-term components of WM. The present analysis STM/ITM. Plausibly, on the basis of of suggests extraverts may have superior STMIITM. different performance equilibria in neuromodulators (in particular noradrenergic, dopaminergic, or serotonergic), cortical STM is more stable, and the Hs tends to operate so as to refresh these memories in extraverts. LTM Conversely, the L TM consolidation processes would be more powerful in introverts. It is already admitted that emotional charge associated with events (and supposed to be larger in introverts than extraverts) favours LTM same emotional component is detrimental consolidation. Could it be that this same to the automatic (non-rehearsal based) maintenance of information in STM/ITM? Consistent with this hypothesis, extraverts tend to show superior STMlITM? recency in free recall, a function normally attributed to STM, but effects of WM tasks requiring active processing are unreliable extraversion on WM (Matthews, 11992). 992). It is well-established that trait anxiety and neuroticism are associated of STM (Eysenck, 11982). with impairment of 982). In this case, effects are stronger WM tasks than for passive storage (e.g. Darke, 11988), for active WM 988), implying that this anxiety effect may relate preferentially to the controlled, cortical of WM. This psychopathological result is also consistent with our component ofWM. interpretation for the effects relating to the introvert-extravert dimension.
Neuropsychology, Brain Imaging Working Memory Imaging and Working an automatic component of of Neuropsychological evidence in favor of of an WM based on hippocampal structures remain important. Indeed, brain WM based during WM paradigms are are still scarce, and imaging experiments performed during WM paradigms for most of of them oriented to testing the controlled aspect of of WM WM according to Baddeley's Baddeley'S theory of of the executive controller. First, First, Brown Brown (1958) ( 1 958) and and Peterson and and Peterson Peterson (1959) ( 1 959) showed that that witholding information in STM STM is dependent on rehearsal, rehearsal, and and rapidly lost if if active rehearsal was was prevented. This fact lies at at the basis basis of of the the distinction between immediate immediate STM STM and and primary memory (James, 1890) which results from STM STM extension extension due due to to the rehearsal rehearsal process. process. Second, responses of of normal subjects to different recall or to different or recognition tests, tests, show in the absence of of rehearsal, rehearsal, aa residual residual memory which the STM decay tends to to asymptotically. This This residual memory is suppressed suppressed after after bilateral bilateral Hs Hs lesion. Therefore, Therefore, we
1 36 136
Chapter Chapter 4
attribute it to to an an hippocampal hippocampal ITM ITM component. Classically, in amnesia attribute to aa mextial medial temporal temporal lobe lesion, STM STM in the the sense sense of of immediate consecutive to memory is fully fully intact, intact, in contrast contrast with the the loss of of new acquisitions acquisitions in LTM (Baddeley & & Warrington, 1970; Cave Cave & Squire, 1991). 199 1). The The declarative LTM picture picture sketched from the study of of amnesic patients is much more subtle than black and and white portrayal, in particular particular when one takes into account the this black role of of the extent of of the lesions in determining the depth of of the deficit. deficit. The The following neuropsychological arguments derive from recall and forgetting curves, or visual recognition obtained by classical tests of of STM STM and LTM both and amnesic patients. In spite of of the thorough and LTM both in normals and analysis of of these results, some repeatedly confirmed revidence has been either overlooked or accounted for. The patients show aa deficit of or even not not accounted The patients of learning and recall for verbal as well as nonverbal material. In terms of of our model, these tests explore in fact both STM these STM and intermexfiate intermediate term memory (ITM), of recall. Since the depth and nature of of the according to the variable delay delay of memory deficit depends on the extent and location of of the Hs lesions, these parameters are taken into account in the interpretation of of the results. Bilateral extensive lesions with complete loss of of hippocampal function will therefore be treated separately from unilateral lesions and/or partial loss of of hippocampal function. But, for the purpose of of separating hippocampal from cortical components of of WM, WM, the most important parameter parameter is whether or not the subjects are allowed controlled rehearsal, as supposed by the articulatory loop in the WM WM model of of Baddeley (1986). Rehearsal allowed. allowed. The results from subjects with complete loss of of Hs function (like patient HM) HM) will be emphasized since a normal or close to normal performance on tests of of STM or ITM in this case would imply all the hippocampal ocampal function is partially more a normal performance when hipp preserved. In this case, the only possibility will be to attribute the performanc, rmance to STM or to the controlled, cortical component corresponding perfo of of WM, since rehearsal is allowed and Hs function is lost. This complete loss of Hs function results either from bilateral and extensive resection of of the of medial temporal lobe, as in HM's case, or from unilateral resection associated (post-mortem diagnosis) or with a severe either with a severe degeneration (post-mortem dysfunction (EEG recording) of the spared medial temporal hemilobe (respectively cases PB and HF). HF). - First, rehearsal is spontaneous, as with verbal material well within the memory span and made of consonant trigrams presented in a variable delay matching-to-sample task (Sidman, Stoddard & Mohr, 1968). 1968)..Then Then STM 1 890) has range can be extended theoretically at will in what William James ((1890) -
JP. J.P. Banquet Banquet et a/. al.
1137 37
named a primary primary memory (see Figure 2d), even for patients patients with total loss of of hipp ocampal function. hippocampal - Second, the material material is easily verbalizable verbalizable but but at at the the limit of of the memory span as in the short version of of the visual maze (Milner, Corkin & Teuber, 968). This test Teuber, 11968). test requires the the memorization of of a sequence sequence of of turns (8 choice points). The patients patients cannot cannot learn the task task even after after many trials. There is complete disruption disruption of of the memory process. Thus these patients patients face an actual cognitive defect, related to the incapacity to simultaneously maintain maintain and and organize accessible accessible information (type of of turns to make), and and to the task task (topographical (topographical translation translation of of the the turns turns on the maze). This implement the is a genuine definition definition of of a WM WM task. In this situation, rehearsal rehearsal is not sufficient due to the memory load at the limit of of the STM span. Furthermore, practice does not improve performance. There is no evidence of of learning over 1125 25 trials, as if if the subject was unable to devise a learning leaming strategy in order to split a too difficult task into accessible accessible subgoals. - Third, if if the material to be memorized is not naturally or easily verbalizable verbalizable (elliptic geometric forms with one variable variable radius to compare to of the STM range is not possible a sample after various delays), the extension of patients (Figure (Figure 2d). Even Even with with very sensitive measures, measures, a limited residual in patients control 6-24 control of of the sample stimulus on the performance performance is restricted to the 116-24 see range, i.e. the the classical decremental STM range. Remarkably, this is not sec subjects or even 99-12 the case with normal subjects 1 2 year old children who demonstrate of performance performance deterioration at delays up to 40 msecs. Similar no evidence of results of of poor performance performance in the case of of extensive bilateral lesions have results of geometrical items (Milner et aI., al., been found in tests involving other types of 11968). 968). Finally, these defects are further corroborated corroborated and extended to the immediate STM by the classical classical test of of digit span. Digit sequences of of various lengths are presented presented at at a typical rate rate of of one digit per second. Subjects speak at their own pace. The percentage of of strings correctly out recalled digits at recalled (Figure 2c) demonstrates demonstrates an inverse linear relation between the recall performance performance and the length of of the string, similar in amnesic and controls. But, consistently (Drachman 966; Baddeley & Warrington, 11970) 970) (Drachman & Arbib, 11966; subjects perform worse (in fact, are are completely unable amnesic subjects unable to recall) on sequences exceeding the normal memory span (7 items). These last results of point to another critical factor in recall performance, besides the delay of retention, namely the memory load. As for the delay duration, medial important role with regard to memory load. temporal lobe seems to play an important Furthermore, this very last result result shows clearly that medial medial temporal lobe Furthermore, -
138
. � 00
1 00 90
_ - - 0- -
AMNESICS CONTROLS
80 I U
�
IZ
� LU Q.
90
70
tf) o
60
I-
�
u
�
50 40
u IZ LU
30
1i LU
Q.
20 10
80 70 60
550040. 40
_ AilNESICS AMNESICS
,, ,,
----O 0--- CONTROLS CONTROLS
B
. , ,
Q.
'.
" 0..
'.
'
0. , -," "
.
'.
'.
'
0- - - - - - - - _ _ _ _ _ _ _ _ _ _ , _ _ _ _
30 20 10
o +-�------�--�� o 2 3 4 5 6 7 8 9 10
ITEM ITEM NUMBERS NUMBERS
o �------�--� 30 o 60
DELAY DELAY (secs) (sees)
Figure 2. A.Two component component free recall recall task. task. Mean Mean percentage percentage correct correct recall (immediate (immediate recall) recall) as as aa function function of oforder order Figure ofpresentation presentation for subjects presenting partial lesions of ofthe the hippocampus or or Korsakoff Korsakoff syndrome s}1ldrome (amnesic) (amnesic) and and of B. Short-term Short-term forgetting. Short-term retention of ofword triads triads for amnesic amnesic and normal controls. controls. B.
Chapter 4
c: o u
-
loo; 1 00
A
1 00
_ - -- 0O--
90 90. (f) V) (!) 0 z
5
�
II(f) V) Ilu V w W IX IX 0 0 U
v
+ z
5
80-
0 "
9
70 70-
"
,
DELAYED DELAYED MATCHING MATCHING ADJUSTING ADJUSTING DELAY DELAY "
,
, ,
\,,
6 600 50 50-
, , "
A H.M. TRIGRAMS
'<4" ."
40
40-
" ""
30 30. 20 20
o
c
, .. "''' ..... ,
'0- - - - _ _ _ _ - - - - 0
-
110. 0 0 0
-
5 5
6 6
7 7
LENGTH LENGTH OF OF STRING STRING
8 8
B H.M.
-
6
TRIALS TRIALS
3.P. Banquet et al.
Iz w W u IX w c.. L
80
4MNESICS CONTROLS CONTROLS
Figure or digits Figure 2. 2. C. C. Digit Digit span: span: Immediate Immediate memory memory ffor Qgits in in amnesic amnesic patients patients and and normal normal controls. controls. Mean Mean percentage percentage of of sequences 970, with sequences correct correct as as aa function function of of sequence sequence length. length. (Adapted (Adapted from from Baddeley Baddeley & & Warrington, Warrington, 11970, with permission). permission). D. D. Adjusting Adjusting delay delay procedure procedure of of the the delayed delaycd matching-to-sample matching-to-sample task. task. Delay Delay periods periods on on consecutive consecutive trials trials of of the the adjusting adjusting delay delay series series allow allow aa quick quick check check on on the the limits limits of of the the subject's subject's ability ability to to remember remember the the sample. sample. On On combination combination of of three three consonants consonants trigrams, trigrams,
HM HM had had no no difficulty difficulty in in remembering remembering the the trigrams trigrams for for as as long long as as 40 40 secs. secs.
On sec with with best best On the the set set of of geometrical geometrical stimuli stimuli (eight (eight different lfferent ellipses) ellipses) progression progression was was reduced reduced to to steps steps of of 11 sec
performance performance at at 55 sec sec delay, delay, achieved achieved once. once.
CL
w
V3
1140 40
Chapter 4
of information out of of the very limited capacity helps to buffer overflow of attended attcndefl STM store. It can be argued in this case that rehearsal is not of the overflow of the immediate memory efficient anymore as a consequence of span. Still, normal subjects demonstrate a residual mnemonic performance in spite of of this overflow. Taken together, these results confirm the integrity of immediate STM for material within the immediate STM span in amnesics. They also confirm the efficiency of of the articulatory loop as a rehearsal device allowing the extension of of the immediate STM span into a primary memory, either for verbal material artieulatory loop is intact and or for material easily verbalizable. This articulatory efficient in patients. Nevertheless, the results also point to the fact that the possibility of of rehearsal is not enough to guarantee recall or learning in patients. If If the material is difficult to verbalize and/or exceeds or even is close to the immediate memory span, as in the short version of of the visual maze (Milner et aI., 968), there is a complete disruption of al., 11968), of the memory process. Therefore, for patients with Hs lesions, there is a clear difficulty in fractioning fraetioning a too difficult task into accessible subgoals. Conversely, the results do not give a strong support for a corresponding rehearsal system which refreshes the visual sketchpad when material is not naturally verbalizable. Otherwise, the performance in geometric tasks should have been close to normal in patients. prevented The tests where rehearsal is prevented during the Rehearsal prevented. delay between learning and recall are similar to experiments designed by WM.. Indeed, they combine withholding information with an Baddeley to test WM interfering task. However, they differ from genuine WM paradigms, in that the items to be remembered are not at all related to the interfering task, and also not related to each other. Conversely, in a WM paradigm, the task is not designed to interfere with item memorization. The items may even be related to the task, and task performance is quantitatively evaluated. Also, in a WM task, STM and and ITM components cannot be dissociated while classical WM ' they can in the memory tests by monotonic variation of of the delay of of recall. The results of of these tests critically depend on the extent of of the lesion. --Bilateral Bilateral lesions lesions of of Hs. Subjects with bilateral extensive lesion of of the HM are clinically evaluated as being in a medial temporal lobe like patient HM of the severe, even dramatic, condition. They depend on continuous rehearsal of information to be retained. Catastrophic forgetting is induced by distraction al., 11968). of this (Milner, 11966; 966; Milner et aI., 968). The most dramatic illustration of psychological condition is reported by Milner: the patient HM was able, by devising an elaborate mnemonic scheme, to remember a three-digit number
J.P. Banquet et al. JP.
1141 41
for 115 5 minutes, but forgot it as soon as he was distracted. The question is then why a normal subject after 5 minutes a three digit aider maintaining for 115 of awareness has no difficulty in remembering it, even number in his field of after distraction. Even hypothesising that 115 5 minutes could represent enough exposure for a "permanent" L TM storage, this type of material does not LTM represent any long-term relevance for the subject. There is little chance that HM could the transient storage takes place at cortical level; otherwise patient HM have learned this information, as he did learn from procedural memory paradigms. Since the main difference between normals and patients, in this case, is the presence or absence of the hippocampal system, it is logical to of the recall defect to the Hs and not the cortex. attribute the primary cause of not provide a clue for the the actual mechanisms responsible for the This does not of information in normals. In normal subjects, continuous rehearsal retrieval of or refreshing memories cannot cannot be directly responsible for this retrieval which looks to us so natural. Indeed, the the very definition of of distraction distraction implies an interruption of the controlled rehearsal rehearsal p process. of information rocess. The retrieval of interruption of the in normals (compared with the non-retrieval in patients with extensive Hs that the continuous process of of rehearsal, or lesions) strongly suggests that of memory has somehow succeeded in rapidly laying down a reactivation of trace, not not at the the cortical level (otherwise both patients and and normals would trace, at the Hs level. The The ITM hypothesis explains have been able to recall), but at of this information in normal normal subjects, in spite of of its irrelevance the retention of for long-term storage storage (LTM and in spite of (LTM consolidation was not involved), and of the interruption (STM was not not prolonged any more interruption of of the rehearsal rehearsal process (STM by rehearsal). Fifteen minutes is well within the WM WM range. This is therefore an argument in favor of of an automatic component to WM. We We will see how recent findings on the multiple-range memory registers operating in the hippocampal hippocampal subsystems, in conjunction with the complex closed-loop system of actual mechanisms responsible of the hippocampus, hippocampus, may provide aa clue to the actual for these these transient transient intermediate-range remembrances (so useful for our our moment-to-moment processing processing of of the continuous flow of of information). A more formal assessment was performed A more assessment of of this this type type of of patient patient was performed in aa visual select from visual memory task, task, based based on on face face recognition. recognition. Subjects Subjects had had to to select from an an array had been array of of 25 faces the the 12 1 2 faces that that had been shown 90 sees secs earlier. Performance was performed Performance fell to to chance chance level in patients patients when the the test test was performed with aa distracting distracting task task interpolated interpolated between the presentation presentation of of the the two two sets of of photographs photographs (Milner (Milner et al., aI. , 1968). 1 968). We We have seen previously how excess in memory load, memory load, or or difficult verbalization verbalization of of the the material to to remember, is equivalent equivalent to to aa prevention prevention of of rehearsal. rehearsal.
142
Chapter 44
- Partial lesion off Hs. The The picture is apparently different when there is a ocampal function. This is the case for patients partial preservation of of the hipp hippoeampal after unilateral temporal lobectomy, or alcoholic patients suffering a Korsakoff syndrome which present an identical pattern of results to the latter 970). As we have seen, delayed recall tests with (Baddeley & Warrington, 11970). interfering tasks present some similarities with the WM paradigms even if they are not identical. The variable delay between item presentation and recall allows a quantitative separation between deficits in STM (20-30 secs), sees), and WM intermediate memory, beyond this range. This is not the case in the WM paradigms, which represent a more natural situation, in the sense that the material to be memorized may be related to the associated processing task. In the short-term forgetting task subjects are presented for a few seconds (e.g. 3 sees) secs) with item sequences (3 words) well within the memory span. They are required to recall the item sequences, after atter delays varying according to trials (0, 5, 10, 115, 5, 30, 60 secs). sees). During these delays they perform a tightly controlled intervening task designed to prevent rehearsal. Manifestly, the forgetting curves present an exponential decay (Figure 2b), and reach asymptote within 30 rnsecs, msecs, which is the maximal range of of STM. For this reason, the corresponding decay can be fully explained by a STM decay. Conversely, the residual information withheld beyond this short-term range is of either the logically attributable to an intermediate store independent of articulatory loop, or any other aware controlled cognitive process, which are both prevented by the experimental design. Therefore the process responsible That the two curves are almost similar in for this memory is automatic. That can be controls and unilateral temporal lobectomy or Korsakoff patients can attributed in both cases to a close to normal cortical function sustaining the attributed hippoeampal ocampal function attributed to STM capacity, combined with a residual hipp the spared side of of the Hs, or to the diffuse and partial lesions of of the Korsakoff syndrome, in a situation where item lists to be remembered are well within the STM span (3 items). This interpretation is strongly suggested by the catastrophic forgetting of of patients with bilateral lesions, when faced with an interference situation. Further, there is also electrical brain imaging evidence that unilateral hippocarnpectomy hippocampectomy alters little or not at all the cognitive ERP that patterns (in particular P300) recorded on the scalp, with the possible exception of of very limited areas of of the temporal lobes where a trend towards asymmetry shows up (Johnson, 11995). 995). of unrelated items In the classical free recall two component task, lists of exceeding the memory span (e.g. 110 0 words) are presented to the subjects at a of the previous experiment. At the end of of the presentation, pace similar to that of
J.P. Banquet Banquet et al. JP.
1143 43
with a variable variable delay, delay, during which which subjects have have to perform an an intervening with task, free recall in any order is performed. The main difference with the previous task is that the memory load here exceeds the memory span. In the absence of an zero delay, immediate recall condition, in spite of the absence intervening task rehearsal after each item presentation is prevented by the of to-be-recalled new stimuli. Very robustly, the continuing presentation of recall curves (Figure 2a) show in controls both a primacy and a recency effect l&st in the list are (Glanzer & Amitz, 11966). 966). The items delivered first and last better remembered than than the ones in the middle. The recency effect is better classically attributed to a persistence of items items in STM. This interpretation is consistent with our model which differentiates STM (present both at cortical and hippocampal level) from ITM (which would be a more specific attribute of Hs). As such, the recency effect disappears both in patients and controls of see delay with performance of of an intervening task is interpolated when a 30 sec between presentation and recall. The primacy effect is also classically of our model, it is attributed to long-term memory. But in the context of supposedly dependent on intermediate-term memory (ITM). This primacy at zero delay even in these patients patients with partial effect greatly deteriorates at of medial temporal lobe functions (Figure 2a). It becomes preservation of see delays, while it still completely abolished in the same patients after 30 sec normal controls. Thus, when the memory load exceeds STM span, persist in normal of the medial temporal lobe function is not not anymore partial preservation of sufficient to prevent intermediate memory deficiency, compared to controls, to as was was the the case case in the short short-term -term forgetting task with a memory load within TM span. the S STM In summary, both patients with bilateral and unilateral lesions of of the Hs present close to to normal immediate STM within the the STM STM for information well within STM span (typically 3 items), and can can further extend this temporal range when span possible and natural (verbal material). Both types of of patients rehearsal is possible present catastrophic catastrophic loss of present of information and forgetting when there is overflow, beyond the the strict STM STM span. This loss of of information is much less systematic and and dramatic are some some indications dramatic with normal subjects. subjects . There There are that the STM span may be reduced in patients. When that When information is well within the memory span patients with unilateral temporal span (and only then), patients lesions or intermediate range or Korsakoff Korsakoff syndrome, preserve close to normal intermediate memory capacity, independent of of rehearsal. rehearsal. This is not the case for bilateral temporal that the side of temporal lesions. It must must be further further emphasized that of the unilateral temporal neutral for the type of temporal lobectomy is not neutral of performance which is the most and the most spared spared (verbal or visual). Thus, both the forgetting curves and
1144 44
Chapter 4
the distinction between b ~ e c n an primacy-recency effect in partial lesions support the LTM, immediate STM and an ITM, rather than an L TM, as was classically stated. Further, the distinctive behavior of of patients with a bilateral extensive ablation Further, of Hs corroborates this ITM as a possible automatic support (non or lesion of rehearsal-based) ofWM. of WM. The development of of brain imaging techniques based either on EEG, PET or functional MR!, MRI, raises the possibility of of a simultaneous and MEG, PET direct investigation of of anatomy and function of of memory, both in normal obtaineA by these techniques vary subjects and patients. The few results obtained according to the experimental paradigms. They lend support to both cortical ocampal components ofWM. and hipp hippocampal of WM. imaging. Electrical brain imaging can be used to Electrical brain imaging. record event-related event-rdated potentials (ERPs). These ERPs can be defined as scalp potentials refl ecting different cognitive processing operations or steps reflecting performed by the brain. There are different ERP responses discriminated on ect automatic of their latency and topography. Some refl reflect the basis of identifi cation of identification of the stimulus, like N200 (negativity at 200 msecs) msccs) Mismatch Negativity; others like Processing Negativity (Naatanen, 982) refl ect (N~tanen, 11982) reflect attended 990) and still attended stimulus processing (Banquet, Smith & Renault, 11990) others like P300 (positivity at at 300 msecs) msees) reflect context processing processing or updating. The The ERP investigation of of WM WM confirms the involvement of of various associative cortical regions, for components later later than modality specific or associative P300, like P600 and over (Ruchkin 99 1 , 11992). 992). Surprisingly, P300 (Ruchkin et r aI., al., 11991, related to to context updating and in particular particular prob�bility probability processing (Banquet, 98 1 ; Johnson & Donchin, 1982) does not refl ect the Renault & Lesevre, Les6vrr 11981; reflect cortical "explicit" component of of WM. These results do not allow the exclusion of of the participation participation in a typical WM WM paradigm, or in other types of of paradigms, of of a deeper structure structure like the Hs. Indeed, scalp electrical recording as in EEG is known to explore predominantly, if not exclusively, the cortical mantle of of the brain. Furthermore, a P300-like P300-1ikr activity has been recorded in the hippocampus (Halgren et 980). But this P300 hippocampal source is r aI., al., 11980). uence on cortical not the generator of of cortical P300s. Further, it has little infl influence P300 generators, since right or left hippocampectomy does not induce any significant cortical asymmetry. There are however two notable exceptions. significant First, far-lateral temporal electrodes (T5-T6) in oddball oddball paradigms present a reduced P300 on the the side of of Hs removal. Second, left hippocampectomy induces a change in P300 behavior rather than P300 asymmetry, along with a deficit in the number of of correctly recognized items in recognition performance deficit paradigms involving stimulus familiarity (Johnson, 11995). 995).
J.P. Banquet et af. al. JP.
1145 45
studies still Positron Emission Tomography. PET or functional MRI studies contribute very little to the solution of the problem of hippocampal memory, despite their long-term promise. involvement in intermediate term memory, Neural correlates of verbal WM involving the articulatory loop were explored by PET measures of regional cerebral blood flow in a task engaging both components of the articulatory loop, the phonological store and the subvocal rehearsal system, compared to a simpler condition engaging only the subvocal al., 11993). rehearsal system (Paulescu et at., 993). This paradigm allowed localization of the phonological store to the left supramarginal gyrus gyms whereas the subvocal rehearsal system was associated with Broca's area. These results are a strong support for the cortical component of WM, and more specifically of the articulatory loop. Nevertheless, since subjects were explicitly instructed to rehearse the consonants to be recognized after a 2 minutes delay period, the of Hs cannot be considered as an argument absence of significant engagement of against the implication of Hs in WM, in as far as a specific memorisation strategy was forced upon subjects. It must also be noticed that the task involved recognition and not recall. of the most salient items of evidence in favor of a combined One of involvement of cortex and hippocampus in Intermediate Transient Memory (ITM) comes from Squire et al. at. ((1992). 1 992). In a delayed (3 minutes) cued recall paradigm without any interfering task subjects subjects learned visually presented word lists ((15 1 5 words). During cued recall, PET scan found significant activation, compared to baseline conditions, of right fight hippocampus and gyrus, plus right (and to a lesser degree leit) left) frontal lobes parahippocampal gyms, (Figure 3). Left hippocampal region region and amygdala did not change their activation level during cued recall. In our model this task corresponds to a of both the delay of of recall and the WM because of test of of ITM supporting WM, memory load, which far exceeds the STM span. The absence of of any interfering task during the delay may have favored rehearsal strategies and a bias towards cortical activation. But this bias may have been limited by the length of of Hs along with the of the item list. list. Nevertheless, the activation of frontal lobes provides strong support for aa WM WM with two components, components, combined in aa closed-loop closed-loop system. The selective cortical and hippocampal, combined activation of of the fight right Hs can be explained as a processing bias introduced introduced by the cued recall (visually presented stem completion) completion) towards processing the visual characteristics of of the word-forms. of Hs in both priming condition condition and cued recall The dual involvement of of the frontal lobes in cued (Figure 3) contrasts with the selective involvement of of memory involved in recall recall. This suggests that that the declarative aspect of ,
Chapter Chapter 44
146 146
A A
Right Hippocampal 30 .=o
t 201-
0
10 _o
.
'E
-
o ~ W
='~
~
~ o Z
// //
. . . . . .
O-
-10 ~. ~ ~
"20
. . . . . . . . . . . . . . .
No response
'
Baseline Priming Memory
B B
Right Occipital 70 60 50 40 30 20 10 No Baseline Priming Memory response
JP. Banquet J.P. Banquet et al.
1147 47
c C
Right Prefrontal 2O
,
,,
,
,
,
,
,
,
10
El
-10
-20
" No
Base'iine Priming
Memory"
response
of the right fight hippocampal (A), right occipital (B), and right fight Figure 33.. Behavior of prefrontal (C) regions, in comparison to the fixation-point control condition, for four task conditions (i: no response; ii: baseline stem completion by first word to come to mind, no stem could form words presented; iii: iii: priming: stem completion by first word to come to mind, half the stems could form words already presented; iv: stems had to be used to recall words from the list presented, half half of of the stems could be completed to form these words). The hippocampal response observed in memory minus baseline subtraction right hippocampal of a 992) did not arise simply as a result of (Figure 2A, Table 11,, Squire et al. 11992) baseline condition. A right hippocampal reduced hippocampal activity in the baseline response was observed both in the priming-minus baseline subtraction and in the memory-minus priming subtraction. Conversely, the right prefrontal response was more specific to the memory task. (Adapted from Squire et al. 11992, 992, with permission).
contrast to the automaticity of of stem-completion priming) is a cortical, (in contrast possibly frontal cortical characteristic. possibly characteristic. At the same time, it justifies the term of automatic component of of WM WM for ITM. The interpretation interpretation of of these results of of the role of of Hs in the is biased by Squire in order to support his theory of aspect of of memory. Thus Hs activation in the simple priming task declarative aspect
Chapter 4
148
is accounted for for by a a covert implication of of explicit memory during the priming condition (Squire et at, al.,
1992). But this explanation is not
parsimonious and is to some extent tailored to fit his hypothesis. Conversely, we attribute attribute the involvement ofHs of Hs in stem-completion priming to the implicit implicitprocedural characteristic of of this task which matches the automatic aspect of of
WM.. Further, if if we make the plausible assumptions, -that the ITM support to WM declarative character of of memory is independent of of the process of of that the fmal of memories is cortical, in consolidation itself, and that final repository of particular frontal, the interpretation of of Squire is not supported by the decrease of of prefrontal prefrontal activation to a level close to that that of of the baseline condition during the priming task.
Neurophysiology: Human Neurophysiology: Human Versus Versus Animal Animal Working Working Memory Memory At variance with some neurophysiologists (Goldman-Rakic,
11994), 994), we
do not not believe that the the criterion "relevant only transiently" is the unique characteristic of of WM WM.. As previously stated, WM WM,, at least in humans, could be characterized by combining temporary storage of ormation and a of inf information capacity f or more complex cognitive processing. Among tests designed to for explore WM WM in in animals, animals, these these two two functions functions are are obvious obvious in in radial radial maze maze
rmance, but perfo performance, but not not so so much in in delayed delayed response response tasks, tasks, where where the the emphasis emphasis is ormation storage. is on on temporary temporary inf information storage. The emergence of of the construct of of WM WM occurred almost simultaneously in in animal animal and human human research. research. In In parallel parallel with with human human neuropsychological neuropsychological studies of of WM WM (Baddeley & Hitch,
11974), 974), the same concept was developed
independently in animal learning to ref er to the capacity to retain inf ormation refer information across trials within a test session (Olton et at, al.,
11979). 979). These authors were
among the first (and perhaps the only ones so explicitly) to propose that that the hippocampal system could be necessary f or animal WM for WM (memory for recent inf ormation of information of current and specific relevance). At variance with this passive witholding function function of of relevant information, we think that that hippocampus hippocampus is also or an ormation selection also equipped equipped ffor an active active inf information selection and and processing, processing, more more in in the the line line with with human human models models of of WM. In In Olton's Olton's radial radial maze, maze, each each arm arm is is baited baited with f ood. Typically, on a series of food. of trials, in the same session, the animal will ood, suggesting a avoid revisiting an arm from which it has already taken f food, retention retention of of behavioral behavioral events events associated associated with with entering entering each each arm. arm. This This WM WM could involve the encoding of of specific episodes associated with specific maze arms. arms. As As such such it it could could represent represent the the ITM ITM equivalent equivalent of of episodic memory. memory. But But
JP. Banquet J.P. Banquet et al.
149
performance could also be guided by the stronger relative familiarity of of cues related to arms more recently visited. In the seventies, recordings performed on awake monkeys trained on response tasks tasks allowed allowed for an extension of of the concept of WM. These delayed response recordings showed that that some prefrontal cortex neurons were activated during the delay period between stimulus and response (Fuster & Alexander, 11971; 97 1 ; Fuster, 11980; 980; Niki, 11974). 974). From the very beginning, these activities were supposed to reflect the cellular expression of of mnemonic processes. The evidence in favor of of a mnemonic process rather than a motor set or any other activity has grown more convincing by the discovery of of some specificity in the response. In particular, particular, in an oculomotor delayed response paradigm, neurons alter their discharge rate only for one or a few target locations, neurons thereby demonstrating a kind of of memory field (Funahashi, Bruce & Goldman Goldman989). This result has been extended to other brain areas, Rakic, 11989). inferotemporal cortex in particular particular during DMS tasks (Fuster & Jervey, 11981; 98 1 ; Miyashita & Chang, 1988). This form of of activity has been qualified as active by some authors, in contrast to a passive form (Eichenbaum, Otto & Cohen, 1994). Passive memory is characterized characterized by a reduced response response to familiar or repeated stimuli. Some neurons in inferotemporal cortex fired much less in response to an immediately repeated stimulus in a serial recognition task (Baylis & Rolls, of the 11987; 987; Rolls, Baylis, Hasselmo & Nalwa, 11989). 989). Item specificity of neuronal response was also demonstrated (Baylis & Rolls, 11987; 987; Miller, Li & Desimone, Desimone, 1993). This paradoxical response has been interpreted as a rapid The decrement in stimulus-elicited firing would reflect a form of of habituation. The of cortical neurons to familiar stimuli. At variance decreased responsiveness of with the active memory this passive response could be interference-resistant and may persist through the presentation of of intervening mismatch choice cues within the same trial (Miller et al., aI. , 11993). 993). Yet, both active and passive representations disappear disappear between trials, suggesting a system reset memory representations al.,, 11993). when the information is no longer relevant (Miller et al. 993). Nevertheless, a gradual cumulative decrement of of response across testing sessions to multiple repetitions of of items confirms our hypothesis of of a graded transition between ITM supporting WM, LTM LTP WM, and transient L TM based on L TP supporting permanent L LTM that this decrement in TM consolidation. It should be noted that response to stimulus repetition is dependent on an automatic, passive type of of processing. When When the the sample stimulus has to be actively maintained in awareness for comparison with several test test stimuli, stimulus repetition does
150
Chapter 4
not induce a decreased response, but an increased one (Miller & Desimone, 1 994). 1994). There are striking similarities between electrophysiological recordings in man, either electrical recording of of ERPs ERPs on the monkey and brain imagery in man, scalp, or metabolic motabolic PET imaging. We have previously mentioned the functional significance of of specific ERP components. In particular, N200 or Mismatch Negativity reflects automatic identification of of the stimulus in an iconic or echoic memory; others like Processing Negativity represent attended stimulus processing; and still others like P300 reflect context processing or updating. When two types of of stimuli are sequentially presented with different probabilities (rare and frequent), the P300 and N200 response to rare and frequent stimuli is at first very similar, but becomes smaller for frequent and larger for rare types of of stimuli. This evolution in response amplitude takes place after only a few stimuli (less than 10) for for P300, but needs more stimuli presentations for N200 (Banquet & Grossberg, 987). This kind of Grossborg, 11987). of probability processing is fully automatic, and in fact better reflected by P300 amplitude when the subject is not aware of 982). of it (Johnson & Donchin, 11982). of response to frequent stimuli can therefore be The decrease in amplitude of to the decrease in neuronal response to to stimulus repetition in compared to monkey. Conversely, the Processing Negativity, corresponding to an attended, selective filtering of of only some preselected type of of stimulus, remains insensitive to event probability and presents an increased amplitude in response to a match (i.e. repetition) condition. The same results hold for PET imaging. A decreased activation was found in the occipital cortical areas in response 992). This response to items that had been recently presented presentexl (Squire et aI., al., 11992). result was interpreted by Squire as a reduction in neural computations required for the processing of of recently presented information. The similarities uman and animal elctrophysiological activities extend to between h human preparatory preparatory set. A contingent negative variation (CNV) paradigm in humans is a formal analog to experiments used in monkey to explore WM WM (Fuster, 11980; 980; Niki, 11974). 974). After a warning warning stimulus S I1 and a delay, an imperative stimulus $2 S2 commands the subject to emit an (usually motor) response. Cellular recording recording experiments in the monkey show that that in the S I-S2 1-$2 interval the CNV presents two distinct components, an early one which has been related to the processing of of S 11,, and a late frontocentral f r o n t ~ t r a l component with a ramp-like activity preceding $2. S2. This late component does not just just correspond to a motor-set, motor-s~, but reflects also timing and perceptual-cognitive set (Ruchkin, Sutton, Mahaffey & 986). & Glaser, 11986).
J.P. Banquet et af. al. JP.
1151 51
Spatio-Temporal Processing in Hippocampus and
Prefrontal Cortex Prefrontal mainly in the Most of the specific functions attributed to Hs and Pc, mainly of temporospatial processing, can be related if not reduced to the domain of global WM capacity proper to these two systems. In fact, some of the paradigms used to probe the WM capacities of these two systems, and already mentioned in the previous paragraph, presented strong explicit or implicit spatial or temporal factors. Nevertheless, in this section we aim at a specific parametric exploration of spatial or temporal functions, and also etc... . Here also we other related functions such as rule formation, planification etc make a distinction between neuropsychological or brain imaging arguments and neurophysiology of the monkey in particular, since both approaches make a specific contribution to the problem.
Neuropsychology and imaging oof spatio-temporal functions functwns in prefrontal Neuropsychology fspatio-temporal cortex and hippocampus that important lesions of of Neuropsychology has provided evidence that of conventional prefrontal cortex (Pc) are compatible with the performance of linguistic, memory or even intelligence tests, which can be sensitive indicators of damage to the temporal lobe. The portraying of of the Pc as the "seat of of of intelligence" was therefore discarded. At the same time, less conventional paradigms brought to light less basic, but no less critical, perturbations of of cognitive and and relational adaptations following important lesions of of the Pc. These patients completely lost the ability to make choices, although sensory and motor sequences remained intact. Reduced flexibility and inventiveness on the the approach new problems, problems, poor adjustment to everyday life and approach to new disturbances of of personality were among the most common signs of of prefrontal dysfunction. These disturbances of of personality are diverse, but could find some of of their common denominators in a disinhibition of of basic drives, combined with with lack of of motivation and and purposeful purposeful behavior. For For these reasons combined these personality changes are are most obvious in family, professional and and social life, all spheres which degree of which require require aa high high degree of instinctual sublimation, not not to to speak speak of of motivation. Since these these familial, professional and and social components of of aa person are are the more distinctive ones, the patient loses his singularity to become common if if not not vulgar: "Gage is no more Gage". In spite of of their limitations, these neuropsychological explorations uncovered uncovered the the structural structural
152 152
Chapter 4
and functional diversity of of Pc, and and sketched a "functional topology" that is presently demonstrated dcmonstratexl by brain imagery. Since Hs seems to contribute to the computation of of the learning leaming of of places and sequences and amygdala participates in the integration of of external and internal drive signals, one can wonder why motivation, based on this drives, intcgrate~ not at at this this subcortical level, but but in the Pc. Several seems to be integrated reflections along these directions can be proposed: - First, goal formation supposes a precise representation representation of of the target level, form and location, which can only be precisely dr defined at cortical level; - Second, goal implementation supposes TM, supposes comparison with L LTM, precisely located in cortex and particularly particularly in Pc, but but not in Hs; and bidirectional relation with - Third, goal enactment supposes a direct and premotor and motor structures structures which is a property of of Pc, but not ofHs; of Hs, - Finally, goal execution needs ne~s control and willed action, which depends on the executive function of of Pc. those reasons seem to to make the prefrontal stage necessary. But the All these most compelling argument in favor of of the necessity of of a prefrontal stage for behavior integration comes perhaps from computational constraints. If If drives were integrated at subcortical level such as Hs for subsequent planning, there would be confusion between perceptual and "executive" representations representations of of sequences of 994b; Zrehen 1995). Veridical of events (Gaussier (Gaussier & Zrehen, 11994b; reproductions of of learned sequences could still be possible. Indeed, short cut relations between Hs and motor programming systems such as basal ganglia do exist through through subiculum and nucleus accumbens. This anatomical connection permits the possibility of of short-circuiting prefrontal prefrontal cortex through the subcortical implementation of of overlearned overleamed automatic programs. But there would be lacking the extra extra degree of of freedom provided by a representation which is directly dependent neither on perceptual units nor from motor units, but lies in fact at 1, below). This extra at the junction junction of of the two two (see Figure 111, degree of of freedom results in a giant evolutionary step because it allows for adaptation adaptation and creation of of new behavioral sequences as needed by specific circumstances. S patial Working Spatial Working Memory, Memory, principal sulcus sulcus and hippocampus. hippocampus. The brain imaging results from tasks inspired by delayed spatial tasks in monkey man either in Brodmann's area 46 which locate spatial working memory in man corresponds to principal sulcus in monkey (McCarthy et aI., al., 1994) or in area 993). Recent results from PET 47 of of the inferior convexity (Jonides et aI., al., 11993). paradigms manipulating spatial 996; spatial working memory factors (Owen et aI., al., 11996; Petrides, Alivisatos, Evans & Meyer, 1993) or simply simulating this spatial
JP. J.P. Banquet et al.
1153 53
information (Burgess et 994) have shown the involvement of ct al., 11994) of the hippocampal hippocampal formation in man during these tasks. rontal cortex. Frontal lobe Temporal ordering and and dorsolateral pref prefrontal lesions present present a specific memory deficit. In delayed comparison tasks, they 982). induce an impairment of of temporal ordering of of recent events (Milner, 11982). subjects have no problem in making the distinction between new and The subjects presented items. This could be a function of of the hipp hippocampus, previously presented ocampus, or even modality specific cortical areas. But they present a specific impairment of items suggesting an incapacity to keep in judging the relative recency of successive successive, trials apart. Considering the pace of of presentation presentation of of the item sequences this recency judgement judgement cannot be based on the rehearsal of of sequences order or the the organization of of material, but rather on the relative temporal order salience of of items in memory. This result is congruent with a defect in STM sees) whereas the distinction between new and old items is (maximum 30 secs) better accounted accounted for by a dysfunction of of Hs-based Hs-basext ITM. A symmetrical better pattern of of deficit in medial temporal temporal lobe lesions, with no impairment on pattern discrimination but a deficit in recognition memory confirms this recency discrimination interpretation. This distinction could be related to the dichotomy in interpretation. experimental psychology between between recency effects (attributed (attributed to STM) and experimental LTM (Atkinson & Shiffrin, 11968). 968). In the primacy effects related to LTM framework of LTM correspond to ITM. STM framework of our hypothesis this L TM would correspond would be cortical, and prefrontal in particular, particular, whereas ITM would have an would hippocampal support. hippocampal The left frontal lobes make aa special special contribution to the the organization and The planning of of responses responses a few few moves ahead. Therefore, the poor performance of of planning patients with unilateral unilateral lesion of of dorsolateral dorsolatcral (DSL) Pc on memory tasks patients rather result result from a failure in control processes of of memory rather than than a could rather deficit in retention per se. The The perturbation perturbation in the temporal ordering of of events deficit aiter DSL DSL lesions is congruent with several interpretations. interpretations. Pc is either the after of processing processing of of such information, or the site of of storage, or both. The site of processing of of this order information in other structures structures such as Hs in processing particular, and the subsequent subsequent storage in Pc cannot be excluded. We We will particular, discuss in the model how this ordering or sequencing function could be the WM WM function of of both both Pc and Hs. dependent on the Other cognitive jUnctions: functions: of behavior and and superior Pc. Patients with a - Feedback regulation of superior Pc lesion present present an incapacity to extract extract information from superior environmental cues elicited elicited by their own responses in order to regulate environmental behavior, in the sense of of their behavior. This of formulating or altering rules of
1154 54
Chapter 4
inability to overcome previously established response tendencies is manifested by the generation of of fewer hypotheses, and perseveration. While superior, cortex lesions produce produce persistent impairment, unilateral, mostly left, frontal cortex of this type. This defect orbital, or inferior frontal lesions produce no deficit of of the valence of of the feedback signal, of of the nature, verbal or is independent of of the the material. This This "rule breaking" behavior does not only concern spatial, of the identification identification or formulation of of the rule. It extends to a specific failure to comply with task instructions and a tendency to error perseveration. As such, of loss of of inhibition it could be viewed as a special "cognitive" instance of of these patients. These attitudes attitudes of of otherwise encountered in the everyday life of perseveration contrast contrast with the hyperflexibility in adaptation to rule changes of of the patients presenting mediotemporal lesions. In this latter case a defect in ITM and the consequent impairment in extracting consistent perceptual or motor schemes for a sufficiently protracted protracted period of of time could be responsible for this inconsistent behavior. and periarcuate cortex. In contrast - Conditional associative learning and with patients presenting lesions of of the anterior temporal cortex, who have normal learning, there is an impairment in learning conditional tasks, either aiter lesions of of the frontal lobes. Lesions producing such spatial or not, after deficits are more specifically located in the posterior part of of the DSL frontal cortex and the periarcuate periarcuate cortex (areas 6, 8). Even at a plain neuropsychological level, prefrontal cortex demonstrates neither structural nor functional homogeneity. Nevertheless, the diversity of of these functions is more an expression of the nature of its cortical or of nature of subcortical subcortical connections, than the result of of structural variations. of prefrontal cortex according In parallel with this compartmentalization of to sensory modality other authors isolate regions according to processing hierarchies. According According to this criterion, upper and mid-dorsolateral parts of of Pc (Brodmann's areas 9) could be devoted to the most executive, high monitoring level of nonspatial WM ofnonspatial WM.. Conversely, low-monitoring requirements (such as those of of delayed matching-to-sample) would be processed by inferior convexity.
Neurophysiology of of spatio-temporaljUnction function in animals. Rats Rats and monkeys have been the main subjects of of investigation, especially especially for the learning of of space space rather than for the learning of of sequences. Hippocampus: Rat Rat place cells versus monkey view cells. cells. It is not the point of of this article to make a review of of the vast experimental literature
JP. Banquet et al. J.P.
155
concerning concerning place place cells and and more generally the learning of of space space in living beings. We just point point to some inter We will just inter or intra-species intra-species experimental experimental disparities that that do not not find aa satisfying issue in experimental experimental accounts, accounts, and and disparities yet can can be integrated integrated in a single unified mechanism mechanism by our model. Experimental results on spatial spatial integration integration seem, at least least at at first look, to Experimental results species to the the other, or even even in the the same species species differ significantly from one species from one one experimental experimental condition to the the other. One of of the the most striking interspecies difference, difference, between between the rat rat and the monkey in particular interspecies particular comes quasi-absence of of place place cells in the the monkey. Instead, from the quasi-absence electrophysiological 995) demonstrate electrophysiological recordings recordings (Rolls & O'Mara, 11995) demonstrate the existence of of "view cells" that that react react not so much to the position of of the animal animal existence but to the direction direction of of gaze. From such facts some authors authors have in space, but inferred that that the hippocampus, hippocampus, at at least for some species, could be the site of of construction of of an allocentric allocentric frame of of reference reference for space space representation. representation. the construction Our 996, 1997a,b) Our model model and its robotic implementation implementation (Gaussier (Gaussier et aI., al., 11996, 1997a, b) demonstrate mechanism based demonstrate that that a unique unique mechanism based on a realistic simulation simulation of of the CA3 network explains explains both the emergence of of CA CA1l place place cell-like activity when the robot camera around the room (see Figures 4 and 5). camera pans 2700 270 ~ around Conversely, when the "visual field" of 800~ of the robot is restricted restricted to to less than 1180 (similar (similar restriction restriction of of the visual field field takes takes place place in the monkey compared compared to rats), rats), the robot becomes unable unable to develop cells that that generalize a place and rather develops analogs analogs of of directional directional view cells. Another Another important important intraspecies intraspecies critical critical factor for the development development of of place cells concerns place concerns the the position of of the the landmarks landmarks in the the arena. In their usual standard standard position position the the landmarks landmarks are are located at at the the periphery periphery of of the the arena, and more or less uniformly disposed around around the circle. Nevertheless, Nevertheless, a few experiments experiments have placed placed the the same same landmarks landmarks at at the the center of of the arena arena 997). In this new experimental paradigm it (Cressant, (Cressant, Muller Muller & Poucet, 11997). becomes becomes very difficult to record record a place cell type of of activity. In most of of the cases, place cell coding was not stable. In a very few cases where stability cases, could could be found, this happened happened after after a prolonged prolonged stay in the the arena. In the the same vein, the model gives a mathematical mathematical account account of of the absence or at least difficulty of of constitution constitution of of stable spatial spatial attractor attractor basins corresponding corresponding to the of a place place when the referent is a landmark landmark and not an absolute absolute learning of direction (Gaussier (Gaussier & Zrehen Zrehen 11995). direction 995). Spatio-temporal rontal lobe in monkey: Location Spa#o-temporal function function of of the f frontal versus f orm in DSL. In monkey, several form several functionally distinct regions have been delineated. delineated. There There is some agreement agreement with neuropsychological neuropsychological results in human. In particular, particular, a a dissociation dissociation exists between the the representation representation of of
1 56 156
Chapter 4
%"%..~,.\
.~h~.If
.~hr
I i chnit~ talkie
i,_.7,.,......,
/
.,""
L__q.,
_,.,._.....,...,,f.?--'"",,,.
.'i
r ........ i LL.J ,,
~.. ..,"
!.
I'T"}
{.2U
a)
..........a)
9
...:
I 09~
:. . . . .
".
...;:~'\
-Ip-"
0S
0.2~"~'.,,~..~ 0
Figure Figure
:
~ ":"..i:...:. ..... .: . . \
0.g~.t
~'"" "
-~
,,<,/""".Sr..
"~.,> .~/" 0 -0,2 "~'~.~ fl" -0.2 -0.4 "--.,., ,./" .-0.4
b)
4. Place cell-like segregation Place cell-like segregation of of space space in in a a robotic robotic experiment. experiment, a) a) Four Four
views views are are learned learned (circles), (circles), others others are are associated associated with with one one of of those those learned learned (boxes). rom each (boxes). As As we we can can see, see, if if the the robot robot learn learn to to reach reach the the cross cross ffrom each learned learned view view it it can can reach reach the the cross cross from from all all the the views views (generalization). (generalization). The The set set of of
11.2 .2 m .2 m m x• 11.2 m area; area; the the 30 cm from the center. The scale is not respected for the 30 cm from the center. The scale is not respected for the .5 m position erent furniture furniture (in (in f act they position of of the the diff different fact they are are about about 11.5 m from from the the center). erent learned cell activities activities f or several positions in center), b) b) The The diff different learned view view cell for several positions in the the same same environment environment (unit (unit is is meter). meter). arrows path. The arrows represents represents a a possible possible path. The views views are are in in a a learned views are learned views are at at
9~:
9
~ i i ~
t
!
' ,.
�
~ .,~.. " ::~: ' ] " :::::::::: [ ~ i
- ......
...:~,,:,.
.... i~'~
..... .!..,..:!~~
: ~ ~,~..-~,x.:~.
..: ~
:t '-Cl ;:::
II) .... II) ....
�
.... :2 " '~
....
.
.....
J.P. Banquet et al.
�
...............................--
:........ .9.............
,~,
.~"
~L."." .....
t:l
.
.......
-"-.
'
'
i....,' :
'
:
".,
;''
.
",~'~..
9
i
............. !
157
4 a). 0
0
r
9 l,-,q
Two of the panoramic images (a and b) referenced in Figure
5. Figure
i
b
�
a
1158 58
Chapter 4 Chapter
which has been removed from the field of of vision external space, for a stimulus which (spatial (spatial form of of delayed matching to sample or delayed alternation), and representation of of the form of of the visual objects. representation Electrophysiological Elcctrophysiological recordings and and lesions studies have localised localiscd the memory for space in dorsolateral dorsolatcral prefrontal region, principal sulcus in particular. The inferior prefrontal convexity ventrolateral vcntrolatcral to the principal sulcus processes color and form, since lesions in these those areas areas induce deficits in of objects (Goldman-Rakic, 11994). tasks requiring memory for the identity of 994). Specific stimulus identity neurons there, are particularly particularly responsive to to pattern pattern delayed response tasks. The reverse is true for neurons in the dorsolateral dorsolateral prefrontal prefrontal cortex, and sulcus principalis coding for location. Neurons hold "on line" these distinct types of of information when the stimulus is no longer present, and are arc clearly distinct from neurons coding for the direction direction of of limb or eye movements that that can can also be found in the same areas. Degeneration studies have confirmexl confirmed the anatomical links between sulcus principalis and posterior parietal cortex (area 7), and between inferior convexity and posterior parivtal inferotemporal infcrotcmporal cortex (area TE) which respectively process spatial and pattern visual information. As yet, it is not clear if spatial information pattern recording takes place in either alloccntrie allocentric or egocentric coordinates or in both. possibility remains of at least least in primates hippocampus could The possibility of both, since at of processing processing for space in an allocentric alloccntric (not necessarily be the locus of cartesian) of reference. cartesian) frame of Response Response pattern pattern and and motor motor control in in inferior ferior and and arcuate arcuate convexity. Inferior convexity has also been attributed diverse other functions. According of to the unlearning hypothesis, these may include inhibition or unlearning of response patterns which do not maximize the probability of of reinforcement, and therefore induce a change of of a predominant response mode. Finally, motor control for interrelating the inferior convexity could be involved in motor stimulus-reinforcement associations associations with behavior, in order to promote motivated behaviors and to prevent indiscriminate motor choices. choices. Autonomic rontal cortex. Another, Autonomic and and motivational motivational control control in m orbital f frontal uncontrovcrsial functional focus is located in the medial modial orbital cortex and uncontroversial endowed with autonomic and emotional control. It could be involved in motivational evaluation evaluation of of reward reward and continuous modulation of of ongoing motivational behavior by competing drives. human, there are arc arguments against a Nevertheless, in monkeys as in h uman, there functional compartmentalization that would be uniquely uniqudy based based on perceptual perceptual modality. Reversible lesions of of the dorsolateral dorsolatcral Pc during a cross-modal of delayed matching-to-sample task produce deficits in somasthesic, somasthcsic, version of
JP. J.P. Banquet Banquet et al.
159
spatial or nonspatial WM WM demonstrating the amodal or supramodal processing taking place in these areas. This could result from a partial convergence of of inputs from posterior cortical areas to the Pc. As in man, at this neurophysiological level, there is evidence both for functional uniformity of of the prefrontal cortex, in particular with regard to WM function and neurotransmitter modulation such as dopamine, and also WM for diversity, as shown by lesion and degeneration studies in animal. Functional Model
of Pc-Hs functional relations which We present first a global model of of Hs in L LTM WM specifies the contribution of TM consolidation and transient WM operation. Then we compare and and contrast the respective roles of of both structures in spatio-temporal processing. In particular we give an integrated view of of the different experimental results concerning place cells, view cells and the experimental conditions leading to the emergence (or not) of of such abstract representation of of space as place cells, and also the plausibility of of more elaborated cognitive maps. We suggest one plausible mechanism (retropropagation of of goals) and its cortical implementation which controls of a sequence of of elementary step-actions in selection and hierarchical ordering of order to lead to subgoals and final goal. In this endeavor we not only take into account cognitive processes but also drives and motivations. The of an algorithmic mathematical model has been simulated. It is also the object of of robot navigation either in open space, or implementation for the control of maze environment. Co rti co-hippocampal relations relations Cor/ico-hippocampal
of the key pathways of of the Hs. There There are many Figure 6 shows some of common points between functions of of Hs and Pc, and one of of the major of research on the functional relations between these two two structures problems of of each of of the two systems to this is to find what are the specific contributions of apparent functional overlap. In particular both systems seem to participate in of WM WM, and the capacity for a WM WM function can be different aspects of expected to contribute a major step to encephalization. On anatomical of these functions in Hs and mostly in Pc (the grounds, the localization of of which is the most pre-eminent anatomical correlate of of primate development of evolution), supports the functional importance of of a "protracted present" for WM provides the animal with the humanization and personification. WM ,
160
patte rn recognition pattern recognition
and and time time spectrum spectrumexpression expression
Associative Associative and and prefrontal pnfrontai Cortical areas Cortical areas
prefrontal prcfrontalcortex conex
Associative Associativecortical cortical (nucleus accumbe
pr efusion refusion confi gurational associations configurntonal associations
(nucleus accumbe
panem completion
bl
A
plastic plasticfusion fusion of of multimodal multimodal information EC2 ECZ
OR operation ?
EC3 EC3
-
I
I
.
I
I
I :, :1 :1 , * - _ ,�- Y X - - - - - - - - - - - - - - - j � ---,
, _
,u
::
_
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Septum ~eptum
4-: _
_ _ _
(ACh)
(ACh)
Novelty Novelty detection, detection.learning learning modulation modulation
I
1I
I
� _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ J
Figure ocampus. Entorhinal Figure 6. 6. Schematic Schematic representation representation of the hipp hippocampus. Entorhinal Cortex (EC) (EC) receives receives information information from the associative associative cortical cortical areas areas.. Highly Highly filtered filtered information from the EC layer 22 is is transmitted transmitted to the Dentate Dentate Gyrus Gyrus (DG) (DG) and to the the CA3 CA3 pyramidal cells. cells. In the model, the DG recognizes recognizes activity activity pattern from EC2 EC2 and develops develops a temporal temporal activity activity spectrum spectrum (factorization (factorization of time time and pattern). pattern). CA3 CA3 recurrent links links allow pattern completion completion and association between an incoming pattern and a previous pattern recognized and maintained in DG granular cells. incoming previous cells. CA3 CA3 allows allows a "plastic" "plastic" fusion fbsion of multi-modality information. information. Integrated recognition of the CA3 CA3 representation representation is performed on CAl CA1 pyramidal pyramidal cells. cells. It can can be either place recognition (i.e. (i.e. rat rat experiments) experiments) or view recognition recognition (i.e. (i.e. primates) and more generally generally sequence sequence recognition. recognition. Information from CAl CAI is then treated by the Subiculum Subiculum (Sub). (Sub). This This information information could become invariant to animal animal orientation due to this structure structure (i.e. (i.e. head direction direction cells). cells).
Chapter 4
I
u
I
J.P. Banquet et al. JP.
1161 61
possibility to escape from a purely reflexive behavior, by taking into account its more or less immediate past and simulating the consequences of planned actions into the future, so the animal's behavior can be grounded not only representations.. upon external stimuli, but also on internal representations A direct consequence of of this WM capacity iiss the possibility to learn and of events or actions hierarchically to plan long temporo-spatial sequences of subordinated to a global but not physically accessible goal through intermediate subgoals. This capacity seems to constitute one of the distinctive of Pc. This temporo-spatial property is crucial for the orientation properties of of the animal in its environment. Nevertheless, even in this and situation of domain of of sequence learning and spatio-temporal integration, several levels of of delmeatexl. At the level of the hippocampus in particular, particular, in analysis must be delineated. of theta them and ggamma rodents, there is evidence of amma rhythms possibly generated by inhibitory feedback circuits. These rhythms can be related to a basic of postsynaptic postsynaptic potentials under the influence of of this inhibitory oscillation of gamma them activity, modulation. The g amma rhythms in particular, overriding the theta of discrete event sequences during have been viewed as the encoding support of exploratory behavior. These sequences would be partially repeated, proceeding further as a travelling window of of activity, at each theta cycle. The hippocampal NMDA receptors are endowed with slower time constants (circa ocampal NMDA hipp 200 msecs) than the short short time constants of of the cortical NMDA NMDA receptors of correlations (circa 20 msecs). Therefore, they will allow for the learning of between successive events occurring less than 200 msecs apart, that that is about about them cycle, and approximately 7 event-coding gamma a full single theta gamma cycles. quite plausible for rodents that that would thus thus be able to to This possibility remains quite separate events about learning this about 20-30 msecs apart. In confirmation that that learning type of at Hs level, aa type of sequences of of elementary events could take place at primacy gradient of been found at at the the synapses between Schaffer LTP has been primacy of LTP collaterals and CA I cells during the presentation presentation of of a sequence of of stimuli (Granger et al., 11994). virtue of gradient the the earlier the the event 994). By virtue of this primacy gradient (Granger in the sequence, the larger the corresponding synaptic facilitation. Therefore at at the presentation presentation of of aa stimulus stimulus of of the sequence the the playback of of the sequence in the correct correct order order is favored. Despite the interest not operate at this Despite the interest of of this this mechanism, our our model does not operate at this low elementary level of of resolution. resolution. It rather tries to to emulate events reaching awareness attention and consciousness. awareness in the human human field of of attention consciousness. At this level individual mental states states have an approximate duration an approximate duration of of the order order of of the second. This does not not certainly preclude the possibility, even in humans, humans, of of automatic automatic subconscious subconscious temporal temporal discriminations of of sensory events in
1162 62
Chapter 4
particular, of of the order of of a few tenths of of millisecs. millise~s. Increasing in duration, our activation states gain also in flexibility and control, and become closer to the of the chapter, brain states that actually control behavior. In the course of arguments from neuropsychology and brain imaging have been provided in favor of of an involvement of of Hs at every phase of of information processing, from the very short-term to the transient long-term. There are also some arguments against the implication of TM store, although opinions of Hs as a "permanent" L LTM diverge on this issue. Thus, we have been led to contrast hippocampal function devoted more specifically to information selection, consolidation, and basic temporo-spatial temporo-spatial correlation from cortical function supporting a direct dialectic confrontation TM during recall and recognition. confrontation between STM and L LTM Cortex-hippocampus complementarity. complementarity. The first claim concerns the complementarity between cortical and hippocampal memory systems. Cortex is mostly endowed with short-term and long-term permanent memory capacity, even though there exist range variations from primary, to secondary, and association areas (Lu, 992). It has (Lii, Williamson & Kaufinan, 11992). more specifically, but probably probably not solely, a capacity for slow learning. Hippocampus Hippocampus is richly endowed with a full spectrum of of temporal ranges, to the the transient long-term, possibly to the exclusion of of the from the short-term to of a large variety of of permanent long-term. More specifically, this combination of memory registers, with just just as much diversity of of close loops of of different sizes, at least least five synaptic relays, seems unique in the brain. It is including one to at of one-exposure learning proper proper to probably responsible for the property of of logical arguments yet to be experimentally hippocampus. On the basis of should take place place not only confirmed, we suggest that fast, transient learning should itself but also at the interface of of the convergent inside the hippocampus itself cortico-hippocampal r gyms), and corti co-hippocampal pathway (either entorhinal cortex or dentate gyrus), of the divergent hippocampo-cortical pathway similarly at the interface of (possibly in the cortex superficial layers). Conversely, experimentally confirmexi slow-permanent learning in the neocortex could could be restricted to the confirmed of polysynaptic cortico-cortical connections. level of Two types types of of transient memories. memories. The second claim concerns the of the two aspects of of hippocampal transient memory, complementarity of Transient Memory (ITM) supporting a WM WM function and Intermediate Transient transient L LTM of WM WM consolidation. These TM supporting the process of of complementary functions cooperate to fulfill the contradictory constraints of relevant information as possible in a large but limited storing as much relevant capacity system. The WM WM function thought to operate both at cortical and capacity hippocampal level is devoted to this segregation between information worthy
JP. J.P. Banquet et al. aL
1163 63
survival, or its human to be "permanently" stored because of its relevance for survival, social relevance for the personal history of the subject, and information which can be forgotten without major damage. For this purpose several hardwired devices implement different functions in various subsystems of the hippocampus: of stimulus intensity, duration, or - Severe filtering on the basis of of entorhinal cortex; repetition, mostly at the level of of noncorrelated patterns of information, reformation, along with - Orthogonalization of lumping correlated information and/or suppression of redundant redundant patterns, possibly in the dentate gyrus; - Correlation of of these orthogonal patterns on the basis of of spatio-temporal of either co-occurrence or sequential ordering. The CA3 subsystem is criteria of specifically equipped for these processes of autocorrelation. There is some of temporal order processing of the events at the level of of CA CA11 evidence of al.,, 11994). (Granger et aI. 994). of a specific processing capacity All these functions argue in favor of devoted to the Hs, based on ITM. This may be possibly more devoted to temporal aspects of of information (like temporal order and timing), than spatial learning, which is the focus of current investigation. More exactly, spatial learning itself would include a strong temporal dimension (in the sequential of snapshots), where temporal order is not relevant. Correlational recording of of Hs, and could account for many aspects processing could be the hallmark of of of its cognitive processing, including temporal processing. But it is certainly not the only form of the hippocampus. On the the basis not of processing performed in the of of CA3 autoassociative architecture, artificial systems can be designed such that simultaneous pattern correlation processing becomes a special case of that of successive pattern temporal order processing. The main difference is that that events occur simultaneously instead of of taking place in sequence. The other class of based on LTP, The of transient memory is transient LTM LTM based LTP, and a support for permanent LTM LTM consolidation. The most prominent sites of of LTP LTP in the hippocampus are at CA I1 level, but also in the entorhinal cortex, dentate dentate gyrus and and CA3. CA3 . LTP LTP is not not specific to the hippocampus. hippocampus. Different areas, in particular, particular, are are also prone to LTP. What What seems more cortical areas, specific, if if not exclusive, to to the the Hs is the the conjunction conjunction of of systems susceptible to shorter versions of to LTP LTP along with with shorter of synaptic facilitation, in the framework of of an an architecture architecture adapted adapted to to facilitate aa smooth transition transition between short-term and and long-term facilitation, without without the need to resort to externally dependent repetitions of of activation. These These two two functions of of ITM ITM and and transient LTM LTM are therefore complementary. Indeed, information or or events have to proceed -
-
-
1164 64
Chapter 4
through the cortical cortical and/or hippocampal WM process in order to get a "certification" as being worth passing to the history of of the individual. individual. This certificate certificate could correspond to the transition from ITM supported by short synaptie facilitation LTM LTP. duration synaptic facilitation (STP), to transient L TM supported by L TP. A unique unique mechanism f or LT M and WM. for LTM W~. The third claim concerns the unique mechanism supporting WM operation and L TM consolidation. This LTM unique process takes advantage of of the reciprocal, reciprocal, topographically organized connections connections between cortex and hippocampus, combined with the endogenous property of of pyramidal neurons, mostly in the CA3 region, to discharge by bursts (Buzsaki, 11989), 989), and synchronise within populations connected by previously facilitated facilitated synapses. Endogenously generated bursts of of activity have been documented in the CA3 region, and are manifest at CA level as CA11 level sharp waves. In contrast with theta rhythm which have been consistently recorded mostly in rodents, sharp waves have been also found in primates and humans. They could fonn form the basis for reactivation reactivation of recently facilitated facilitated neuronal populations either at hipp ocam pal level or at cortical level hippocampal level level or both. They would correspond to what we call the print mode of the Hs, by contrast to the read mode. In episodic learning, this print mode would subserve the function, attributed to the repetition repetition or practice process during the formation of of procedural memories. Two general mechanisms will cooperate in the reactivation reactivation of of cortical cortical patterns, starting either from these endogenous hipp ocampal bursting and hippocampal ocampal reactions synchronization capacities, or from hipp hippocampal reactions to cortical activation of of cue patterns. patterns. First, there is resonant or reverberating activity between reciprocally connected networks. It has been used in neural network modelling as a mechanism for synaptic weight modification leading to class 976a,b). Second, neural population synchrony, a learning (Grossberg, 11976a,b). consequence of of the reverberant activity, has been presented as a plausible temporal coding used locally locally by the brain for the coalescence of features features into a unified percept (Singer, 1983; 989). The main contention of 1983; Gray et aI., al., 11989). of the model is that ocampus as a site for both that the central central location of of the hipp hippocampus convergence from, and divergence towards the different cortical areas allows the Hs to act as an infonnation 983; Banquet information selective pacemaker (Banquet, 11983; & Contreras-Vidal, 11994). 994). As such, it can selectively synchronize distant cortical cortical areas that have been previously coactivated. coaetivated. This mechanism provides provides for both the refreshing of of memories during WM operation, and the TM reactivation of of cortical cortical patterns during the more lengthy process of of L LTM consolidation. In this last case it will create the conditions for the slow facilitation facilitation of of distant polysynaptic cortico-cortical cortieo-cortieal connections, which
JP. Banquet et al. J.P.
1165 65
eventually will make make recall or recognition independent of of hippocampal function. Transient storage by the hippocampus. The last claim is largely but not universally accepted. The information processed through the hippocampus is not bound to be permanently stored there, at least in humans. Beside anatomical arguments, like the relatively limited capacity of of the system, the strongest support for this contention comes from the limited range of of the retrograde amnesia in hippocampal lesions. Most of of the memories from the distant past are spared. This raises the issue of of the necessary transfer of of the information to cortical and/or cerebellar areas, which are supposed to be among the most pre-eminent sites of of permanent storage.
Model of of dentate-CA3 dentate-CA 3 function The experimental sources for the inspiration of of our mathematical model are of of two different types. First, the learning of of space by animals as demonstrated by rat place cells or monkey view cells, and eventually the of cognitive maps in an allocentric frame of of reference, more formation of of space is considered as plausibly in primates. In our model this structuring of a degraded form of of sequence learning. Second, in trace conditioning with a response there is evidence of of adaptive timing of of the behavioral motor response response, as we shall see in the dentate gyrus model. these two two subsystems of of Hs, dentate gyrus (DG) and CA3, In our model these are endowed with specific functions which go beyond the classical attributes of of pattern pattern orthogonalization (dentate gyrus) gyms) and autocorrelation autocorrelation (CA3) which are the usual characteristics characteristics of of these subsystems in neural network modeling. These two two classical functions are in fact implemented in such a way as to add a temporal dimension to both of of them, on the basis of of experimental evidence. The addition of of this temporal characteristic to these networks is globally of the hippocampal hippocampal registers which are justified by the temporal properties of capable of of fast transient short-term or long-term learning thanks to the synaptic properties of of short-term and long-term potentiation particularly To this propensity for rapid transient transient learning must be added developed in Hs. To the loop-like organization of of the different circuits which allows for a few repetitions or iterations of patterns of activation even though they have been externally triggered only once (Buzsaki, 11989). 989). Thus, these loops endow Hs rather to with the potential to maintain for some limited duration, or rather punctually reactivate, a recently activated pattern that has been transiently of the recruited populations. learned in the synaptic connections of
166
EC2
Modality 11
-<1--+--_____
Modality 2 2 Modality
�...-+ _ _ .... .r
Modality 3
-1--4-----...
!
I
No ACh No ACh
DO 0 DG
1.
11
t -4
0
intermediate connections
Proximal connections CA3 competitive structure
Ach modulation
Septum
"L_J connections. Figure 7. Detailed model of the CA3 connections.
9 -§ Chapter 4
ACh
Distal connections
� ..,
�
J.P. Banquet et al. JP. Banquet
1167 67
of these these temporal properties, the model implements As a consequence of two different temporal functions: of timing, i.e. a flexible evaluation of of time intervals or - First, a function of of granule cells of of the dentate gyrus; gyms; durations, at the level of of the CA3 - Second, the corresponding temporal or order function of of linking together successive similar or different events of of a network consists of process results from a combination of of sequence (see Figure 7). This chaining process autocorrelation and crosscorrelation autocorrelation crosscorrelation as we will see in the paragraph on the CA3 model. Factorization of o f time and and pattern pattern in the dentate gyrus. The The model Factorization presented here is based on the same principles of of population learning and of time by a limited assembly of of neurons whose dynamic range of of coding of activation varies along a biologically plausible continuum. This timing of results function has been attributed to the dentate gyrus on the basis of of trace conditioning of of the nictitating membrane derived from the paradigm of response in the rabbit (Berger & Thompson, 11978; 978; Solomon, 11980). 980). The of first monitoring and then anticipation of of the behavioral response, evidence of of pyramidal cells shows up first at the level of of CA3 by firing patterns of neurons. This lets us suppose that that processing the duration duration of of time intervals to the CA3 system. Dentate gyrus is the immediately takes place upstream to preceding stage (see Figure 6). Granule cell population there is sufficiently present variations in size and time constants which could support important to present There are several lines of of evidence for neural different dynamics of of activation. There dynamics in the brain varying according to a continuum, in particular in spinal motoneurons and there is also evidence of of spectral decomposition of of space in the visual cortex. This hypothesis forms the basis for a population of time by cells endowed with these different dynamics. Similar neural coding of mechanisms underlying the hippocampal adaptive timing function during conditioning were proposed by Grossberg and Merrill ((1992), 1 992), among others. have already proposed proposed a model including a timing system which gates or We have modulates the flow of of information in a categorization system (Banquet (Banquet & modulates The main difference between this last Contreras-Vidal, 11993a,b; 993a,b; 11994). 994). The that the onset of of the input stimulus is not model and the present one, is, first, that held on by an hypothetical external device but the STM is directly implemented through the positive feedback from the GC to MC and their return to the GC (see Figure 8 for the DG architecture and Figure 9 for a of its temporal dynamics). A second difference is that the same simulation of of cells process, or register, .both .both pattern and time, thus realizing populations of of the two two parameters. This basic computational competence a "factorization" of -
Chapter 4
1 68 168
may may also also be be shared by by cerebellum cerebellum which is more more suited for the the motor motor response response (Ivry & Keele, 1989; 1 989; Bullock, Fiala Fiala & Grossberg, 1994). 1 994). There There is at present present no no decisive decisive argument against the fact fact that the cerebellum cerebellum could could at support the upstream upstream system system forwarding timing timing information to to CA3. support
time axis axis time
Pattern
GC "
...-,--r . -,,. ,--
,r.I .I,,
i l
I
;,
i
-,--~ ~
i
~ ,
..
.t,
!
-,--
---r-
!
:l ( ( C
EC2 E C2
,L
/
I _
BC ---- ~ I "~
~.
( 17
-
!
',/,,
c,c ~ BC,
I
L-O t~
" .....,.
BC
~
.-
~
.
.
9
.
.
excitatory excitatory link link
.: .
Reset inhibitory link inhibitory link
.
MC MC
. . - . . .
"
. . . . . .
..
"
. ,
. . . . . . . . .. .
I to CA3 CA3 to
Figure 8. Detailed representation of of the different neurons and connections in ECZ is DG used in our simulations. simulations. Instantaneous information information coming from EC2 learned by the Granule Cells (GC). We suppose activity spectrum of of GC depends on their size. GC are linked together in a functional cluster through a particular Mossy Cell (MC). This recurrent recurrent connection allows allows GC to develop time spectrum activity while EC2 stimulation is over. Basket Cells (BC) allow choice of of a winner cluster of of GC: pattern pattern recognition of of EC2 information. Another MC and BC subset allow the reset of of the active GC cluster at the onset of of new significant input from from EC2 to avoid pattern pattern recognition problems.
Neurons activity
activity of a little GC
activity of another GC in the same cluster
/
acti vity of activ~ty of the the
associated associared Me MC
J.P. Banquet et al.
o 0
time time
Figure cells (GC) (GC) randomly Figure 9. 9. Simulation Simulation of of aa cluster cluster of of "spectral" "spectral" cells randomly selected selected in in the the Dentate Dentate Gyrus Gyrus (DG) (DG).. A A pattern pattern is is presented presented at at time time o. 0. As As aa result, result, the the GC GC develop develop aa time time spectrum spectrum activity activity depending depending on on their their size. size. Activity Activity of of the the GC L
169
is is maintaining maintaining due due to to recurrent recurrent connection connection through through aa MC MC cell cell.. The The decrease decrease of of GC GC activity activity is is due due to to an an habituation habituation term term at at the the synaptic synaptic level level (between (between MC MC and and GC) GC)..
$ %
1 70 170
Chapter Chapter 4
Nevertheless, Nevertheless, the the plausibility of of multiple multiple local local timing systems dedicated dedicated either either to perceptual, perceptual, motor, or or even cognitive processes processes argues argues in favor favor of of an an to hippocampally hippocampally based based local system for for the the timing of of perceptual perceptual events. Figure how aa particular particular punctual punctual event can can be extended in time time Figure 9 shows how of cells endowed with randomly varying time constants, constants, until the the by aa battery of subsequent significant event takes place. The The two events could be the the CS and and the US of of aa trace trace conditioning sequence. But more generally, our our system deals with any arbitrary sequence of of either distinct or or repetitive events or both, occurring time intervals. Thus, can account occurring at at variable time Thus, it can account for the learning of of any independently of any pair, and and therefore sequence, of of events independently of the presence presence or or not of learning and conditioning becomes becomes aa not of aa reinforcer. reinforcer. Reinforcement learning particular case of more general associative learning. particular case of this more It cannot the presently available cannot be excluded, on on the basis basis of of the of maintaining aa pattern of experimental evidence, that that the the function function of of STM or ITM ITM could not be performed by CA3 itself. There is information in STM indeed evidence of such aa maintenance of under aa "dormant" form of such of information under of short-term or long-term synaptic facilitation. In order to be used by the of of information processing (match-mismatch, filtering, system for purposes of correlation . . . ) this dormant form needs to be reactivated or reenacted as aa correlation...) pattern of of activation. In aa sequence of of dissimilar events, this this coherent pattern reenactment is not not possible from one event to the the next. Therefore the pattern of activation must be maintained in an active form in order to be compared or of associated to the next one. The circulation of of information in a loop could in theory also perform this function of maintaining of of information in an a pattern of active state. Also, event-locked them theta activity could be another mean used by Hs to protract in time an active pattern. Our system of of population coding by a battery of of cells seems parsimonious, flexible and and therefore plausible. plausible, but but it does not not exclude these other possibilities. In our model, relations of of Hs with other structures in particular frontal cortex for temporal order processing, and cerebellum for timing are viewed as special cases, in the temporal domain, of of the more global function of of Hs as a rehearsal system for the rapid acquisition of any kind of of information. This interpretation interpretation seems to be confirmed by the fact that that learning this type of of information, time intervals in particular, is not really suppressed by hippocampal hippocampal lesion. It just just requires more trials to be learned. A ,4 temporal order network in CA3-CAl. CA3-CA/. Different models of temporal order processing have been designed. The most recent models aim in particular to emulate temporal processing attributed to prefrontal cortex. Guigon, Dorizzi, Burnod Bumod and Schultz ((1995) 1995) have designed a model directly
Banquet et al. J.P. Banquet
1171 71
inspired from the electrophysiological studies performed in monkeys. The implementation of processing units which can commute between two stable synaptic activations activations allows the states of activity (bistable units) in response to synaptic sequences.. After Atter learning, the sustained activation of a learning of temporal sequences of a past event, selective given neuron represents the selective memorisation of of a future event, or the prediction of of a reinforcement. Thus, the anticipation of model reproduces the functions of delay neurons encountered in frontal cortex. Bapi and Levine ((1990, 1 990, 11994) 994) and Levine and Park ((1992) 1 992) designed network models of frontal cortex function. In these models, sequence learning is secured by storing in LTM (synaptic weights) the transitions between the events of of a sequence. The different types of of learned sequence are encoded in a compressed form, and then categorized. It presents some similarities with our of WM WM for temporal order and probability coding (Banquet & own model of Contreras-Vidal, 1992a,b; I 992a,b; 11993a,b; 993a,b; 11994). 994). Both are in particular inspired from the same design principles of of Grossberg ((1978). 1 978). Dehaene, Changeux and Nadal ((1987) 1 987) implemented a network that can learn temporal sequences based on biological properties of of allosteric receptors. our model, the dentate gyrus gyms performs simultaneously the double In our of time and and pattern. This output output of of the dentate gyrus gyms can then be used coding of by the CA3 stage. This stage represents a multimodal event or state as a of activation. It performs then for each event a double correlation pattern of correlation: -First, aa zero-delay "auto-correlation" of of the the event with itself itself thanks to to the temporal conjunction of input to CA3 and and the fast the of the direct pathway input spectral component of of of the indirect input transiting by the trisynaptic loop of the dentate gyms (Figure 6). gyrus -Second, aa delayed auto- (stimulus repetition) or or "cross-correlation" (successive stimuli different) between the present direct input to CA3 and and the previous indirect input which has been maintained maintained in an active state thanks thanks to of the dentate gyms gyrus granule cell population the slow spectral components of (Figure 9). CA3 learns this cross-correlation between successive events. This last recording of states between events. These last process results in the recording of transition states transition states states allow at any points in the sequence prediction, priming or at any playback of of the the subsequent subsequent event according to the functional mode of of the playback system (i.e. learning, recognition or reenactment of of aa sequence). The transition from aa functional mode to to the next is endogenously determined by transition the type the type of of neuromodulation neuromodulation (ACh (ACh in particular; particular; Hasselmo & Schnell, 1994) 1 994) and the biological rhythms. and the
1172 72
Chapter Chapter 4
For the purpose purpose of of linking a specific behavior to a particular view of of the PerAe (Perception-Action) architecture (Gaussier & environment, as in a PerAc Zrehen, 11995), of these thesetransitional 995), linking of transitional states to elementary motor actions has been found to better support the planning and implementation of of specific sequences, rather than than the mere mere relation of of perceptual states to actions (Figure 10; Gaussier et al., 11997a,b). 997a,b). These transition states do not differ in their coding from simple states since they are also encoded as correlation patterns in CA3 network. Their meaning is different since they correspond to the correlation between patterns which do not occur simultaneously in time. The timing system of of the dentate gyrus gyms is set in derivation on the main flow of of information that that transits through CA3, and thus modulates or even gates and infonnation controls the activity of of CA3 CA3.. This property makes the system very flexible with respect to the nature nature and variable timing of of successive significant events. The learning not only of of autocorrelations between simultaneous patterns of of activation, but also of of "cross-correlations" between successive patterns can sub serve several functions. First, during the learning process of subserve of a repetitive sequence, the predictive activity of of the system results in priming the next event in the sequence when the present pattern of of activation acts as a trigger for the transition pattern. This priming is critical for the implementation of the match-mismatch process that operates to recognize the successive events of a sequence at CA3 level. There, the previous pattern maintained in DG and of of reaching CA3 by the mossy fibers is matched with the present pattern of activation forwarded by the direct connections from EC to CA3 pyramides activation distal dendrites. Therefore, the learning process is modulated according to the of predictedness predietedness or novelty of of the events. Hasselmo and Schnell. degree of ACh septal modulation of of the Hs activity could ((1994) 1 994) have shown how ACh of the learning process, as a function of of provide for an automatic control of of the events. One can suppose that that a match inducing novelty or familiarity of of a new transition and sequence. The familiarity preempts the learning of resulting "resonance" between betwee� learned and input patterns provides for both of a state and the reinforcement of of previous learning. the transient stability of Conversely, mismatch between the expected or primed pattern and the input pattern fosters, by means of of a high Ach modulation, the prevalence of of input over the stored pattern and therefore the recording of of a new transition pattern. Second, in sequences of of events that occur only once or a few times, the Hs provides the the unique facility for fast fast transient learning, thanks to its capacity of recent patterns of of coherent activation. This reenactment for reenactment of correspond to a synchronous discharge of of CA3 neurons which activates CA I of CAl CA1 cells manifests as sharp waves (Buzsaki, pyramidal cells. The activity of
1173 73
J.P. Banquet Banquet et al. JP.
,%. ~o'~ r
q
#7/.~ ~,
..~o"/
/,"
/
l
,~~5'~'0-~
l
)
__~
~
~
~"
T//~
.~S~"/
'~W-'----",jta~ ~l),~'- ~ C
'.,,
', AE
/
Limbic System
Max operator
. . . . . . . . . . /. . . . . "'.
~,,*V 0A I t - . . .
//I\\
Motivation
Activation diffusion
/
: , ..... ~..... / efferent copy of CA! pyramidal cells
...................
-"
/__. /'~
.... .K'-->" / ....
". . . .
direction of movement
/
/.__~ ~
Motor Area
--~"
Associative motor area
Figure 10. Goal backpropagation: the model is used to learn transitions between stable states such as to go from A to B (AB) and next from B to C (BC)... CA1 neurons. CA3 neurons are used to merge the recognition of (BC) . . . on CAl the current state coming directly from the entorhinal cortex EC2 and delayed information about the previous state that that we suppose comes from the Dentate Gyrus and more precisely from the Granular Cells (the GC developing a time spectrum expression of of the information coming from EC2). The CA3 representation of of the the state transition is then learned on CAl CA1 neurons and copied on prefrontal cortical neurons which learn the "anti-causal" (backpropagated) links between pairs of of transition states. These prefrontal if they are neurons also learn associative links with motivation nodes if activated simultaneously with them. Assuming the activation of of a motivation coming from the limbic system (the will to eat for instance), the model of the goal to subsequent subgoals on the prefrontal explains the propagation of cortex. When a subgoal transition can be achieved because the robot state of the transition, the associated learned action is corresponds to the first part of the trigger and allows the robot to reach a new state which is the starting point for triggering the recognition of of a new transition. When several BD) the level of of transitions are possible from a current state (like BC or BO) activation of of the prefrontal neurons associated with the different possible transitions allows the choice of of taking the shorter or the most interesting pathway from the current state to the final goal.
1174 74
Chapter 4
11989). 989). These sharp waves present some of of the characteristics, in particular in amplitude amplitude and and frequency, of of the stimuli capable of of inducing LTP. They could therefore participate in the reactivation of of cortical patterns necessary for the TM storage at L LTM at cortical level. The linkage between successive events constituting transitional states would result in a playback not of of single events but of of sub-sequences of of events.
From hippocampus to pref rontal cortex prefrontal of states and transitions forming a sequence of of events is, in This memory of the model, further ~rther integrated at at CA CA11 level, with a possible partial partial restitution there of of the "cortical" topology thanks to the direct pathways from entorhinal cortex third layer to CAL CA1. From CAl CA1 there are direct as well as indirect connections to prefrontal cortex. At this prefrontal site the long-term registration of of a sequence can be considerd as a result of of the successive of nodes which can be assimilated to cortical columns. By the very activation of orderly nature of of a sequence sequence encoded by a spatial pattern of of activation, one could assume that the best neurophysiological support for such a storage would be a unidirectional facilitation of of the synaptic weights of of a specific path path in a network architecture, as it has been implemented in Bapi and Levine ((1994). 1994). It is plausible that such an oriented unidirectional facilitation takes place in primary or secondary cortical areas. Nevertheless, the prefrontal cortex is the most plausible site for the linkage between sensory and motor I). This sequences at least at a high level of of controlled processes (Figure I11). does not preclude the possibility of of sensory-motor links at at subcortical or even lower levels as schematized by Figure I11. I . At cortical level, the execution of of a sensory-motor sequence is necessarily linked, at at least implicitly, to the completion of of a goal selected by motivation. A goal in the model corresponds in fact to a secondary goal, ii.e. .e. a situation which allows the satisfaction of of a basic drive or of of a sublimation of of this basic drive. At the executive controlled level of of the prefrontal cortex there there is clear, even if only subjective, evidence that that the goal is usually present and therefore activated at the very onset of of the sequence. Thus it can can influence the choices of of subgoals and the hierarchical unfolding of of specific endeavors to reach them. The most parsimonious implementation of of this psychophysical reality requires the instantiation of of a bidirectional facilitation of of the different pathways leading from the starting point to the goal of of a sensory-motor sequence. In this way, the activation of of a goal induces a retropropagation of of activity, similar to a priming by top-down activation from the categorial
-
-
-
-
Prerronta\ cortex L--.,.--r--�--.---J ' - "
,2
Posterior association cortices
' G 175 175
J.P. J P. Banquet et al. ,
Drives
" - "V"'"
Premotor cortex
e~
// cL
Primary Primary sensory sensory
Primary motor
cortex cortex
cortex
,// Perception Perception
Action Action
,
,- - - - - - - - - - - - - - - - - - - - - ,
, ,
L-____-.;' ,
External External environment environment
I I I
I I
,I
, - - - - - - - - - - - - - - - - - - - - - �
Figure 111. 1 . Flowchart Flowchart of of multilevel multilevel information transfers: 1) between sensory and motor areas (horizontal (horizontal connections) connections) 2) between between sensory (or motor) and associative associative areas of of increasing complexity complexity (vertical connections) 3) between planification capabilities (prefrontal cortex). internal drives and planification
nodes 976a,b). nodes in an ART ART architecture (Grossberg, 11976a, b). Yet, Yet, here the priming process concerns an entire entire sequence sequence of of events, and accordingly, accordingly, is implemented implemented according according to a gradient. This subliminal backward backward priming of of a sequence sequence in conjunction with with a bottom-up activation from subcortical structures structures such as the hippocampus helps the selection of of the best sequence of of actions to reach a specific specific goal. This process of of goal retropropagation is not solely solely efficient efficient for selecting the optimal way way for goal attainment. It also operates in deciding deciding the order of of goal goal satisfaction, satisfaction, when when several goals are
1176 76
Chapter 4
simultaneously active or in competition. As such, it implements aspects of of
performed Similar disambiguation between several hierarchy setting perf ormed by Pc. Similar possible possible sequences takes place at Hs level during the playback of of the
if the prediction of of the future event is based on several previous sequence if of just the immediate preceding one. events instead of When such a neural network is used for the control control of of a mobile robot,
different of behavior in order to select and reach a goal are exhibited by diff erent types of the robot. These These choices are are dependent on three parameters of of the system: the
strength of of the diff different of the connections relative strength erent drives, the weights of implicated in the diff different erent paths, and the required match level between perceived and memorized steps towards the goal. Any node of of the cognitive map learns leams transitions between pairs of of learned places. The level of of activation activation of of these nodes results from the addition of of bottom-up (match-related) (match-related) and top toppath length-related) activations. These combined activations down (drive and path of of the nodes can lead to aa variety of of behaviors of of the system which have e. If counterparts in real lif life. If the top-down influences are too weak, the system or the attainment of is unable to follow aa specific path ffor of a specific goal. It is susceptible susceptible to distraction distraction by any new input previously associated associated with a if the the saliency of of the the top-down input is too different behavior. Conversely, if strong, the the recognition of of aa situation could be biased biased in the the direction of of a situation corresponding to the satisf action of its goal. The initiation of such the satisfaction of The of erroneous erroneous recognitions recognitions can can be be self-reinforcing. self-reinforcing. All All these these situations situations correspond correspond to pathologies of of frontal lobe. Fronto-Hippocampal Fronto-Hippocampal Function Function and Personality Personality in humans presents three key "primitives": "primitives": I1)) A temporal function which seems to to obey some principle of of symmetry of of past and future, memory and prospective, with respect to the present. This temporal memory function is mostly based on the the capacity to evaluate and record the order of of occurrence occurrence of of event sequences (a kind of of segmentation lost
eomplementarily the capacity capacity to recognize new from in frontal patients), and complementarily familiar events (a kind of of fusion of of events lost in hippocampal patients). We have seen the importance of of the Hs in the recording of of "one exposure" events. The consequence of of the suppression suppression of of this memory function is illustrated by anterograde amnesia, i.e. the incapacity starting at a period of of life, corresponding to some time prior to the lesion, to build up a continuing
history, just just as if if the factual factual lif lifee of of a person had stopped at this moment. Yet, the Pc is also involved in this historical function, as one of avoured of the ffavoured
J.P. Banquet Banquet et al.
1177 77
cortical cortical sites for the permanent recording of of these event memories. Symmetrically, the the prospective function supposes the capacity to project in the the future future an orderly orderly sequence of of planned planned events in order to either actually perform them or merely simulate them. This planning capacity is also an hallmark of of personality. This function is as important as the previous one, and in fact intimately linked to it. Our capacity to make plans, i.e. to project our actions in the future, is narrowly dependent on a library of of past behavioral schemes and of of their consequences. As our personal history goes back in time as far as our early childhood, our ability to project our life in the future concerns more or less remote time. The range of of this prospective capability is closely linked to the strength and integration of of our personality, and supports our motivation. Mostly from neuropsychological neuropsyehological studies, the role of of the Pc in this function is well documented. Pc is essential not only for the the strict and and logical ordering of of events or actions. Furthermore, it operates in the the determination of of an hierarchy of of subgoals and actions to reach a The incapacity to forecast the consequences of of actions predetermined goal. The could be responsible, along with the neutralisation of of affective life, for the incoherent and self-destructive behavior eventually encountered in prefrontal patients. of the temporal 2) Working memory can still be considered as a part of function. Nevertheless its unifying role, and its implication in practically other function related related to personality deserves a separate account. The every other historical and prospective function, in particular, could not exist without the support of of an "extended present", i.e. a working memory. The capacity to link successive, logically related events oriented towards the performance of of a task, or the accomplishment of of a goal, is essential to the development of of WM is not not present in early childhood and and this absence explains personality. WM the non-permanence of of hidden objects in the field of the of consciousness as internal representation, and therefore the incapacity to perform delayed tasks. This representation, capacity progressively develops during during infancy, and probably supports the unfolding of of logical reasoning. This logical function is a prototypical illustration of the characteristic of of WM WM defined as both both maintenance and of the of information over an extended period of manipulation of of time. Classically, Pc is endowed with WM are temporal and and plausibly other WM capacity, but but so also are structures. Our contention is that Hs also partakes of of cortical or subcortical structures. of WM, even though an automatic aspect of though delay neurons have not been recorded there, as in different cortices. We We have proposed at at least three recorded subsidiary mechanisms that could support this function at Hs level.
1178 78
Chapter 4
3) Emotions and motivations are arc another important facet of of personality. One could hardly contend that a robot is a person even though by its previous experience experience it can have a semblance of of history. Emotions could be viewed as of basic drives resulting from sublimation, but certainly not a suppression, of including the instinct for survival and even the instinct of of death. Motivation could result from an integration and a trade-off trade-off between the need for drive satisfaction, satisfaction, emotions, and social constraints. Motivation is usually dependent on the degree of of satisfaction satisfaction of of these needs. Here Hero also limbic system and Pc act conjointly. Amygdala is as important for emotional life lifo as hippocampus act proper is important for correlational and WM ito WM functions. Similarly, orb orbitofrontal and meAial medial Pc are essential for the integration of of drives, emotions and motivations while more cognitive infonnation information processed in dorso-Iateral dorso-lateral Pc. The suppression of of any type of of affective affr162 colour, positive or negative, in the of severely damaged prefrontal patients, as after lobotomy, induces a life of disengagement from real real life. life. This underlies the the importance of of the integrative function of of Pc, in particular, particular, between cognition and emotion. These Those different functions are not compartmentalized. In particular particular emotional charge of of events, as previously mentioned, modulates the process of memory of consolidation in the hipp hippocampo-cortical of these consolidation ocampo-cortical system. The cooperation of different functions is perhaps best perceived in the mechanism of of attainment of of goals. 4) Attainment of of goals can be considered as the uttennost uttermost expression of of the cooperation between Pc and limbic system. This function presents sensory aspects which consist of of recognition of of goals and evaluation of of the outcomes of of action, and a motor aspect made up of of the setting and execution of of motor programs. In the classical learning theory, such as proposed by Skinner ((1953), 1 953), the necessary necessary chaining of of sequences of of sensory-motor events results from associative (or operant) conditioning of of a neutral stimulus by a reinforcer. Cascades of of secondary, and higher order, conditioning could account account for linking sequences of of events together. Obviously, this process can be and has been accounted for without without the extensive implication of of prefrontal cortex as in our model (Gray 99 1 ). The main structures concerned are (Gray et ct aI., al., 11991). hippocampus, amygdala and basal ganglia. They certainly correspond to a kind of of automatic operation mode for the attainment of of goals. Nevertheless, several problems arise arise if if the the basic components of of sequences, plans, or chained actions remain limited to stimulus-response conditional associations. In particular, 948) does not obey any clearly particular, latent learning (Tolman, 11948) defined drive satisfaction, motivation or goal attainment. The conditioning process seems to work correctly for simple sequences of of actions. But, taking
J.P. Banquet et al. JP.
1179 79
into account simply the present states in the recognition or reproduction of into of actions leads to a combinatorial explosion explosion of possible long sequences of an active representation of more more than paths, which can only be avoided by an one pair of events. Furthermore, the strength of secondary or higher higher order reinforcers seems to sharply decrease with the distance to the unconditional Sigmficant progress has been achieved by the identification of of latent stimulus. Significant learning useful to build cognitive maps, even if these maps are used of any prespecified goal. We think that the definition of goals independently of based on the satisfaction of basic (or not so basic) needs conjointly with the of more or less complex maps is a further step required to account learning of for complex behaviors. The possibility for diffusion, and in particular of goals allows the discovery of of solutions that have never retropropagation of been experienced during learning, and thus are created from new by the of creativity. This functioning mode system. This is an actual illustration of of freedom in the system, independent from requires a supplementary degree of both sensory and motor processing, but still bridging the two systems. This extra degree of of freedom is provided by Pc. The efficacy of of an algorithmic extra version of of the model for the parsimonious solution of of several problems of of robotic learning and navigation either in free space or maze constraint does of biological plausibility for the system. not automatically deliver a certificate of Nevertheless, the fact that this efficacy has been obtained thanks to a of essential neurobiological constraints, makes stringent taking into account of us confident that that the model is oriented in aa relevant direction. Conclusion Conclusion
Two different forms of supported of memory, "active" and "dormant", supported respectively by post post synaptic potentials (PSPs) and synaptic potentiation are present everywhere in the brain. The interplay between the two forms of of memory and in particular the transition and/or and/or the modulation of of one form by the other are at at the bases bases of of the different processing modes and memory capacities capacities of of the the brain. brain. Variations in the the implementation modalities and and in the the ranges of of these two two types of of memory along with variation of of connectivity give functional functional specificities to to the the different systems. This This is specially true true for Hs Hs and Cortical processing processing depends essentially on two two memory registers, and Pc. Cortical and permanent permanent LTM. LTM. The transition from the long-term to to the short STM The transition the shortSTM and term term seems to to be be direct direct and and normally encounters encounters few problems. Plausibly, at at aa gross level of of analysis analysis the the anatomical anatomical substrates substrates are are topographically identical (STM at aa fine (STM representing active active forms of of LTM). LTM). Nevertheless, at
1 80 180
Chapter 4
grainexi grained level, level, the the neurophysiological supporting supporting mechanisms must must be necessarily different different as as previously mcntionexi, mentioned, involving respectively either electrical or or durable durable structural-chemical changes. Conversely, the transition transition from from STM STM to to LTM LTM store store follows follows aa more intricate path, probably probably for for the sake of optimizing the amount amount of of information infonnation stored, but but also for securing the of storage of unique events up the unique history of of each living storage of events which built up being. Between these two two extreme ranges (STM and LTM), LTM), only minor variations from primary to associative areas can be recorded recorded at the sole cortical level, with aa tendency temporal range range of tendency to to an increase in the temporal of memory of complexity in processing performed perfonned by these areas with the increasing level of (Lii (Lil ~ et al., aI., 1992). 1 992). Prefrontal and temporal cortices arc are endowed with delay neurons that that can can bridge a a gap gap between two sensory or sensori-motor events. temporal range in the usual experimental tests of of this Nevertheless, the the tg~nporal i.e. less than 30 property remains largely largely in the domain attributed to STM, STM, i.e. secs. SCCS.
The specificity and vantage point of of Hs concerns both topographical and temporal facets. The topographical aspect of of Hs specificity as a unique compact site of convergence and output output divergence has been extensively of input convergence of Hs which emphasized. It has been credited with the correlational function function of implies some loss of of the the cortical cortical topology. This functional characteristic is corrected and and complemented by aa loose topological correspondence between hippocampal system in the longitudinal direction. This loose cortical and hippoeampal correspondence could be transfonned transformed into a dynamic learning-dependent precise mapping between hippocampal and cortical neuronal populations in order to implement the topologically specific consolidation function. function. This function could be implemented thanks to the fast-transient learning capacities present both within the hippocampus itself, and also at the interfaces between cortex and hippocampus. hippoeampus. The emphasis placed on the spatial aspects of of Hs function was detrimental to the exploration of of the no less important temporal function. This function results from the capacity of of Hs to interact very flexibly with a whole spectrum of of registers from the short-tenn short-term to the long longtenn, nnanent L TM. The unique term, and also possibly to be detached from pe permanent LTM. characteristic of of Hs would be the conjunction of of this array of of registers with a wide wide variety of of loops of of various sizes providing for an easy transition between donnant-inactive onns of dormant-inactive and re-activated fforms of memory. differences Beyond these range diff erences between cortex and Hs memory registers, differences of some more subtle diff erences could exist in the implementation modalities of active memory. Extensive research has has been conducted on delay cells cells in Pc Pe or temporal cortex, as a support for WM. Indeed, this type of of activity can bridge
JP. J.P. Banquet Banquet et al. aL
1181 81
the gap between two sensory or sensory-motor events. Up to now, delay cells have not been located at the level of of the Hs. Nevertheless, the equivalent function in Hs could be performed by different mechanisms subserving slightly different functions. First, loop iterative activation could operate the punctual reenactment of of recorded patterns of of activation either during information processing in WM, of L LTM TM consolidation. or during the more lengthy process of Second, spectral timing as performed in our model by DG could also operate the the function equivalent to that of of the cortical delay neurons. This of maintaining significant information in an active state, function consists of while waiting for correlation with a new significant event. This process creates the chaining chaining of of basic components of of the sequence. The Hs functional specificity would be in multimodal fusion and correlations. Finally, event locked and and modulated theta activity could constitute, at least for some species, a basic mechanism for the maintenance of a pattern in an active state, thus making possible a cross crosS correlation with forthcoming significant patterns. These types of of complementary "hardware" constraints in the implementation of of active memory and in the range of of "donnant" "dormant" registers determine the type of of cooperation established between the two structures Hs and cortex. Further complementarity results from the direct contact of of the cortex with environment, favoring externally triggered activation. Conversely, That property Hs is the only brain structure so easily prone to autoactivation. That leads, in the pathological domain, to seizure activity. The specific import of of Pc to this processing chain seems to result from its unique position at the top of the hierarchy of of sensory-motor and and motivational streams of of information of (Figure 11). 1 1 ). Both, its independence from and its close contact with multisensory and complex motor representations or codes provides the entire extra degree of of freedom. This feature gives to the brain the system with an extra capacity for: - Recording and simulating both sensory and motor sequences independently of of their actual implementation, in relation with planning and adaptation; of goals and subgoals, and goal - Motivated hierarchical selection of attainment; and creativity. - Finally, invention and These properties properties can be considered as the highest expression of of all these capacities. -
-
-
1182 82
Chapter 4 References References
Atkinson, R. C., & Shiffrin, R. M. ((1968). memory: A proposed 1 968). Human memory: system and its control 1. T. Spence control processes. In K. W. Spence & J. (Eds.), The psychology of of learning and motivation: motivation" Advances in 1 85). New York: Academic Press. research and theory (Vol. (Vol. 2, pp. 8989-185). 1 986). Workin Working Baddeley, A. D. ((1986). g memory. Oxford: Oxford University Press. Baddeley, A. D. ((1995). 1 995). Working memory. In M. S. Gazzaniga (Ed.), The cognitive neurosciences (pp. 755-764). Cambridge, MA: MIT Press. Baddeley, A. D., & Hitch, G. J. 1. ((1974). 1 974). Working memory. In G. A. Bower (Ed.), The psychology of of learning and motivation: motivation: Advances in research and theory (Vol. (Vol. 8, pp. 47-90). New York: Academic Press. Baddeley, A. D., & Warrington, E. K. ((1970). 1 970). Amnesia and the distinction between f Verbal Learning and between long-and short-term memory. Journal oof 176-189. Behavior, 9, 176Verbal Behavior, 1 89. Banquet, 1. 1 983). Inter and intra-hemispheric J. P. ((1983). intra-hemispheric relationships relationships of the EEG during sleep in man. m a n . Electroencephalography Electroencephalography and Clinical Neurophysiology, 55, .55, 551-59. 1-59. Banquet, 1. 1 987). Probing cognitive processes through J. P., & Grossberg, S. ((1987). the structure of of event-related event-related potentials during learning: learning: An experimental and theoretical analysis. Applied Optics, 26, 493 4931-4946. 1 -4946. J. P., Renault, B., & Lesevre, Les~vre, N. (1981). Effect of of task and stimulus Banquet, 1. probability on evoked potentials. 14. potentials. Biological Psychology, Psychology, 13, 203-2 203-214. Banquet, 1. J. P., Smith, M. 1., J., & Renault, B. ((1990). top1990). Bottom-up versus top down: down: An alternative to the automatic attended dilemma? Behavioral 3, 233-234. and Brain Sciences, Sciences, 113, 233-234. Banquet, 1. (1 992a). Temporal order, timing, J. P., & Contreras-Vidal, 1. J. L. (1992a). categorization. and probability context effects on pattern recognition and categorization. In I. Aleksander 1. Taylor (Eds.), ArtifiCial Aleksander & J. Artificial neural networks 2 (pp. 1 885-1 890). Amsterdam: Elsevier. 1885-1890). Banquet, J. 1. P., & Contreras-Vidal, J. 1. L. ((1992b). 1 992b). An integrated neural network-event network-event related potentials model of temporal and probability context effects on categorization. categorization. In Proceedings o f the International of Joint Con ference on Neural Networks (pp. 54 1 -546). Hillsdale, NJ: Conference 541-546). Lawrence Erlbaum. J. P., & Contreras-Vidal, J. Banquet, 1. 1. L. ((1993a). l993a). Learning temporal contexts and priming preparation modes for pattern recognition. recognition. In Proceedings o f the International World Conf erence on Neural Networks (pp. 26(pp. 1126of Conference NJ" Lawrence Erlbaum. 1131). 3 1). Hillsdale, NJ:
JP. Banquet et al.
1183 83
J. P., & Contreras-Vidal, 1.J. L. ((1993b). timing and Banquet, 1. l 993b). Spectral timing J. Taylor (Ed.), integration of multimodal systemic processes. In 1. Artificial neural networks (pp. 350-354). Amsterdam: Elsevier. P.,. , & Contreras-Vidal, JJ.. L L.. ((1994). Banquet, JJ.. P 1 994). Medium and long-term network model of of cortex-hippocampus memory in context processing: A network. f the International Conf erence on Neural of Conference relations. Proceedings o 647-654.. Networks, 4, 647-654 J. P., Contreras-Vidal, JL., Gaussier, P., Gissler, A., & Burnod, Bumod, Y Y.. Banquet, 1. cortico-hippocampal ((1997). 1 997). The corti co-hippocampal system as a multirange temporal processor: A neural model. In R. Park & D. Levin (Eds.), Fundamentals oof f neural network modelling f or neuropsychologists. Boston: MIT for Press. S.,. , & Levine, D. S. ((1990). of Bapi, R. S 1 990). Networks modelling the involvement of the frontal lobes in learning and performance of of flexible movement f the sequences. sequences. In Proceedings of of TWl!lth Twelth Annual Conference oof Cognitive Science Society (pp. 9915-922). 1 5-922). of the frontal lobes in Bapi, R. S., & Levine, D. S. ((1994). 1 994). Modelling the role of task performance. II:: Basic structure and primacy effect. sequential task effect. Neural Networks, 7, 11167-1180. 1 67-1 1 80 . Baylis, G G.. c C.,. , & Rolls, E. T T.. (1 (1987). of neurons in the inferior 987). Responses of temporal cortex in short-term and serial recognition memory tasks. 6.5, 6614-622. Experimental Brain Research, 65, 14-622. Berger, T. W., & Thompson, R. F. (1978). ( 1 978). Neuronal plasticity in the limbic system during classical conditioning of of the rabbit rabbit nictitating membrane response. I: The hippocampus. Brain Research, 114.5, 45, 323-346. Br6bion, Brebion, J. 1. (1994). ( 1 994). Memoire de travail, comprehension de texte et vieillissement. Unpublished doctoral dissertation, University Ren6 Rene Descarte, Paris. Brown, J. tests of 1 958). Some tests of the decay theory of of immediate memory. 1. ((1958). Quarterly Journal Journal of O, 112-21. ofExperimental Psychology, 110, 2-2 1 . Bullock, D., Fiala, 1J.. C., & Grossberg, SS.. ((1994). of timed 1 994). A neural model of response learning in the cerebellum. Neural Networks, 7, 1101-1114. 1 10 1 - 1 1 1 4 . cerebellum. Neural Burgess, N., Reece, Reece, M., & O'Keefe, O'Keefe, I. 1. (1994). ( 1 994). A model of of hippocampal hippocampal function. function. Neural Neural Networks, 7, 1065-1081. 1 065-1 08 1 . Buzsaki, G. (1988). ( 1988). Polysynaptic long-term potentiation: A physiological role of the Brain of the perforant perforant path path - CA3/CA1 CA3/CA I pyramidal cell synapse. Brain Research, 4.55, 455, 192-195. 1 92-195 . Buzsaki, G. G . (1989). ( 1989). Two-stage model of of memory trace formation: A role for 'noisy' brain states. Neuroscience, NeurOSCience, 31, 551-570. 5 5 1 -570.
1184 84
Chapter 4
Cave, C. B., & Squire, 1992). Intact and long-lasting visual object Squire, L. R. ((1992). f Experimental priming in amnesic patients. Journal o of Experimental Psychology: Psychology: Learning, Memory, and Cognition, Cognition, 18, 509-520. Cressant A., Muller R., & Poucet B. ((1997). 1 997). Failure of of centrally placed objects to control the firing fields of ocampal place cells. f of hipp hippocampal cells. Journal oof Neuroscience, 17. 1 7. Neuroscience, Darke, S. ((1988). 1 988). Anxiety and working memory capacity. Cognition Cognition and 145-154. 54. Emotion, 2, 145-1 Dehaene, S., Changeux, 1. 1 987). Neural networks that J. P., & Nadal, 1. J. P. ((1987). Proceedings oof f the National learn temporal sequences sequences by selection. selection. Proceedings Natwnal 1. Academy o fScience of 2727-2731. of of the USA, USA, 84, 2727-273 Denham, M 996) A model M.. 1., J., & Boitano, 1. J. (1 (1996)A model of of the interaction interaction between between
pref rontal cortex, prefrontal cortex, septum septum and the hippocampal hippocampal system system in the learning and recall oof f goal-directed sensory-motor sensory-motor behaviours. behaviours. Technical Report NRG-96-0 1 , University of NRG-96-01, of Plymouth School School of Computing. Drachman, D. A., & Arbib, M. ((1966). Drachrnan, 1 966). Memory and the hippocampal complex. Archives of 1. of Neurology, Neurology, 15, 52-6 52-61. Eichenbaum, H o, T., & Cohen, 1994). Two functional H.,, Ott Otto, Cohen, N N.. 1J.. ((1994). components of the hippocampal memory system. system. Behavioral Behavioral and Brain Science, 117, 449-518. 7, 449-5 18. Science, arousal: Cognition Cognition and per performance. M.. W W.. ((1982) Eysenck, M 1 982) Attention and arousal: formance. New York: Springer-Verlag. Funahashi, SS.,. , Bruce, C., & Goldrnan-Rakic, Goldman-Rakic, P. W. (1 (1989). Funahashi, 989). Mnemonic coding of of visual space by neurons in the monkey's dorsolateral dorsolateral prefrontal cortex revealed by an oculomotor delayed-response task. Journal o f of Neurophysiology, 61, 33 331-349. 1 -349. Neurophysiology, J. M., & Alexander, G. E. ((1971). relatext to short shortFuster, 1. 1 971). Neuron activity related term memory. 73, 652-654. memory. Science, Science, 1173, Fuster, 1J.. M. ((1980). 1 980). The pref rontal cortex: prefrontal cortex: Anatomy, Anatomy, physiology and neuropsychology o f the f rontal lobe. lobe . New York: Raven Press. of frontal J. M. ((1995). cerebral cortex. cortex. Cambridge, MA: MIT Fuster, 1. 1 995). Memory in the cerebral Press. Fuster, 1. J. M., & Jervey, J. 1. P. ((1981). 198 1). Inferotemporal neurons distinguish and Science, 2212, behaviorally relevant features of visual stimuli. stimuli. Science, retain behaviorally 1 2 52555. 5. architectures for autonomous Gaussier, P., & Zrehen, S. ((1994a). 1 994a). Complex architectures agents. In P. Gaussier & J. D.. Nicoud (Eds.), (Eds.), PerAc PerAc (pp. (pp. 278-290). 1. D Lausanne: IEEE Press. ,
J.P. Banquet et al. JP.
1185 85
S.. ((1994b). Gaussier, P., & Zrehen, S 1 994b). Navigating with an animal brain :: A neural network for landmark identification and navigation. In Proceedings of Vehicles (pp. 399-404), Paris. of Intelligent Vehicles Zmhen, S. ((1995). Gaussier, P., & Zrehen, 1 995). PerAc: A neural architecture to control Systems, 116, 291-320. artificial animals. Robotics and Autonomous Systems, 6, 29 1-320. A., & Banquet, 1. J. P. ((1996). Gaussier, P., Joulain, C., Revel, A, 1 996). Are shaping techniques the correct answer for the control of of an autonomous robot ? of UKACC UKACC International Conf Conference '96, erence on Control '96, In Proceedings of IEEE. C.,. , Revel, A A. Zrehen, S, & Banquet, 1. J. P. ((1997a). Gaussier, P., Joulain, c l 997a). Autonomous robot learning: What can we take for free? To appear in Proceedings of of IEEE conference on ISlE, ISIE, IEEE. C.,. , Zrehen, S. Banquet, 1. J. P, & Revel, A A. ((1997b). Gaussier, P., Joulain, c l 997b). Visual navigation in an open environment without map. To appear in Proceedings of erence. of IROS Conf Conference. Glanzer, M., & Amitz, A. A R. ((1966). 1 966). Two storage mechanisms in free recall. Behavior, 5, 35 351-360. Journal oof f Verbal Learning and Verbal Behavior, 1 -360. Goldman-Rakic, P. W. ((1988). 1 988). Topography of of cognition: Parallel distributed networks in primate association cortex. Annual Review Review oof fNeuroscience, 111, 1 , 1137-156. 37- 1 56. Goldman-Rakic, P. W. ((1994). 1 994). Working memory dysfunction in schizophrenia. Journal o f Neuropsychiatry of Neuropsychmtry and Clinical Neuroscience, 6, 348-357. J., & Lynch, G. ((1994). Granger, R., Whitson, J., Larson, 1., 1 994). Non-Hebbian of long-term potentiation enable high-capacity encoding of properties of fScience oof f temporal sequences. Proceedings of of the National Nanonal Academy oof the USA, USA, 91 91,, 110104-10108. 0 1 04- 1 0 1 08 . Gray, C 1 989). Oscillatory C.. M., Konig, KOnig, P., Engel, A A.. K K.,. , & Singer, W. ((1989). responses in cat visual cortex exhibit intercolumnar synchronization which reflect global stimulus properties. Nature, Nature, 338, 338, 334-337. A., Feldon, J., Rawlins, 1. J. N. P., Hemsley, D. R., & Smith, A A. D. Gray, 1J.. A, ((1991). 1 99 1 ). The neuropsychology of of schizophrenia. Behavioral and Brain. Sciences, 114, 4, 11-20. -20. Sciences, Grossberg, S. ((1976a). l976a). Adaptive pattern classification and universal recoding. I: Parallel development and coding of of neural feature detectors. detectors. receding. Cybernetics, 23, 12 121-134. Biological CybernetiCS, 1 - 1 34 . S.. ((1976b). Grossberg, S l976b). Adaptive pattern classification and universal receding. recoding. II: Feed-back, expectation, olfaction, and illusions. Biological Cybernetics, 23, 1187-202. 87-202. CybernetiCS,
1186 86
Chapter 4
Grossberg, S 1 978). A theory of S.. ((1978). of human memory: memory: Self-organization and performance of sensory-motor codes, codes, maps, and plans. In R. Rosen & F. Snell (Eds.), (Eds.), Progress in theoretical theoreacal biology (Vol. (Vol. 5, pp. 233-374). New York: Academic Press. Grossberg, S., & Merrill, 1 992). A neural network model MerriU, J. ((1992). model of of adaptively timed reinforcement ocampal dynamics. reinforcement learning and hipp hippocampal dynamics. Cognitive Cognitive Brain Research, 11,, 3-38. Guigon, E., Dorizzi, B., Bumod, Y., & Schultz, 1 995). Neural correlates Schultz, W. ((1995). of learning in the prefrontal cortex of the monkey: monkey: A predictive predictive model. model. Cerebral Cortex, 35-147. Cortex, 2, 1135-147. Halgren, E., Squires, Halgrcn, Squires, N. K., Wilson, C. L. Rohrbaugh, J. W., Babb, T. L., & CrandaU, P. H. (1980). Endogenous potentials generated in the human Crandall, ( 1980). Endogenous hippocampal formation and amygdala by infrequent infrequent events. events. SCience, Science, 210, 803-805 803-805.. Hasselmo, M. E., & Schnell, 1994). Laminar selectivity SchneU, E. ((1994). selectivity of the cholinergic suppression suppression of of synaptic transmission in rat hippocampal region region CA Il:: Computational modeling f modeling and brain slice slice physiology. physiology. Journal oof Neuroscience, 114, 4, 3898-3914. 1994). The localization of general memory Horel, J. A. ((1994). memory functions. functions. and Brain Sciences, 117, Behavioral and 7, 482. Howarth, E., & Eysenck, H. J. ((1968). paired1 968). Extraversion, arousal, and paired recall. Journal oof Research in Personality, 3, f Experimental Research associate recall. 1 14-1 1 6. ll4-116. Ivry, R 1 989). Timing functions R. B., & Keele, Keele, S. W. ((1989). functions of the cerebellum. Journal of 34- 1 50. of Cognitive Neurosciences, 11,, 1134-150. James, W. ((1890). 1 890). The prinCiples fpsychology. New York: principles oof York: Holt. R.,, Jr. ((1995). P300: Evidence Evidence from Johnson, R 1 995). On the neural generators of the P300: temporal lobeetomy lobectomy patients. Electroencephalography Electroencephalography and Clinical 10-129. Neurophysiology, 44 (Supplement), (Supplement), 1110-129. R., Jr., & Donchin E. ((1982). Sequential expectancies expectancies and decision Johnson, R, 1982). Sequential environment: An An electrophysiological approach, making in a changing environment: Psychophysiology, 19, 183-199. 1 83-199. S.. G. ((1993). connections: A speculative Jones, R. R S 1 993). Entorhinal-hippocampal connections: view of 6, 58-64. TINS, 116, of their function. function. TINS, Smith, E. E., Koeppe, Koeppe, R R. A., Awh, Awh, E., Minoshima, Minoshima, SS.,., & Mintun, Jonides, J., Smith, M. A. ((1993). revealed by PET. 1 993). Spatial working memory in humans as revealed Nature, 363, 363, 623-625 623-625..
J.P. JP. Banquet et al.
1187 87
D.. SS.,. , & Park, R. ((1992). Levine, D 1 992). Frontal lesion effects on verbal fluency in a of Conference network model. Proceedings o f the International Joint Conf erence on Neural Networks, 2, 39-44. J., & Kaufman, Lii, Z. L., Williamson, S. 1., Kaufinan, L. ((1992). 1 992). Human auditory primary and association cortex have differing lifetimes for activation 572, 236-241. traces. Brain Research, 5 72, 236-24 1. A. P. Smith & D D.. M. Jones (Eds.), Matthews, G. ((1992). 1 992). Extraversion. In A of human per performance. Vol. 3: State and trait (pp. 9595-126). Handbook of formance. Vol. 1 26). London: Academic Press. A. M., Puce, A, A., Nobre, A A. C. Bloch, G., Hyder, F., McCarthy, G., Blamire, A Goldman-Rakic, P. W., & Shulman, R. ((1994). 1 994). Functional magnetic imaging of of human prefrontal cortex activation during a spatial working memory task. Proceedings o of f the National Academy oof f Science of of the USA, 91, 8690-8694. Miller, E. K., Li, L., & Desimone, R. ((1993). of neurons in anterior 1 993). Activity of of inferior temporal cortex during a short-term memory task. Journal of 1460-1478. Neuroscience, 113, 3, 14601478 . Neuroscience, Desimone, R. ((1994). Miller, E. K., & Desimone, 1 994). Parallel neuronal mechanisms for Science, 263, 520-522. short-term memory. Science, Milner, B. ((1966). 1 966). Amnesia following operation on the temporal lobes. In C. W. M. Whitty & O. L. Zangwill (Eds.), Amnesia. London: Butterworths. S.,. , & Teuber, H.-L. ((1968). of the Milner, B., Corkin, S 1 968). Further analysis of of H H.. M. hippocampal amnesia syndrome: 14-year follow-up study of Neuropsychologia, 1 5-234. Neuropsychologia, 6, 2215-234. 1 982). Some cognitive effects of of frontal lobe lesions in man. Milner, B. ((1982). of the Royal SOCiety, Society, London, B, 298, 2211Philosophical Transactions of 1 1226. of pictorial short shortMiyashita, Y., & Chang, H. S. ((1988). 1 988). Neuronal correlate of 331,, 68-70. term memory in the primate temporal cortex. Nature, 331 Niki, of prefrontal units during right and left iki, H. ((1974). 1 974). Differential activity of N delayed response trials. Brain Research, Research, 70, 346-349. Naatanen, 1 982). Processing Negativity: An evoked-potential reflection of N~i/it~en, R. ((1982). of selective attention. Psychological Bulletin, 92, 605-640. J., & Nadel, L. ((1978). O'Keefe, 1., 1 978). The hippocampus as a cognitive map. Oxford: Oxford University Press. Olton, D. S., Becker, 1. 1 979). Hippocampus, J. T., & Handelmann, G. E. ((1979). Science, 2, 3313-365. and memory. Behavioral and Brain Science, space and 13-365 .
1188 88
Chapter 4
Owen, A. M., Morns, Morris, R. G., Sahakian, Sahakian, B.
1., J., Polkey, C. E., & Robbins,
1 996). Double T. W. ((1996). Double dissociations dissociations of of memory and executive functions ollowing lobe excisions, temporal lobe in working working memory tasks ffollowing 19, 11597597excisions or amygdalo-hippocampectomy in man. Brain, Brain, 1119, 11615. 615. Paulescu, E 1 993). The neural E.,. , Frith, C C.. D D.,. , & Frankoviak, R R.. SS.. 1J.. ((1993). Nature, correlates of the verbal component of working memory. of of Nature, 362, 362, 342-345. 342-345. Peterson, 1 959). Short-term retention of Peterson, L. R., & Peterson, Peterson, M. 1. J. ((1959). of individual /Experimental verbal items. Journal o of Experimental Psychology, Psychology, 58, 193-198. Petrides, 1 993). Dissociation Petrides, M., Alivisatos, B B.,. , Evans, A. C., & Meyer, E. ((1993). of human mid-dorsolateral from posterior dorsolateral frontal cortex in of /the National Academy / memory processing. Proceedings Proceedings 0of Academy 0/ of Science Science 0of USA, 90, 873-877. the USA, effect Rolls, E. T., Baylis, G. C., Hasselmo, M. E., & Nalwa, V. ((1989). 1 989). The eff ect of learning on the face selective responses responses of of neurons in the cortex in the of superior temporal sulcus in the monkey. monkey. Experimental Experimental Brain Brain Research, Research, 76, 153-164. Rolls, E. T., & O'Mara, S S.. M., (1995) View-responsive View-responsive neurons in the hippoeampal complex. Hippocampus, Hippoeampus, 5, 409-424. 409-424. primate hippocampal Ruchkin, D. S . , Sutton, S . , Mahaff ey, D., & Glaser, 1. ((1986). 1 986). Terminal Ruchkin, S., S., Mahaffey, J. CNV in the absence of of a motor response. Electroencephalography Electroencephalography and Clinical Neurophysiology, Neurophysiology, 63, 445-463 445-463.. Ruchkin, 1991). EventRuehkin, D. SS.,. , Johnson, R., Jr., Canoune, H., & Ritter, W. ((1991). related potentials during arithmetic and mental rotation. Electroencephalography Electroencephalography and Clinical Neurophysiology, Neurophysiology, 79, 79, 473-487. Ruehkin, D. S., Johnson, R., Jr., Grafinan, Grafman, 1., J., Canoune, H., & Ritter, W. Ruchkin, Distinctions and similarities among working memory processes: ((1992). 1 992). Distinctions An event-related event-related potentials study. Cognitive Cognitive Brain Brain Research, Research, 11,, 53-66. Schacter, D. L., Chu, C. -Y., & Ochsner, K. N. ( 1 993). Sehacter, Oehsner, (1993). Implicit memory: A /Neuroscience, 116, 6, 1159-82. 59-82. selective review. review. Annual Review oofNeuroscience, Scoville, 1 957). Loss of Seoville, W. B., & Milner, MiMer, B B.. ((1957). of recent memory after bilateral hippocampal hippoeampal lesions. Journal 0/ of Neurology, Neurology, Neurosurgery, Neurosurgery, and 1 -2 1 . Neuropsychiatry, Neuropsyehiatry, 20, 111-21. Shiffrin, R. M., & Schneider, W. ((1977). 1 977). Controlled and automatic inf ormation processing. information processing. II: Perceptual leaming, learning, automatic attending and a general theory. Psychological Psychological ReView, Review, 84, 1127-90. 27-90.
J.P. Banquet et al.
1189 89
Sidman, M 1. P. ((1968). 1 968). Some additional M.,. , Stoddard, L. T., & Mohr, J. quantitative observations of of immediate memory in a patient with bilateral hippocampal lesion. Neuropsychologia, 6, 245-54. Singer, W. ((1983). 1 983). Neuronal activity as a shaping factor in the self selforganization of of neuron assemblies. In E. Basar, H. Flohr, H. Haken, & Synergetics of the brain. brain. New York: Springer SpringerA. J. 1. Mandell (Eds.), S ynergetics of Verlag. Skinner, F. (1953). ( 1 953). Science and human behavior. behavior. New York: McMillan. Solomon, P 1 980). A time and a place for everything? Temporal P.. R R. ((1980). processing views of of hippocampal function with special reference to attention. PhYSiological 1. Physiological Psychology, 8, 254-6 254-61. Squire, L. R R.,, Ojemann, 1J.. G., Miezin, FF.. M., Petersen, SS.. E., Vdeen, T. 0 O.,. , of the hippocampus in normal & Raichle, M. E. ((1992). 1 992). Activation of humans: A functional anatomical study of f the of memory. Proceedings oof 837-4 1 . National Academy o fScience oof f the USA, of USA, 89, 11837-41. Tolman, E 1 948). Cognitive maps in rats and men. The Psychological E.. C C.. ((1948). ReView, 55, 189-208 1 89-208 Review, Zrehen., S. ((1995). 1 995). Elements o f brain design f or autonomous agents. of for agents. Unpublished PhD thesis, Swiss Federal Institute of of Technology, Lausanne. Zrehen, S., & Gaussier P 1 997). A neural architecture f or motivated P.. ((1997). for landmark-based landmark-based navigation. navigation. ETIS internal report (submitted for publication) publication).. Author Author Note Note
This research was supported by INSERM, NATO and DGAIDRET DGA/DRET 911470/A000/DRET/DS/DR. Grant # 9 1 1 470/AOOOIDRETIDSIDR
This Page Intentionally Left Blank
P A R T II PART II PERSPECTIVES FROM FROM EMOTION EMOTION RESEARCH RESEARCH PERSPECTIVES
This Page Intentionally Left Blank
Cognitive Science Science Perspectives Perspectives on Personality Personality and and Emotion Emotion Cognitive
- G. G. Matthews Matthews(Editor) (Editor)
-
1997 V. 1997 Elsevier Elsevier Science Science B. B.V.
CHAPTER S5 CHAPTER Affective Influence in Perception:
Some Implications of the
Amplification Model Amplification Ki tayama Shinobu Kitayama Will affectively charged stimuli be perceived any differently from affectively neutral ones? Will affect inherent in a focal stimulus disrupt affectiveiy perceptual processing? Or will it facilitate the latter so that affective stimuli field? After years of waxing and waning (e.g., stand out in the perceptual field? Dixon, 11980; 980; Erdelyi, 11974, 974, 11985), 985), the influence of stimulus affect on immediate perception remains as a topic of considerable significance. Mechanisms underlying immediate conscious perception are both logically (e.g.,, Helmholtz, 11884) (e.g. 884) and empirically (e.g., Marcel, 11983a) 983a) preconscious. if it can be shown that that the immediate perception of of a stimulus stimulus is indeed indeext Thus, if influenced by affect inherent in the stimulus itself, we will have identified a window through which to observe what Kihlstrom ((1990) 1 990) has called the unconscious. psychological unconscious. . Championed by Freud and his successors (e.g., Freud, 11895/i966), 895/1 966), the functional structure of of the unconscious, especially the one involving affect, has turned out to be one of of the most formidable problems in psychology, often evading scientific scrutiny. However, with rigorous experimental experimental evading methodologies and theoretical tools now available at hand, recent investigations detectionless processing (e.g., Bargh, Bond, Lombardi, & investigations on detectionless Tota, 1986; 1 986; Carr & Dagenbach, 1990; 1 990; Greenwald, Klinger, & Lui, 11989; 989; Marcel, 1983a,b; Niedenthal, 1990; 1 983a,b; Niedenthal, 1 990; Shevrin, 11990), 990), automatic processing Schneider, 11977), (e.g., Uleman & Bargh, 1989; 1 989; Shiffrin & Schneider, 977), and implicit memory (e.g., Schacter, 1989) significant steps toward more 1 989) have taken significant comprehensive comprehensive and accurate understanding of of the unconscious. unconscious. And a new theoretical framework has begun to emerge (e.g., Erdelyi, 1985; 1985; Lewicki, 1986; 1 986; Kihlstrom, Kihlstrom, 1990; 1 990; Marcel, 1983b; 1 983b; Rumelhart, 1989; 1 989; Zajonc, 1980). 1 980). The present paper seeks to contribute to this literature. The literature. We will examine whether and and how the the perceptibility perceptibility of of a faintly shown stimulus can vary with the affective significance significance of of the stimulus stimulus itself. itself. The goal is to identify distinctly affective phenomena in a perceptual identification identification task, and integrate them with current current theories of of cognition, affect, and attention, understanding other forms of attention, thereby laying a solid foundation for understanding of
194
Chapter 5
"hot cognitions" (Abelson, 11963). 963). Should perceptibility depend on affective significance, significance, that would suggest that affect can be preconsciously elicited by an impinging stimulus and, moreover, that the elicited affect influences subsequent processing required to develop a conscious percept. Accordingly, of revealing the nature of of a preconscious the present work has the potential of interaction between affect and cognition. The current approach emphasizes a non-associative, energizing consequence consequence of of affect (cf. (of. Osgood, 1962) and analyzes how stimulus affect influences the perception of of the stimulus itself. Specifically, it will be proposed that affect evoked through preattentive processing amplifies attentive processing, thereby either enhancing or impairing the emerging conscious percept. This work therefore will supplement a presently dominant, 986; Bower, 11981; 98 1 ; largely associative approach to affect (e.g., Bargh et aI., al., 11986; 986; Greenwald et aI., Fazio, Sanbonmatsu, Powell, & Kardes, 11986; al., 1989; 1989; !sen, Isen, Shalker, Clark, & Karp, 11978; 978; Johnson & Tversky, 11984; 984; Lang, 11084; 084; 1 990; Zajonc, 989), which has proved Niedenthal, 1990; Zajone, Murphy, & Ingelhart, 11989), powerful in analyzing how affect of of one stimulus (prime) can bias the perception of of another (target). It is typically assumed in this literature that the activation of of affective affective information, caused by the prime, can can spread to related of associative memory, thus biasing the information within a network of of the target. perception of
historicalperspective: "New Look" and its a aftermath A historical perspective: The "New ftermath The general issue of of affect-cognition interaction in perception can can be back to the literature of traced back of "New Look" in perception in the 11950s 950s (Bruner, 11957). It was then proposed that perception depends not only on 957). exogenous factors, but also on endogenous factors including perceptual set, 1955, for a expectation, motivation, personality, and affect (see e.g., Allport, 1955, review). In a pioneering experiment on affect and perception, McGinnies MeGinnies of affectively affcctivdy charged words. He briefly ((1949) 1 949) examined the perception of flashed either a taboo word or a neutral word, and found recognition threshold to be considerably higher for the taboo word than for the neutral word. He maintained that this resulted from perceptual processes; affect ("anxiety") evoked by a taboo word recruited the process of psychological ("anxiety") defense, which blocked further perceptual processing, thus diminishing the that affect conscious percept (see also Blum, 11954). 954). Other studies observed that sometimes enhanced perception (e.g., Postman, Bronson, & Gropper, 11953). 953).
S. S.
Kitayama
1195 95
then suggested the the operation of The proponents of perceptual defense then perceptual vigilance. Critics quickly pointed out an apparent paradox involved in McGinnies's assertion that one can feel "anxiety" without knowing the identity of the stimulus. They maintained that the finding could be explained most parsimoniously by post-perceptual response biases such as reluctance or 1958: see readiness to report a taboo word (Eriksen, 1963; Goldiamond, 1958: Erdelyi, 11974; 974; Dixon, 11980 980 for reviews). There are some methods, however, that allow one to examine perceptual accuracy independent of post-perceptual response bias. For instance, subjects may be asked to choose the item shown of equivalently valenced words. Research employing this and from a pair of other similar procedures has demonstrated that the affective tone of a (e.g.,, Bootzin & Natsoulas, stimulus does influence accuracy in perception (e.g. 11965; 965; Dorfman, 11967). 967). Thus, McGinnies was correct in this regard. Further, to be reviewed below, recent cognitive research has strongly suggested that of a number of of preconscious conscious perception is the end product of operations. Hence, McGinnies's notion that affect can be induced by an impinging stimulus before the stimulus is consciously identified is no longer considered paradoxical (Erdelyi, 1974). of defense/vigilance mediate the Nevertheless, his theory that processes of of affect in perception has faced serious challenges. Neither he nor his effects of of defense or vigilance. Hence, no successors articulated the mechanisms of prediction about of affect is possible. Further, about the direction of of the influence of of a stimulus evidence suggests that that affect can influence the perceptibility of whether its valence is positive or negative (Broadbent & Gregory, 1967; 1 967; Kitayama, 1990, 1 990, 1991). 1 99 1 ). In retrospect, then, it would seem that McGinnies was correct in that that the the effect he observed was, at at least in part, part, perceptual. However, his hypothesis of of defense/vigilance as an underlying mechanism is increasingly suspect.
The present approach current paper presents aa model of of affect-cognition interaction The current designed to to account account for perceptual perceptual influence of of affect. We We reconsider this old rubric of of defense and vigilance, from problem, traditionally studied under the rubric by aa number of of theoretical and and methodological aa new perspective afforded by innovations innovations accomplished accomplished in the the interim. Informed by by current theories of of cognition, affect, and and attention, the the model hypothesizes that that affect induced through through preattentive processing of of an impinging stimulus amplifies
1 96 196
Chapter 5 Chapter
subsequent attentive processing. We We will review past past studies studies on preattentive subsequent processing, attention, and affect, which together form attention, and affect, fonn the empirical and and theoretical theoretical basis basis for the proposed model. It It will be shown shown that that the model coherent account account for for an an anomalous pattern pattern of of past past findings findings in the the provides aa coherent defense/vigilance model has guided guided more more recent defense/vigilance literature. Further, the model Evidence for the model will be reviewed, empirical investigations on the topic. Evidence and two new experiments will be reported. reported. Finally, broader implications of of for future research the proposed model will be discussed and directions for explored. explored. The present The present attempt to to test implications implications of of the model strategically just one form fonn of of perception: perception: word perception. perception. Studying focuses on just Studying this relatively simple case should make it possible to exert exert precise experimental control and, thus, to test the model in aa more rigorous fashion. Further, in the domain of of word perception, there is considerable overlap between research on affect of past research on aff affect affect and research on cognition. Much of ect and perception, dating back to McGinnies's original contribution, was done with words as stimuli, stimuli, and word perception has been extensively studied in current 989). cognitive psychology (see e.g., Posner, 11989).
The of Affect-Cognition Interaction The Amplification Model of
A model A model of ect-cognition interaction in early perceptual processing of aff affect-cognition (Kitayama,
11990, 990, 11991; 99 1 ; Kitayama & Howard, 11994) 994) is illustrated in Figure
11.. The basic tenet of ect induced of the model is that aff affect induceA through preattentive processing
of of
an
impinging stimulus amplifies amplifies subsequent attentive
processing, thereby either enhancing or impairing the conscious percept of of the stimulus. Unlike the def ense/vigilance hypothesis, this model assumes that an defense/vigilance influence ect in perception results from interaction among three influence of of aff affect component
processes
commonly implicated implicateA in ordinary processes of of
perception, i.e., preattentive processing, attentive processing, and aff ect. affect.
Preattentive processing. According to current cognitive theories of perceptual processing of of lexical lexieal materials such as words (e.g., McClelland & Rumelhart, 198 1 ; Posner, 11978) 978) and 1981; and certain graphic stimuli such as faces Damasio, Tranel & Damasio, (Bauer, 11984; 984; Damasio, D amas io, & Van Hoesen, 11982; 982; Tranel 11985), 985), an impinging stimulus is initially processed automatically without any involvement of of attention. Through preattentive processing, the graphic and emantic perceptual codes that that correspond to the stimulus can be possibly the ssemantic
1197 97
S. Kitayama S. ii
)"
Preattentive processing
,11111
i
i|1 ii iii
I Engagement of of attention in a relevant perceptual code
i
i
|l
Activation Activation of of alTective affective circuits and subsequent of amplification of attentive processing
Attentive
processing
Conscious
percept
Response out (i.e., reading out features, relevant reatures, selecting responses, etc.)
affectFigure 1. 1 . A schematic illustration of of the amplification model of of affect cognition interaction in early perceptual processing. The model is composed of, as its major components, preattentive processing, attentive processing, and of attentive processing, and activation of of affect and subsequent amplification of components are highlighted in bold squares. response. These components
activated before the conscious percept of of the stimulus is developed. A number of semantic priming paradigm have shown that that semantic of recent studies with aa semantic information (and, by implication, graphic information as well) can be activated by aa word that undcteetable in that is pattern masked and thus made undetectable consciousness consciousness (e.g., Allport, 1977; 1 977; Balota, 1983; Carr, McCaulcy, McCauley, Spcrber, Sperber, & Parmclee, ct aI., al., 11989; Parmelee, 1982; 1 982; Carr Carr & Dagcnbach, Dagenbach, 1990; 1 990; Grccnwald Greenwald et 989; Fowler, Wolford, Wolford, Sladc, Slade, Tassinary, Tassinary, 1981, 1 98 1 ; Marcel, 1983a, 1 983a, 1983b). 1 983b). This
198
Chapter 5
to the the hypothesis that attention, in the sense of of literature lends strong support to of processing mechanisms selectively and serially applied to a confined a set of spatial sp~/tial or semantic region, is not not necessary for the activation of of meanings of of an an impinging lexical lexieal stimulus, let alone the activation of of its shape. The The foregoing conclusion might seem at at odds with cognitive research on traditionally assumed that preattentive vision is very crude vision, which has traditionally (Egeth, 1977; Neisser, 1967). 1 967). For example, Sagi and colleagues (Braun & Sagi, 11990; 990; Sagi & Julesz, 11985) 985) have proposed that preattentive vision that a field contains a discrepancy, but does not enable one allows one to note that object. Similarly, for Treisman ((1988), to identify any object. 1988), preattentive vision is of a field (such as "green" or "square"), sufficient to identify separate features of but not any object defined by a conjunction of of more than two features (such as "green square"). However, the vision literature does not necessarily contradict contradict the above evidence for the preattentive activation of of shape and of a perceptual object. The two lines of of research research typically use meaning of conspicuously different stimulus materials. On the one hand, the vision literature literature has focused primarily on simple and to a large extent, arbitrary arbitrary graphic stimuli (e.g., colors, lines, simple geometric figures) that have no obvious, unique meanings. On the other hand, the studies attesting to the presence of of preattentive activation of of shape and meaning employ meaningful stimuli that are routinely encountered in everyday life, viz., mostly lexical materials such as words, but occasionally certain complex and realistic graphic graphic materials materials such such as as faces. faces. It goes without saying that some kind of of preexisting processing structures structures such as the ones exemplified in connectionist networks are required for preattentive activation of of shape or meaning to take place (McClelland & Rumelhart, 11981). 98 1). These structures will develop gradually from everyday These structures encounter with relevant stimuli (e.g., LaBerge & Samuels, 11973; 973; Shiffrin & Schneider, 11977) 977) although those those for certain phylogenically significant stimuli 985). Thus, the such as faces may be be hard-wired through through evolution (Field, 11985). extent of of preattentive processing can vary vary from very crude (as in the case of of arbitrary arbitrary and/or meaningless stimuli for which no ready-made processing structure structure is available) to very thorough and sophisticated (as in the case of of meaningfulllexical meaningful/lexical materials for which elaborate processing structures have been established and, thus are readily available). A series of of studies by Shevrin and his colleagues provided some evidence (Shevrin & Fritzler, 1968; 97 1). They found that evoked potential to a Shevrin, Smith, & Fritzler, 11971). subliminally shown picture is significantly more intense if if the the picture is than if if it is meaningless. Further, the the meaningfulness of of the meaningful than
S. S
Kitayama
1199 99
picture systematically influenced subsequent free associations associations as well, suggesting that semantic activation was actually caused by the subliminally shown picture if if it was meaningful. All in all, then, once a meaningful stimulus commonly encountered in daily life such as a word is presented, it will be processed automatically, and daily corresponding perceptual perceptual codes are activated. Although this activation is preconscious preconscious and quite weak especially if the stimulus is impoverished, it has been shown to be sufficient sufficient for affect associated with the stimulus to be O 985). Perhaps, the activated codes summon elicited covertly (Ohman, ( hman, 11985). affective circuits of of the brain located in the limbic or subcortical regions. LeDoux 1 987, 11989) 989) has reviewed neuroanatomical evidence suggesting LeDoux ((1987, numerous neuronal connections between between sensory processing areas of of the brain and the limbic regions. Furthermore, Furthermore, several studies (e.g., Corteen & Wood, Lazarus & McCleary, 11951; 11972; 972; Lazarus 95 1 ; Zajonc, 11962) 962) have shown a reliable autonomic response to a subliminal affective stimulus. Although this literature has been criticized on methodological grounds (e.g., Merikle, 11982; 982; literature Holender, 11986), 986), more recent research with a strict criterion for awareness has also shown that affect can can be elicited by undetectable stimuli (e.g., has Dawson & Schell, 11982; 982; Greenwald 989; Niedenthal, 11990; 990; Tassinary Greenwald et aI al.,., 11989; Tassinary et aI., 984; see also Kunst-Wilson 980). al., 11984; Kunst-Wilson & Zajonc, 11980). Attentive processing. It is reasonable, then, to postulate that affect turn influences subsequent attentive elicited via preattentive processing in tum of processing, which is generally believed necessary for conscious perception of the stimulus (Neely, 11977; 977; Posner & Snyder, 11975). 975). Unlike preattentive processing, attentive processing is selective, limited solely to a perceptual code to which attention has been directed. Thus, once a relevant perceptual activated by an impinging code has been automatically and preconsciously activated operations need be performed. First, attention is shifted and stimulus, two operations directed to the relevant code and, second, once so directed, attention furthers directed the code's processing (cf. Posner, 11980). 980). Through attentive processing, a more more elaborate elaborate perceptual perceptual and, perhaps, semantic image of of the stimulus is developed, which corresponds to the immediate conscious percept of of the stimulus. Finally, the conscious percept may be scanned and its more specific features may be read out out to control subsequent action (Allport, 11989). 989). to grasp the the proposed relationship between preattentive In order to of attention as a spotlight is processing and attentive processing, a metaphor of useful (Crick, 11984; LaBerge, 1983; Moser, 1988; 984; 1 983; 1 988; Posner, 11980). 980). According preattentive processing activates activates the relevant one of of to this metaphor, preattentive numerous perceptual perceptual codes in long term memory. This activation itself,
200
Chapter 5
to produce produce conscious conscious perception. For For the latter to to however, is not not enough to the latter occur, The spotlight must first first be occur, an an attentional attentional spotlight spotlight must must be deployed. The spotlight must shifted to the relevant to illuminate the the code. This to the relevant code and and then used to This illuminating of of the the relevant relevant information information amounts to to additional processing processing illuminating performed on on the latter latter and, as as such, is thought to to enable conscious conscious performed perception of perception of the stimulus. In this this formulation, attentive attentive processing is distinct preattentive processing processing in its selective nature. Whereas from preattentive Whereas preattentive preattentive activation can can occur simultaneously at at multiple loci (e.g., letter, graphic, activation semantic are valid as well as, especially when semantic codes that that are when the the stimulus is impoverished, those that that are invalid; see below), attentive processing can be focused on only one of Another implication of of the current current formulation of them. Another that attention and consciousness consciousness are distinct even though there is is that substantial overlap between them. Generally, preattentive substantial preattentive (nonselective) processing processing takes place without conscious awareness, whereas whereas attentive processing is mostly conscious. However, attentive (selective) (selective) processing processing required to produce conscious awareness is necessarily preconscIous. preconscious. Amplification by a ffect. One widely postulated property of affect. of affect is arousal, or its ability to amplify a variety of of psychological functions. In his pioneering work, Tomkins ((1962, 1 962, 11980) 980) has proposed that various basic can be described in terms of of differential emotions such as joy and anger can of amplification of of a nervous system. Although Tomkins's analysis patterns of of affect has may no longer seem feasible, arousal or an intensity dimension of of everyday vocabularies of of been shown to be essential in defining a variety of emotion and concepts in general, and suggested to be universal across of affect cultures (Osgood, 11962; 962; Russell, 11980). 980). Another major dimension of identified in this literature is pleasantness. From the very beginning, it has been widely recognized that an of affect can can have a variety of of consequences on amplifying property of drive, this assumption is central psychological processes. Under the guise of drive, to a behavioral theory of learning proposed by Hull, Spence, and Taylor in the 1950s (e.g., Spence, 11956). 956). It also is at the core of the Yerkes-Dodson modem extensions by H. law (Yerkes & Dodson, 11908), 908), as well as its modern to analysis of of personality dimensions of of extraversion! extraversion/ Eysenck ((1967) 1 967) to introversion and impulsivity. It has also proved applicable to social (Zajone, 11965). facilitation (Zajonc, 965). More recently, Revelle, Humphreys, and their colleagues (Humphreys & Revelle, 11984; 984; Revelle & Loftus, 1990; see also M. Eysenck, 11976) 976) have elaborated on some specific consequences of arousal on different stages of memory processes.
Kitayama S. Ki tayama
201 20 1
As As has has been been pointed pointed out out by by aa number number of of researchers, researchers, the the notion notion of of generic, uni-dimensional uni-dimensional arousal arousal involving all aspects aspects of of the the sympathetic sympathetic nervous system system and and those those of of cortical processes processes seems too too simplistic simplistic (e.g., Lacey Lacey & & Lacey, Lacey, 1968). 1968) . Nevertheless, Nevertheless, the hypothesis that that affect affect amplifies amplifies aspects of of psychological psychological processes processes remains both both reasonable reasonable (Lindsley, some aspects 1 95 1 ) and and empirically empirically viable viable (Stembach, (Sternbach, 1968). 1 968). And, as such, it has the 1951) potential potential of of clarifying clarifying ways in which which affect affect influences influences cognition. Extrapolating Extrapolating from this this literature, literature, it may may be hypothesized that that affect affect elicited through through preattentive preattentive processing processing amplifies attentive attentive processing. This simple hypothesis which affect with an an hypothesis suggests both both (i) the the conditions conditions in which affect associated associated with impinging word perception of of the the word word itself itself and impinging word is most most likely to to enhance the perception (ii) those in which the the affect affect is most likely to impair the perception. perception. Enhancement and ofperception by affect. First First and and most Enhancement and impairment of by affect. if attention attention has accurately accurately been directed to a relevant perceptual obvious, if perceptual code (i.e., the one corresponding to an impinging word), affect and ensuing amplification of attention enhance the veridical perception of amplification of attention should enhance of the impinging stimulus. In this case, affective stimuli will be more accurately than neutral stimuli (affective enhancement). perceived than Suppose, that a stimulus is presented in an extremely Suppose, however, that impoverished manner, as is often the case in perceptual perceptual identification identification perceptual code will not experiments. Under these conditions, the relevant relevant perceptual receive strong activation. As we have reviewed earlier, this weak activation seems sufficient to produce a degree of of affect, thus amplifying subsequent attentive processing. Nevertheless, the weak activation will cause considerable difficulty in computing exactly which perceptual code corresponds to the impinging stimulus, especially because residual activations past experience are likely to remain for many other irrelevant codes caused by past of this difficulty in 983; Higgins & Bargh, 11987). 987). Because of (e.g., Jacoby, 11983; locating the relevant code, attention may be misdirected to an irrelevant code. locating Under Under these these conditions, affect produced through preattentive processing will amplify attentive processing processing that has been directed, accidentally, to invalid perceptual information and, as a consequence, it will impair an emerging conscious percept. In this case, affective stimuli will be less accurately perceived than neutral stimuli (affective impairment). In terms of of the spotlight metaphor introduced earlier, preattentive of an affective stimulus activates the corresponding code and, as a processing of consequence, evokes associated affect, which in tum turn increases the of the attentional spotlight. However, because the activation of illumination of the relevant code is weak, perhaps no stronger than residual activations
202
Chapter 5
of remaining in irrelevant codes, the spotlight is likely to be locked on to one of the irrelevant codes, accidentally illuminating the latter and, thus, causing an of valid perception: impairment of In sum, the present present model (the amplification model hereafter) predicts that that affective impairment should be most likely to occur when the presentation presentation of of a target target word is extremely impoverished. It further implies that the impairment effect should disappear disappear or even reverse itself once the difficulty in locating the relevant code is alleviated, that is, when the activation of of the relevant code is increased relative to the activation of of other irrelevant codes. Under these conditions, attention will be successfully directed to the relevant perceptual code and, as a consequence, affect and subsequent amplification of of attentive processing processing should enhance the emerging percept. The general prediction tested, therefore, can be be stated in terms of of a ffoctive enhancement (higher accuracy for affective than for neutral stimuli affective within a given experimental condition) or affe ctive impairment (lower affective accuracy for affective than for neutral neutral stimuli within a given experimental condition): Any variable that increases the activation of of a relevant perceptual of other, irrelevant codes will code relative to the activation of increase the likelihood o fa ffective enhancement and/or of affective and~or decrease the likelihood of ffective impairment. of a affective impairment. Evaluation Criteria of of the Amplification Model
Two points must be made explicit before setting out to test test implications Two of the amplification model. First, the model predicts that stimulus affect of of response bias either should influence perceptual accuracy independently of for or against reporting an affective stimulus. Earlier studies in the defense literature were criticized largely because they used recognition and vigilance literature threshold as a dependent variable. With this measure it is extremely difficult to separate perceptual accuracy from response bias (Eriksen, 11963; 963; Goldiamond, 11958). 958). As noted above, however, there are some methods, most that allow notably forced choice between two affectively equivalent stimuli, that one to control for response bias (Natsoulas, 11965). 965). In the following, we will draw primarily on those studies that have adequately controlled for response draw bias. of attentive processing in a preattentively Second, the engagement of perceptual code is only one of of several distinct operations that that can activated perceptual
S. S.
Kitayama
203
(see Figure 1). 1). The model contribute to overall perceptual identification (see therefore assumes that any variable that can enhance attentional engagement may also either improve or impair other operations, and thus either increase of stimulus or decrease the overall perceptual identification independently of
a ffect. affect. To illustrate, consider stimulus complexity, which is likely to have opposing effects on preattentive processing and response selection. To begin of features it with, the more complex a stimulus is, the greater the number of contains. Because each of these features will serve as an additional constraint in preattentive processing, as stimulus complexity increases, the corresponding perceptual code may be more unequivocally activated. According to the amplification model, under these conditions the percept will be more accurate for affective stimuli than for neutral stimuli (affective enhancement). In addition, however, once the percept has been developed, the of the respondent will subsequently have to make an overt response. Most of studies to be reviewed or reported in the current paper examine accuracy in a the respondent supposedly scans scans and and compares the forced choice, whereby the percept with available alternatives. Because complex stimuli contain more features to be compared and matched in the choice, stimulus complexity should make response selection more difficult, thereby leading to poorer overall performance. In short, stimulus complexity is likely to increase the likelihood of of affective enhancement, while while simultaneously decreasing choice performance. Once these two effects of of stimulus complexity are super-imposed on on each performance for affectively neutral neutral stimuli to decline other, one will observe performance with stimulus complexity. Relative Relative to this base base line defined by the the neutral stimuli, performance performance for comparable comparable affective stimuli should improve. Yet, this improvement due to stimulus affect may or may not compensate compensate for the decline of of overall performance due to choice difficulty. This means that performance for affective stimuli may or may not actually improve with The crucial prediction of of the amplification model in this stimulus complexity. The case then, then, is that decline of of performance performance as aa function function of of stimulus case that aa decline complexity is less for than for neutral stimuli. complexity for affective stimuli than In general, general, it is safe assume that any variable variable that can enhance safe to to assume that any that can attentional engagement (e.g., stimulus complexity) may attentional may also either improve or impair other operations impair other operations (e.g., response selection), and and thus thus either increase or or decrease decrease overall overall performance performance independently of of stimulus affect. Accordingly, the amplification model must the must be be evaluated evaluated in terms of of its ability to to predict either affective enhancement enhancement or or impairment, impairment, rather rather than than its ability to to predict either
204
Chapter 5
of perf performance affective an absolute increase or decrease of onnance for aff ective or neutral stimuli with the manipulated variable. With these evaluation criteria at hand, we now tum turn to a review of of extant extant studies pertinent to some predictions of of the amplification model.
Empirical evidence pnmmg has amply Expectation and word requency. Research on priming word f frequency. demonstrated that that when individuals are led to expect the identity of of a target,
& Posner & Snyder, 11975). 11987; 987; Neely, 11977; 977; Posner 975). It can be hypothesized,
the corresponding perceptual code receives extra activation (e.g., Higgins Bargh,
theref ore, that therefore, that in a perceptual perceptual identification task, the difficulty of of locating a
relevant perceptual code is relieved by a correct expectation about the target affective affective word. Thus, aff ective enhancement will be more likely and aff ective impainnent impairment less likely in the presence of of a valid expectation than in its absence (Kitayama,
11990). 990). Initial support for the prediction was uncovered in
a review of ense and vigilance. Because most of the literature of of perceptual def defense studies in this literature examined recognition threshold and f ailed to control failed f or response bias, their status as evidence f or the current analysis is uncertain for for (Eriksen,
11963; 963; Goldiamond, 11958; 958; see e.g., Dixon, 11980; 980; Erdelyi, 11974, 974, ffor or
two experiments which manipulated expectation reviews). Nevertheless, two 11954; 954; Lacey, Lewinger, & Adamson, 11953). 953). In these studies, when there was no expectation,
tended to support the amplification model (Freeman,
for affective than ffor recognition threshold was higher f or aff ective words than or neutral words (affective impairment); but when an expectation about the identity of of a target (aff ective impainnent); for affective that was provided, recognition threshold f or aff ective words was lower than that for (affective al.,, 11953, 953, ffor or a for neutral ones (aff ective enhancement; see also Postman et aI. similar result). Additional evidence f or the present analysis can be f ound in more recent, for found methodologically more sophisticated studies that assess perceptual accuracy of response bias (either readiness or reluctance to report independently of 1 990) located nine such affective rather than neutral stimuli). Kitayama ((1990) of these studies (Bootzin & Natsoulas, 11965; studies (see Table 1). In most of 965; Dorfman, 1967; 1967; Dorfinan, Dorfman, Grossberg, & Broadbent & Gregory, 11967; 967; Dorfinan, Kroeker, 11965), the dependent variable was correct response rate, with 965), the for appropriate adjustments made f or response bias. Two additional studies used different 1972) diff erent methods to minimize response bias. Chapman and Feather ((1972) examined the ability to detect (rather than identify) a novel graphic stimulus using a signal detection procedure. They assigned an affective tone to the using
S.
Kitayama
205
stimulus by classically conditioning an electric shock to it. Sales and Haber Haber ((1968) 1 968) minimized response bias by having subjects report ers of report individual lett letters of word rather than than the word word itself, and and analyzed the number number of of letters a flashed word correctly reported.
Table Table 11.. Comparison of of nine experiments experiments in which effects of of response bias were 990). were minimized (adapted (adapted from Kitayama, 11990). Experiment
Expectation Expectation
Outcome Outcome
Chapman 1 972) Chapman & & Feather Feather ((1972) Dorf man ((1967) 1 967) Dorfman Dorfman et al. ( 1965) 1 965) Dorfman Wertheimer ((1958) 1 958) Mathews Mathews & & Wertheimer Minard 1 965) Minard ((1965) Van Egeren ((1968) 1 968) Bootzin & 1 965) Bootzin & Natsoulas Natsoulas ((1965) Broadbent 1 967) . Broadbent & & Gregory Gregory((1967) Sales & 1 968) Sales & Harber Harber ((1968)
Clear Clear Clear Vague Vague Vague Vague Vague Absent Absent Absent
Enhancement Enhancement Enhancement Enhancement Enhancement Enhancement Inconsistent11 Inconsistent Inconsistent22 Inconsistent
effect No effect Impairment Impairment Impairment Impairment Impairment Impairment
11 Significant Significant impairment impairment effect effect was found found for "high-hysteria" "high-hysteria" subjects, subjects, but no effect effect was obtained for "high-psychasthenia" "high-psychasthenia"subjects. subjects. obtained 2 ound ffor or males, 2 Significant Significant impairment impairment effect effect was ffound males, whereas whereas significant significant enhancement enhancement effect effect was found found for females. females.
Among these these nine studies, three obtained affective enhancement Among (Chapman & Heather, 11972; Dorfman, 11967; Dorfman et aI., al., 11965). 972; Dorfinan, 967; Dorfinan 965). (Chapman Interestingly, all the three studies inadvertently used a procedure that assured that that the subjects had a a clear expectation expectation about about the target stimulus. In two experiments by Dorfinan Dorfman subjects were shown the target target word experiments word plus a nontarget word word immediately before the the target was actually flashed. They were nontarget that one of of the pre-target words would be flashed on that that trial. Chapman told that and 1 972) had and Feather Feather ((1972) had subjects keep in mind the target target stimulus while seeing a visual display. Thus, both both methods provided subjects with a clear a other experiments implanted subjects with vague expectation. Some other expectations by familiarizing them with experimental stimuli at at the beginning of the session. In these studies there was no systematic pattern. Mathews and of Mathews Minard ((1965) of affect to depend Wertheimer ((1958) 1 958) and Minard 1 965) found the influence of
206
Chapter 55
Van Egeren (1968) failed to on certain individual difference variables, and Van of affect on perceptual accuracy. The remaining did not find any influence of of the above procedures and reported affective impairment (Bootzin use any of & Natsoulas, 1965; Broadbent & Gregory, 1967; Sales & Haber, 1968). Kitayama (1990) subsequently conducted an experiment in which expectation expectation was systematically manipulated. In this experiment word frequency was also varied. It was hypothesized that as word frequency increased, the valid perceptual code would be more strongly activated and, as a consequence, affective impairment would become less likely and affective of a target enhancement more likely. Subjects were exposed to a 25 ms flash of word. They then chose the target word from a word pair. In half the trials, this word pair was given before the flash to create an expectation. Further, on some trials no target target was presented although subjects were led to believe that it was actually shown. Analysis of of the data from these trials revealed no response bias for or against reporting affective stimuli, so choice hit rate was used as a measure of of perceptual accuracy. Consistent with the amplification model, both expectation and word frequency increased the likelihood of of affective enhancement and decreased the likelihood of of affective impairment. As can 0-50 can be seen in Figure 2, when words were low in frequency (1 (10-50 occurrences per per million) and and an expectation was absent, affective words were identified significantly less accurately than neutral ones (affective impairment). This pattern, however, was reversed to show a reliable enhancement effect when high-frequency words (more than 100 occurrences per million) were examined and and an expectation was present. Finally, the of affect in the remammg remaining two conditions (high influence of frequency/unexpected and low frequency/expected) was no greater greater than than that (1990) 990) study thus generally in the former two conditions. The Kitayama (1 confirmed the predictions of of the amplification model. Nevertheless, it was not totally conclusive. First, it tested only a small number of of words (12 in total). Second, it found the predicted effect of of expectation only for high-frequency words. There was no such effect for low-frequency words: as can be seen in Figure 2, affective affeetive impairment of of evidently equal strength was observed regardless of of expectation. Stimulus contrast. The The amplification model states that affective impairment is most likely when the presentation of of a target is extremely impoverished. Another recent set of of experiments with a larger number of of stimulus words ((126 126 in total, ranging from 8 to 65 per million in frequency of of occurrence) has provided support (Kitayama, 1991). In Study 11,, a target target was
S.
207
Kitayama
0.68 0.66
"
0.64
--
['7
Affeetivq
1
Neutral
0.62 0.60
--
0.58 0.56 Im qD
0.54 0.52 0.50
i
.......
I
i
..0
Word Frequency/Expectation
of target affectivity, Figure 2. Perceptual accuracy (hit rate) as a function of expectation (adapted (adapted from Kitayama, 11990). word frequency, and expectation 990). = .72 ft-L) if-L) in a black background (.67 ft ilpresented in dark gray (luminance = L), so the contrast between the target and the background was extremely low. As predicted, a significant affective impairment effect was obtained - that is, of affective words was less accurate than the identification the identification of of neutral words. In Study 2, however, the contrast contrast was increased so that that the of if-L). Under the latter condition, there target was shown in lighter gray (.75 ft-L). was no influence of of affect. Exposure time. time. According to the amplification model, a relevant perceptual code needs to be located quite early in the processing bef before ore attention is directed. It then follows that effects of of exposure times should range in which they are manipulated. When relatively depend crucially on the range long exposure times are manipulated, it will be only late in the processing that of the relevant code. Thus, these variations begin to increase the activation of of extremely impoverished stimulus contrast, attention under the conditions of should be misdirected to an irrelevant code regardless of of the exposure times. of Kitayama ((1991) In Study 11 of 1 99 1 ) described above, three relatively long exposure times ((100, 1 00, 150, 1 50, and 200 ms) were tested. As predicted, affective of evidently equal magnitude was observed in all the three impairment of
208
Chapter 5
conditions. By contrast, when relatively short exposure times exposure time conditions. are tested, an increase in exposure times should increase the activation of of a relevant code early in processing, thus mitigating the difficulty in locating the code. Thus, an increase in relatively short exposure times should result in a lesser likelihood of of affective impainnent impairment and a greater likelihood of of affective enhancement. Kitayama ((1989) 1989) showed that affective impainnent impairment observed with a 25 ms exposure (Kitayama, · 11990) 990) could disappear with a 40 ms exposure. Word length. Virtually every past past study in this area area has examined examinexl only relatively short words (less than 6 letters long). However, current cognitive models of 9 8 1 ) suggest of word recognition (e.g., McClelland McCleUand & Rumelhart, 11981) that that word length may systematically change the likelihood of of affective affectivr enhancement and impainnent. impairment. According to these models, the initial, preattentive preattentivr processing of of a visual input proceeds process in parallel, leading to simultaneous activation of of parts of of the entire input. Currently, there is no what defines functional parts of of a word. Drawing on consensus about exactly what some prominent models of of word recognition (e.g., McClelland McCleUand & Rumelhart, 11981), 98 1 ), we assume here that "word-parts" correspond fairly closely to individual letters, although, for the purposes of of the present argument, however, it makes little difference whether the units are letters or something else. turn impose Once individual letters have been activated, they in tum of the input, pennitting permitting only a significant constraints on the likely identity of limited number of of English words as reasonable candidates for the input. All else being equal, as word length increases, a greater number number of of letters should be activated and the letter-level infonnation information should more strongly constrain of a four fourthe word-level identity. To illustrate, imagine that the processing of half of of the constituent letters, say, "LxxE." letter word successfully activated half There are several 4-letter candidate words that meet these constraints, say, if half of of the "LIVE," "LIKE," "LOVE," "LAKE," and so on. In contrast, if constituent letters are activated in a word that is 110 0 letters long, say, "AxTRAxxlxx," there will be very few 110-letter "AxTRAxxIxx," O-letter words other than "ATTRACTIVE" "ATTRACTIVE" that that fully meet the constraints. Thus, as word length increases, the perceptual code corresponding to an impinging word will be more unequivocally and uniquely activated and attention will be more likely to be directed to the valid perceptual code. As word length increases, there of affective enhancement as opposed to affective should be a greater chance of impainnent. impairment.
S.
Kitayama Ki tayama
209
In a recent experiment, both word affect and word length were systematically manipulated (Study 11 in Kitayama, 11991). 991). Words comparable to Kitayama's ((1990) 1990) low-frequency words were presented with an extremely contrast and no expectation expectation was provided - a condition diminished stimulus contrast the amplification model predicts to be highly conducive to affective impairment. A reliable affective impairment effect was observed. However, this impairment effect was also observed for longer words, thus failing to confirm the model's prediction. Perhaps, with the extremely diminished stimulus contrast contrast examined examined in this study, there was only marginal activation of a relevant perceptual code regardless of of word length. It would seem of reasonable that that an increase increase in word length could contribute to the unique and of a relevant perceptual code only when there was unequivocal activation of enough stimulus input. In the experiments to be reported below, therefore, presented with greater stimulus contrast. The perceptibility of of targets were presented the target target was then reduced by presenting a masking stimulus immediately after the disappearance disappearance of of the target. Under these conditions of of backward pattern masking, an increase in word length was predicted to decrease the likelihood of of affective impairment and to increase the likelihood of of affective enhancement. Valence o f aff ect. One potential divergence between the of affect. defense/vigilance hypothesis and the amplification model concerns the effect of of the valence (positive or negative) of of affect. Unlike the amplification model, the defense/vigilance hypothesis has never been explicit enough to advance clear-cut predictions for affective enhancement and impairment. Yet, it would seem to predict that that the processing is either prohibited (the defense) or enhanced (the vigilance) if if and only if if "anxiety" (or, equivalently, enhanced "psychodynamic conflict") is evoked. Since "anxiety" is more closely linked with negative than positive affect, the perceptual influence of of affect should be obtained primarily with negative affective words. In contrast, the obtained amplification model is non-committal in this regard. It is possible that of an attention is amplified once the significance or the interest value of perceptual impinging stimulus has been detected. If this is the case, the perceptual not depend on the valence (positive or negative) of influence of of affect need not of the affect; for the significance or interest value can be signalled by any any affect either positive or negative, that examined accuracy accuracy in perception In virtually all the past studies that of response bias for or against affective stimuli, only taboo independently of Kitayama, 11990, (mostly affectively negative) words were used (see Kitayama, 990, for a The exclusive use of of taboo taboo words words was justified justified on the supposition review). The
210 210
Chapter 5
that Freudian Freudian defense defense should should mediate mediate the the perceptual perceptual influence influence of of affect affect (e.g., (e.g., that Blum, 1 954; Erdelyi, Erdelyi, 1974; 1 974; McGinnies, McGinnies, 1949), 1 949), but but ironically made made it Blum, 1954, to test test the the supposition supposition itself. itself. A A few few studies, studies, however, however, examined examined impossible to impossible both positive positive and and negative negative affective affective words words while while controlling controlling for for response response bias. both It was was found found that that the the perceptual perceptual influence of of affect affect was was mostly identical identical It whether affect affect was was positive or or negative negative (Broadbent (Broadbent & & Gregory, 1968; 1 968; whether Kitayama, 1990, 1991). This evidence is consistent with the amplification Kitayama, 1 990, 1 991). This with but raises raises some doubt doubt on on the the defense/vigilance hypothesis. model, but
The present experiments attempt to to further of the model, In an attempt further test test the implications of the amplification model, for the two experiments were two were conducted. As noted above, initial support for of past studies that model was obtained obtained in a review of that differed in the extent to to which aa valid expectation was available available to to subjects. It was hypothesized that a valid expectation should activate the relevant code prior to the presentation presentation of a target target stimulus, thus alleviating the difficulty in locating of locating the code in identification. The expectation, therefore, should increase the perceptual identification. likelihood of of affective enhancement and decrease the likelihood likelihood of of affective manipulated impairment. So far, however, only a few studies have actually manipulated expectation (Fr (Freeman, ct al., 1953; Kitayama, 11990; ct expectation eeman, 11954; 954; Laccy Lacey et 990; Postman et of the 953). Although these studies supported the predictions of al., 11953). Freeman Laccy et ct al. amplification model, they were not conclusive. Fr eeman ((1954), 1954), Lacey (1953), and Postman et al. ((1953) 1953) measured recognition threshold, so for their findings. fmdings. Although response response bias response bias may in part account f or their of was controlled in the Kitayama ((1990) 1990) experiment, only a small number of of the ffmdings. words ((12 1 2 in total) were tested, leaving open the generality of mdings. of expectation was further examined in the present Thus, the effects of experiments. to current models of of Another variable tested was word length. According to (e.g.,, McClelland & Rumelhart, Rumclhart, 198 1981), an increase in word word processing (e.g. 1), an length should impose impose more constraints on the the identity identity of of the the word word and, and, thus, thus, length unequivocal activation activation of of the the relevant relevant perceptual perceptual code. code. Thus, Thus, conduce to unequivocal affective affcctive enhancement should should be be more likely likely and and impairment impairment less less likely likely with an increase of 1991) studied of word word length. Only Kitayama ((1991) studied the the effect effect of of this variable, and failed to to find find any evidence. To To test test the the conjecture conjecture that that this variable, failure was was due due to to the the highly highly degraded degraded input, input, the the current current series series of of experiments experiments failure employed aa pattern-masking pattern-masking procedure, procedure, whereby whereby aa target target stimulus stimulus was was employed presented with aa relatively relatively high high stimulus stimulus contrast, contrast, but but was was immediately immediately presented
S.
Kitayama
2211 11
word frequency was held followed by a pattern mask. In these experiments word constant at the level comparable to the low-frequency condition of the 1 990) experiment (the level also examined in the Kitayama ((1991) 1 99 1 ) Kitayama ((1990) study) to maximize the comparability across different studies. Finally, both positive and negative affective words were tested to determine whether the of affect depends on its valence. perceptual influence of Predictions. We predicted that both expectation and word length would of affective enhancement and decrease the likelihood of increase the likelihood of affective impairment. Thus, our first two predictions were: ((1) 1 ) Affective impairment will be most likely when short words are used and no expectation is provided. (2) Affective enhancement will be most likely when long words are tested and an expectation is provided.
It was not certain exactly how expectation and word length would jointly operate. According to the amplification model, in order for these variables to of affective enhancement or have additive impacts on the likelihood of impairment, two conditions must be met. First, expectation and word length must additively increase the activation of of a target perceptual code relative to of other irrelevant codes. Second, the relative increase of of the the activation of activation of of the target code must linearly increase the likelihood of of affective enhancement (or decrease the likelihood of of affective impairment). Neither enhancement assumption has been explicitly tested in the literature. Thus, no a priori priori prediction could be made made regarding whether the two variables would interact or have additive effects. Thus, our third prediction was:
The influence of two conditions (i.e., short (3) The of affect in the remaining two words/expected, long words/unexpected) will fall somewhere between the above two extremes (short words/unexpected, and long words/expected); in other of affect in the former conditions greater other words, the influence of conditions will be no greater than than that that in the latter. In addition to to improve attentional engagement to their hypothesized role to and thereby to to increase the likelihood of of affective enhancement and to reduce and that of of affective impairment, there are some suggestions in the literature that that that and word length may have some extraneous effects on forced both expectation and choice performance. To demonstrated that To begin with, it has been demonstrated that explicit
212 212
Chapter 5
formulation formulation of of aa "hypothesis" "hypothesis" or or expectation expectation can can impair impair perceptual perceptual identification identification (e.g., Bruner Bruner & & Potter, Potter, 1964; 1 964; Lawrence Lawrence & & Coles, 1954). 1 954). Perhaps, Perhaps, aa clear clear expectation expectation may may direct direct one's attention attention to to only one one type type of of information that that is potentially potentially available available (e.g., orthographic, orthographic, phonemic, information perceptual identification). identification). Information on the ignored semantic, etc., in perceptual ignored may then then be be unavailable unavailable in response selection, leading poorer dimensions may leading to poorer performance Note that performance in the the presence presence of of an expectation than than in its absence. Note that this restrictive restrictive effect effect of of expectation expectation should occur occur equally regardless regardless of of this stimulus stimulus affect. affect. Similarly, word word length is also likely to to depreciate depreciate overall performance independently of of stimulus stimulus affect. In aa forced choice task performance task tested in present research, research, the the respondent respondent scans the percept and compares the present the percept compares it with alternatives. The The longer the word, the greater greater the number of of features features available alternatives. (e.g., individual letters) letters) that that must must be be matched matched and, therefore, the more difficult response selection should be. Furthermore, as we shall show below difficult 2 1 9), this difficulty in response selection for longer (p. 219), longer words words may be exacerbated by the fact that of long words exacerbated by that any any given pair of words tend tend to share aa greater a pair pair of greater number of of common letters than a of short words. All in all, as word length increases, response word response selection will be more difficult and, further, effect of of word length on response response selection will occur of word this effect occur regardless regardless of word affect. In sum, we hypothesized that affect. that both both expectation and word word length length would would performance in perceptual depreciate overall overall performance perceptual identification, while simultaneously of affective simultaneously increasing the the likelihood of affeetive enhancement and decrease that ective impairment. Taken together, we predicted a general that of of aff affeetive decline of of perceptual perceptual identification with expectation and word length, and further expected this decline of of performance to be significantly less ffor or aff ective words than affective than for neutral words. Notice that that this this latter prediction prediction amounts amounts to the three predictions stated stated above. Experiment Experiment 11
Method Overview and subjects. There were 128 trials, divided into two blocks, diff ering in the length of differing of the words words (long (long versus short). The order of of the two blocks was counter-balanced over subjects. On each trial subjects were ms flash of of either an affectively positive, negative, or neutral exposed to a 33 ms target word, immediately followed by a pattern mask (a string of "&"s of of the target word and same length as the word). They were then presented with the target of the same length, and asked to choose the an equivalently valenced word of
S. S.
Kitayama
2213 13
one presented. Whereas word length was a within-subject variable, expectation was a between-subject variable. Thus, for half the subjects the two choices were presented immediately before the target was flashed to create an expectation, but for the other half there was no such expectation. undergraduates at at the University of of Oregon (both males and and females) Seventy undergraduates participated in the experiment to partially fulfill their introductory psychology course requirements. All the subjects claimed to be a native English speaker. Materials. The one-hundred twenty-eight words used in the present experiment are listed in Appendix A. There were an approximately equal of affectively positive, negative, and neutral words. Thirty three number of undergraduates who did not participate in the present experiment were asked of each word ((11 = = unpleasant, 5 = = pleasant). The to judge the affective quality of = 4.3, s = = .23) than the neutral positive words were rated as more positive (M = . 1 , s = .2 1 ), which in tum words (M = 33.1, .21), turn were rated to be more positive than half of of the words of of each affect the negative words (M = 11.8, . 8, s = ..30). 30). About half type were long (more than nine letters long) and half half were short (fewer than of occurrence ranged from 8 to 65 appearances appearances six letters long). Frequency of per million words, as determined by Kucera and Francis's ((1967) 1 967) norms. The of occurrence for the six word categories (3 affective types mean frequencies of • 2 length types) were practically identical (varying from 27 to 37 x occurrences per million). This frequency range roughly corresponded to to the of the Kitayama ((1990) low-frequency condition of 1 990) experiment. Sixty-four pairs were formed between with equivalent valences and lengths, as shown were between words words with nontarget once, Each word served as target target once and as nontarget in Appendix A. Each resulting in 128 experimental trials. The order of of these these trials were randomized randomized upperwithin each block for each each subject. All stimulus words words were shown in upper case letters. The experiment was Equipment. The was controlled by an AMDEK-286 AMDEK-286 personal computer presented on an an computer with an an AMDEK-132 AMDEK-1 32 VGA VGA adapter. Stimuli were presented AMDEK-732 and contrast AMDEK-732 color color graphic graphic monitor. Both the brightness and contrast of of the screen were kept maximal. To To the monitor was attached a translucent translucent tube. The inner the tube 23.5 cm, and and the the length, 60 60 cm. The inner contour contour of of the tube was was 17 1 7 cm •x 23.5 One the tube tube was the monitor, and subjects One end end of of the was fitted fitted to to the monitor, and subjects watched watched the the screen screen through the other end. Target presented at through the other Target words words were were presented at the the center center of of the the monitor. monitor. The The height of of the the words was was approximately approximately 4 mm. The The length of of aa five-letter word word was was 12 1 2 mm, and and that that of of aa nine-letter word word was was 23 mm, mm, resulting angle of the short resulting in the the visual visual angle of approximately approximately 1.15 1 . 15°~ and and 2.06 2.06°~ for the short and the and the long words, words, respectively. The The experiment experiment was was controlled controlled by by the the MEL MEL (Micro Experimental Experimental Laboratory) Laboratory) system developed by by Schneider (1988). ( 1 988). =
=
=
=
2 14 214
Chapter 5
This system synchronizes the the presentation of of stimuli with the refresh cycle of of This the monitor, enabling accurate accurate control control over the exposure durations. The experiment experiment was conducted under under normal room illumination. With the VGA facility, it is possible possible to create 62 shades of of gray in addition to white and background (luminance = .67 ft-L [foot-Lambert]), a target black. In a black background was shown in the 23rd shade"of 1 .02 ft-L). This level of shade"of gray ((1.02 of stimulus contrast was considerably higher than the level employed in the Kitayama ((1991) 1 99 1) study ((15th 1 5th shade of of gray; .72 ft-L). It-L). A fixation point, a mask, an expectation, and response response choices were shown in the 32nd shade of of gray i~-L). (2.65 ft-L). Procedure. Subjects were tested individually. Upon arrival, they were randomly assigned to one of of the four conditions representing representing the presence or the absence of of an expectation expectation and the order of of the short-word and the long longword blocks. They were instructed instructed to look into the tube, and to place their left and right fight index fingers respectively respectively on the Z and the M keys of of the computer keyboard. The subjects were told that that the experiment was concerned with perception of of briefly shown words. After the procedure was described described (see perception below), they were given the following instruction: liAs "As I mentioned to you, words are are presented very briefly. We want to know how accurately accurately people can recognize a word under such impoverished r ~ u r e is set up so that you cannot perfectly viewing conditions. So, the pprocedure of the word. see the word, yet you can still recognize some fragments or parts of may be able able to recognize a letter or two, or even a a small For instance, you may part of of some letter. Or you may may be able to recognize recognize the contour of of the word. part partial information has has proved very useful in performing this task. I will Such partial explain to you exactly what can be learned le~meA from this sort of of experiments later. explain can For the time being, even though though you you might occasionally feel that that you are For merely guessing, don't be discouraged discouraged or disturbed by this. Instead, try to pick up as many physical cues from the flash as possible. In this way your responses will be most accurate. Even when you don't think that that you have enough information to make a choice, give us your very best guess. Please never use any intuition or gut-feeling in making judgment; once you do this, never performance better than chance. From past research research we know rmance cannot be better your perfo that that it is essential that that you try to pick up physical features such as letter if you are to perform this task at a better-than better-thansegments and overall contour if chance level. level."II
S.
Kitayama
2215 15
In the expectation-present condition, each trial started with the fight of the center of the screen presentation of a pair of words with one to the right see, the pair was replaced by a fixation point and the other to the left. After 2 sec, center of of the screen. The subjects in this condition condition had been told that at the center of the two words would be a target in the upcoming trial and that it was Was one of important to keep the pair in mind in order to perform well in the task. In the expectation-absent condition the trial began with the presentation of of a fixation expectation-absent expectation conditions were otherwise identical. Thus, in both point. The two expectation conditions, when the subject simultaneously pressed the two response keys, point disappeared disappeared and, 200 ms later, a target word was presented the fixation point for 33 ms, immediately followed by a pattern mask. A sequence of "&"s of served as a mask. The mask was presented the same length as the target word serve<;l for 11500 ms. Immediately after the disappearance disappearance of the mask, the 500 corresponding corresponding word pair pair (the one shown at the beginning of of the trial in the expectation-present condition) was shown. The subjects had been instructed instructed expectation-present to press either the Z key (with the left index finger) if if the word on the left side of the fixation point had had been presented, or the M key (with the right index of finger) if the word on the right side had been presented. Response time was measured from the onset of of the choice pair in milliseconds. The next trial measured automatically started started after an interval whose length varied randomly between automatically see with the average average of of 4.5 sec. At the completion of of the first block, 2 and 7 sec the subjects subjects were given a short short break break prior to the second second block. The The the experimental the practice experimental trials trials were were preceded preceded by 26 practice practice trials. trials. On the practice trials trials both affective not used the experimental experimental trials trials were both affective and and neutral neutral words words not used in the presented. presented. On On the first first practice practice trial the target target word was shown for 200 ms. On each On each of of the the subsequent subsequent practice practice trials, trials, the exposure exposure time was gradually gradually reduced that by the reduced so that the 21st 2 1 st trial, it was was set at at the exposure level used used in the the experimental experimental trials, trials, i.e., 33 ms.
Results and and discussion Two were examined. the percentage Two dependent dependent variables variables were examined. First, First, the percentage of of correct correct choices choices (hit (hit rate rate or or the the accuracy accuracy score) was was computed computed for each each condition. condition. In the the present present procedure procedure the the subjects subjects chose between between two two affectively affectively equivalent equivalent stimuli. stimuli. Hence, Hence, there there was was no no room room for response bias for or or against against affective affective stimuli safely taken an stimuli to to come come into play, and and the the accuracy accuracy score can can be safely taken as an unbiased unbiased index index of of the the perceptibility perceptibility of of the the target target (Natsoulas, (Natsoulas, 1965). 1 965). Second, the the time time required required to to make make aa choice choice was was analyzed. Because Because few existing existing studies studies in this this area area reported reported response response time, it was was not not clear clear whether whether the the
2 16 216
Chapter 5
perceptibility under the present conditions could be reliably and adequately indexed by response time. It was considered reasonable, however, that that as the perceptibility of of aa target word word increased and the accuracy score increased,
response time would decrease. Accuracy. Two separate analyses were perf ormed on the accuracy Accuracy. pefformexl scores. First, mean accuracy scores were calculated separately for each subject and then submitted to' to. an Analysis of of Variance (ANOVA) with subjects as a a random variable. A significant F F statistic obtained in this analysis (designated as FI) F1) indicates the generalizability gcneralizability of of the effect across subjects. Second, mean accuracy scores were computed separately f or each for word pair and then thon submitted submitted to an ANOVA with word pairs as a random variable. A significant F F statistic from this analysis (designated as F2) 1 973) has indicates the of the effect across word pairs. Clark ((1973) the generalizability of recommended the use of mmF', which indicates the extent to which the effect of minF', can be generalized generalizeA simultaneously over both subjects and word pairs. A minF' mmF' can be computed from from the corresponding FI F 1 and F2 F 2 according to the formula, FI .F'21(FI+F2). Some have argued, however, that this statistic is F 1,,F2/(FI+F2). too conservative, prone to Type II errors (Wike & Church, 1976); both F F I1 and F2 F 2 will have to be quite large bef before mmF' can and ore the corresponding minF' present paper, thus, F F I1 and F F 2 are approach statistical significance. In the present used to gauge the reliability of ects. MinF' will be reported only when it of eff effects. attains statistical significance. A preliminary analysis involving two between-subject variables (expectation and the order of of long- versus short-word blocks) and two within withinof block subject variables (word affect and word length) showed no effect of dropped in subsequent analysis. The ). This variable, thus, was dropped order (Fs < I1). of Table 22.. Note, relevant mean accuracy scores are given in the first half of for first, that in all the conditions the accuracies f or positive and negative affective words were virtually identical, lending support to the prediction that of stimulus aff affect the influence of ect on immediate conscious perception should not of the affect. Consistent with the amplification model, depend on the valence of yet somewhat contrary to theories based on the notion of of Freudian defense (e.g., Blum, 11954; 954; Erdelyi, 11974; 974; McGinnies, 11949), 949), this apparent irrelevance of of the valence (positive or negative) indicates that attention was of an impinging stimulus amplified once the significance or the interest value of had been detected. Because the effect of of the valence was negligible, the two had valence categories were subsequently combined.
S.
2217 17
Kitayama Ki tayama
Table 2. Means (and standard standard deviations) of of accuracy and response time in Table perceptual perceptual identification as a a function function of of word affect, word length, and expectation (Experiment 1). Expectation Expectation Absent Absent (N=36 (N=36))
N=34) Present Present ((N=34)
Word
Word Word Affect Affect
Length
Positive Positive
Negative Negative
Neutral Neutral
Positive Positive
Negative Negative
Neutral Neutral
Accura'Y Accuracy
Long Long Words
.63 .63 (.11) 1I) (.
.63 .63 (.14) (. 1 4)
.57 (.12) 12) (.
.60 ((.13) . 1 3)
..57 57 (.14) (. 14)
.59 .09) ((.09)
Short Words
.61 .61 (.16 (.16))
.61 .61 (.14) (. 14)
.67 10) (.10) (.
.63 .63 (.15) 1 5) (.
..57 57 (.15) (. 1 5)
.67 (.15) (. 15)
Response ms) Response Time Time ((ms) Long Long Words
827 6 3) (363) (3
799 (326)
890 (431) (43 1)
1760 1760 (571) 1) (57
11792 792 (740)
1684 1684 (571) (57 1)
Short Words
762 (397)
862 (458)
719 (402)) (402
1651 1651 (481)) (481
1645 1645 (490) 90) (4
11571 571 (431) (43 1)
A 2 xx 2 xx 2 (expectation, word affect, an andd word length) ANOVA ANOVA showed a significant main effect for word length (F (/71(1,68)=22.5 , p<.001, showed (1 ,68)=22.5, p<.OO l, l ( 1 ,60)= 19.2, p<.OO l , and minF'( 1 , 125)=lO.4, p<.OO l). This effect, F2(1,60)=19.2, p<.001, minF'(1,125)=lO.4, p<.001). F 2 however, was qualified by the interaction between word affect and word and p<.005, length (F p<.OO l, F (F1(1,68)=12.9, p<.001, F2(1,60)=11.1, p<.005, and l ( 1 ,68)=12.9, 2( 1 ,60)= 1 1 . l , minF'(1,126)=5.97, p<.05) as well as by the interaction among word affect, minF'( 1 , 1 26)=5 .97, p<.05) word length, and expectation (F (F1(1,68)=4.04 p<.05, and F F2(1,60)=5.70 l (1 ,68)=4.04,, p<.05, 2( 1,60)=5 .70,, p=.02). Thus, as predicted by the amplification model, the influence of of affect p=.02). on perceptual accuracy depended on both expectation and word length. The relevant means are shown in Figure 33.. Based on the amplification model, we hypothesized that both expectation accurate engagement engagement of of and an increase in word length would improve the accurate processing in the the relevant perceptual code. From this hypothesis, attentive processing were advanced. First, affective enhancement should should be most three predictions were
2218 18
Chapter 5
~176 i 0.68
['-]
Affective
m
0.66 0.64 0.62 "~
0.60
gh
0.~8 0.56 0.$4
/ O'' / / Word Length/Expectation Figure 33.. Perceptual sensitivity (hit rate) as a function of of word affect, word length, and expectation (Experiment I). 1).
likely when both of of these two factors are present. In support of of this prediction, accuracy was higher for affective words than for neutral words, (t 1 , t2(1 13)=2.98, p<.01, minF'(246)=4.37, p<.05), p<.05), (t1(136)=2.95 p<.01, t2(113)=2.98, p<.01, and minF'(246)=4.37, l( 136)=2.95,, p<.0 when (i) the words were long and (ii) an expectation was present. Second, it was predicted preAictexl that affective impairment should be most likely when (i) the words were short and (ii) an expectation was absent. Indeed, in this condition accuracy was lower for affective words than for neutral words (t} 136)=11.93, .93, (q ((136)= p<.06, 1 13)=2.02, p<.05). p<.06, and t2( t2(113)=2.02, p<.05). Finally, the current analysis also predicted that the influence of of affect in the remaining two conditions should be no greater than in the above two conditions. Consistent with this prediction, there was no influence of of affect when (i) words were long and (ii) an expectation was absent, ttss < 11.. Further, there was an affective affoctive impairment effect when (i) 1 36)=2.6 1 , p<. Ol, words were short short and and (ii) an expectation was present (t}( (t1(136)=2.61, p<.01, and 12(1 13)=2.80, p<.O I); but the magnitude of t2(113)=2.80, p<.01), of this effect was virtually identical to the comparable effect in the short/expectation-absent condition, ttss < 11.. When comparisons were made across the experimental conditions
S.
Kitayama
2219 19
separately for affective and neutral words, three pairs were reliable: the expectation present versus absent conditions for long, affective words of (p<.06), and the long versus short neutral words at both of the levels of expectation I). expectation (ps<.O (ps<. 01). We We also hypothesized that an increase in word length would lead to a greater difficulty in response response selection in a forced choice task. Further, we that this increase of of difficulty in response selection due to word reasoned that length should be exacerbated exacerbated by the fact that that a pair of of any arbitrarily chosen words tend to share a greater number of of letters as they become longer. To check this possibility, we counted the number of of letters shared by the words OVA showed that ANOVA that regardless of of word affect, the long in each pair. An AN word pairs shared a considerably greater (5 .0) than the greater number of of letters (5.0) l .7) (}1 1 ,58)=83 .2, p<.OOO l ) . The same pattern emerged short word pairs ((1.7) (F(1,58)=83.2, p<.0001). even when we controlled for the total number of of letters in a word by testing the ratio of of the number of of shared letters to the total number of of letters. Regardless of of word affect, this ratio was significantly higher for long word 1 ,58)=1 8.9, p<.000 1). In both pairs (.50) than than for short word pairs (.33) (}1 (F(1,58)=18.9, p<.0001). measures, there was virtually no overlap between the short word and the long word conditions; so it was not possible to statistically test the relation between word length and performance independently of of the number of of shared letters. As predicted by the current analysis, word length generally depreciated of word performance in perceptual identification, but this detrimental effect of length was significantly less for affective than for neutral words. It was also response selection, causing a predicted that expectation would tend to impair response general decline of of performance, especially for neutral words. Some hint hint of of such an effect was found for long/neutral words; it, however, was far from significant. We We will discuss this latter effect after Experiment 2 is reported. time. To the the extent that response time decreases as the Response time. perceptibility of of a target improves, we should find response time to vary systematically with word affect, word length, and expectation. Affective enhancement would be indicated by a shorter response time for affective than for neutral words and affective impairment, by a longer response time for the former than for the latter. The mean response times are given in the lower half of of Table 22.. A preliminary analysis indicated that there was no systematic difference between positive and negative words with one exception. In the short-wordlexpectation-present condition, response time was longer for short-word/expectation-present negative words words (862 ms) than than for positive words (762 ms). Note, Note, however, that both response times were longer than that that for neutral words (7 (719 that 1 9 ms).
220
Chapter Chapter 5
Hence, they they are are both both demonstrating demonstrating aff affective impairment. Furthermore, Furthermore, the the Hence, ective impairment. accuracy measure has suggested that aff affective is no no stronger stronger for for suggested that ective impairment impairment is accuracy measure negative negative than for for positive positive words words in in this this as as well well as as in in the the remaining remaining conditions. conditions. At this point, the replicability At this point, then, then, neither neither the replieability nor nor the the theoretical theoretical significance significance of of the above above anomaly anomaly is is entirely entirely clear. clear. In what what follows, follows, as as in in the the analysis analysis of of the the the accuracy score, the two two valence valence categories categories (positive (positive and and negative) negative) of of affective affective accuracy score, the words were were combined. combined. words An on the An ANOVA ANOVA performed performed on the mean mean response response times times showed showed aa significant significant (l,68)=76.0, p<.001 , F main ( l,60)=12.79, main effect effect for for expectation expectation (F (F1(1,68)=76.0 , p<.001, F2(1,60)=12.79 , I 2 I ). Also I ; and 1 .7, p<.OO p<. O01, and minF'(l,76)=7 minF'(1,76)=71.7, p<. 001). Also significant significant were were the the two two main main p<.OO I) effects effects for for word word length length (F (F1(1,68)=4.12 and F F2(1,60)=26.1, p<.001) I (l ,68)=4. l2,, p<.05, and 2(l,60)=26. 1 , p<.OO and ( 1,68)=5.24, p<.05, F ( 1,60)=3.40, p<.08). Further, and word word affect affect (F (F1(1,68)=5.24, F2(1,60)=3.40 , Further, I 2 these were qualified qualified by by the the interaction these main main effects effects were interaction between between word word length length and and I (l,68)=4. l , p<.05, F word (1,60)=5 . l 8, p<.03), as well as by the word affect affect (F (F1(1,68)=4.11, p<.05, F2(1,60)=5.18, as well as by the I 2 second order order interaction interaction among among expectation, expectation, word word length, length, and and word word affect affect second (F O). (F1(1,68)=4.55 p<.04, F F2(1,60)=2.76, I ( 1,68)=4.55,, p<.04, 2 ( 1,60)=2.76, p=. I10). It must It must be be noted noted first first that that response response time time was was shorter shorter when when an an expectation expectation was present when it it was was absent absent ((1677 ms). This main main effect effect of of (8 10 ms) ms) than when 1 677 ms ). This was present (810 expectation simply simply indicates indicates that that response response time time was was much much briefer briefer when when expectation choices given in Further, consistent choices had had been been given in advance. advance. Further, consistent with with the the corresponding corresponding effect effect of of word word length length on on accuracy, accuracy, response response time time was was significantly significantly briefer briefer for for 1274 ms ). These long long words words (1214 ms ms)) than than for for short short words words ((1274 ms). These two two main main effects, however, effects, however, interacted interacted with with word word affect, affect, as as predicted predicted by by the the amplification amplification model. model. support of of the the amplification amplification model, model, affective affeetive enhancement enhancement was was found found In support in the response time in the long-wordlexpectation-present long-word/expectation-present condition; condition; response time was was shorter shorter for for words. Further, affective affeetive words words than for for neutral neutral words. Further, affective affeetive impairment impairment was was evident in in the the short-wordlexpectation-absent short-word/expectation-absent condition; condition; response time was was evident response time longer longer for for affective affeetive words words than for for neutral neutral words. words. Finally, Finally, the the influence influence of of affect the remaining all somewhere affect in in the remaining two two conditions conditions was was predicted predieted to to ffall somewhere between between the the above above two two extremes. extremes. Both Both in in the the long-wordlexpectation-absent long-word/expectation-absent condition and in in the condition and the short-wordlexpectation-present short-word/expectation-present condition, condition, affective affeetive impairment But, consistent consistent with impairment was was apparent. apparent. But, with the the current current analysis, analysis, the the magnitude this impairment magnitude of of this impairment effect effect was was no no greater greater than than the the comparable comparable ). effect observed in effect observed in the the short-wordlexpectation-absent short-word/expectation-absent condition condition (ts (ts < I1). Overall, Overall, the the pattern pattern for for response response time time paralleled paralleled that that for for accuracy. accuracy. First, First, in both measures affective enhancement was found when long words in both measures affeetive enhancement was found when long words were were expected. Second, also also in measures aff ective impairment expected. Second, in both both measures affeetive impairment was was observed observed when when short short words words were were unexpected. unexpected. Finally, Finally, the the influence influence of of affect affect in in the the
S.
Ki tayama Kitayama
22 2211
remaining two conditions was no stronger than in the above two conditions. It seems clear, then, that under the present condition of of backward pattern masking, both expectation and word length are importantly implicated in determining the precise effect (i.e., enhancement or impairment) of of stimulus affect on perceptual perceptual identification. It appears that that the failure by Kitayama ((1991) 1991 ) to obtain an interaction between word length and word affect can be attributed to the fact that 199 1 ) study, the perceptibility of that in the Kitayama ((1991) of a target was reduced reduced by an energy mask, which diminished stimulus contrast. an increase of of word length to contribute to The assumption is that in order for an the unique and unequivocal activation of of a relevant perceptual code, there 199 1 ) experiment, stimulus must be enough stimulus input. In the Kitayama ((1991) input was marginal to begin with, whereas in the present procedure procexture of of backward pattern masking, stimulus input was presumably great enough for the influence of of word word length to show up. Experiment 2
Experiment 2 was designed to determine the generality of of the results of of that allegedly the first experiment. Specifically, one can always suspect that perceptual effects might be due to artifacts operating at the post-perceptual, response end (Eriksen, 11963; 963; Goldiamond, 11958). 958). It would seem important, therefore, to ensure that that different methods of of assessing perceptual sensitivity of produce similar results. In the first experiment each word pair consisted of equivalently valenced words to exclude any response bias for or against affective stimuli (Natsoulas, 11965). 965). In Experiment 2 we used an alternative Dorfinan et aI., al., approach to control for bias. Following Dorfman ((1967; 1 967; Dorfman 11965), 965), each pair was now composed of of an affective (either positive or negative) word and a neutral word. Thus, response bias for or against an affective choice was allowed. A signal detection framework (Green & Swets, 11965) 965) was employed to measure this response bias and to yield an unbiased of perceptual sensitivity. In addition, in Experiment 2 expectation was index of employed as a within-subject variable and word length as a between-subject variable, reversing their roles in Experiment I1..
Method Method of a large number of of Overview and and subjects. The experiment consisted of Overview subjects. The trials (240 and 252 trials in the long and short word conditions, respectively). two thirds of of the trials, subjects were exposed to a brief flash of of either an On two
222 222
Chapter 5
affective or or aa neutral neutral target target word. word. On On the the remaining remaining trials, trials, no no word word was was affectivr flashed flashed although although the the subjects subjects were were led to to believe believe that that one one had had been been flashed. flashed. The The data data from from the the latter latter trials trials were were used used to to estimate estimate response response bias bias for for or or against against an an affective affective choice. choice. After After the the flash, the the subjects subjects were were presented presented with with the the target target and and nontarget nontarget words, words, and and asked asked to to choose the the one that that had had been been presented. On On half half the the trials trials the the pair pair was was given prior prior to to the the flash flash as as well, presented. resulting in about the the other other half resulting in an an expectation expectation about the target. target. In the half no such such expectation was was provided. provided. For For half half the the subjects subjects the the words expectation words were were relatively long long and they were were relatively seven undergraduates and for for the the other other half half they relatively short. Twenty Twenty seven undergraduates at at the the University of of Oregon Oregon (both (both males and and females) females) participated participated in the the experiment to partially partially fulfill their their introductory introductory psychology course course experiment to requirements. native English requirements. All the the subjects subjects claimed claimed to to be native English speakers. speakers. Materials and eqUipment. The equipment as as in Experiment Experiment 1 1 was was Materials and equipment. The same same equipment used. The to those used in Experiment The stimulus stimulus words were were virtually virtually identical identical to Experiment 1. between an word and and aa neutral neutral word of the the 1 . Pairs were were formed formed between an affective affective word word of same length. There and 20 negative-neutral positive-neutral and negative-neutral word There resulted resulted 20 positive-neutral pairs positive-neutral and and 24 total), and 118 8 positive-neutral pairs for for the the long words words (40 pairs in total), negative-neutral word word pairs pairs for for the words (42 pairs pairs in total), negative-neutral the short words total), as as shown shown in Appendix of trials were were constructed: the first kind, an an Three kinds of constructed: In the Appendix B. Three affective word word in each pair pair was was a a target; word in the affective target; in the second, second, a a neutral neutral word the pair was was a a target; target; and and in the the third no word was pair was flashed flashed as as a a target. target. The The third third of trials trials were were used to response bias or or ective kind of to assess response bias ffor or against against an an aff affective choice. There corresponding trials f or each of There were the the three three corresponding for of the pairs. pairs. The The resulting 1 20 trials trials (40 x resulting 120 x 3) in the the long word word condition and the 126 trials (42 xx 3) in the short short word word condition were tested tested under under each each of of the two two 120 xx 2) trials 126 xx 2) expectation orty ((120 expectation conditions. conditions. Two Two hundred hundred fforty trials and 252 ((126 trials trials were thus thus constructed constructed in the the long and the short short word conditions, conditions, respectively. respectively. The The order order of of trials trials was was randomized randomized for each each subject. All stimulus stimulus words words were were printed printed in upper-case upper-case letters. Procedure. Procedure. Subjects Subjects were tested tested individually. Upon arrival, arrival, they were randomly assigned to one of the two word length conditions (long versus assigned to of the word versus short). These conditions or the stimuli used. The long conditions were identical but ffor word word condition consisted of of 240 trials, while the short short word word condition consisted of of 252 trials (see above). The instructions instructions and the procedure proczxture were the same as in Experiment or the f ollowing changes. Experiment 11,, except ffor following A trial in which an expectation was available started with the presentation presentation of of a pair of of words with one to the right fight of of the center of of the screen other to the left. lefL On a trial in which the expectation expectation was not available, and the other .5 a pair pair of of strings of of eight "_"s "-"s appeared appeared instead of of the the two two words. After 11.5
S S.
Kitayama
223
(of either words or "-" strings) was replaced replaced by a central fixation sec, the pair (of point. The subjects had been told that that when the pair was given in advance, of them them would be presented presented on that trial and that that it was important important to keep one of pair in mind in order to perform well in this task. the pair When the the subject simultaneously pressed pressed the the two two response keys, the When fixation point disappeared disappeared and, 200 ms later, on two thirds of of the trials a target target word word was presented presented for 3333 ms, immediately followed by a pattern pattern mask, which lasted for 11500 500 ms. On the remaining one third of of the trials, no stimulus was presented presented before the the mask was shown for the same duration. The subjects had had been told told that that a target was was always presented "behind" the The post-experimental interview revealed that that no one had any suspicions suspicions mask. A post-experimental about this manipulation. manipulation. As in Experiment Experiment 11,, the the mask was a string of of "&"s about of the same length as the the target. Immediately after the disappearance disappearance of of the of corresponding word pair pair (the one shown at the beginning of of the mask, the corresponding trial in the expectation-present expectation-present condition) was shown. In a randomly chosen half of half of the cases an affective word appeared appeared on the right side and a neutral word, on the left, whereas in the other half the location of of the affective and neutral words was switched. Response time was measured measured from the onset the neutral of the choice pair, and the next trial automatically started after an interval of see with the average of of 2 sec. see. whose length varied randomly between I1 and 3 sec of the first first half half of of the entire set of of the trials ((120 At the completion of 1 20 and 1126 26 trials in the the long long and and the the short word conditions, respectively), the the subjects trials were given a short break before the second half. As in Experiment 11,, the experimental experimental trials trials were preceded preceded by 26 practice practice trials. Finally, two minor procedural procedural points may be noted. Experiment 2 consisted of of about about twice as many trials 1 . To complete the experiment within an hour, the trials as Experiment 1. sec to 11.5 .5 duration with which an expectation was shown was reduced from 2 see duration sec, see, and and the the intertrial intertrial interval was was reduced from the average of of 4.5 sec see to the average of of 2 sec. see. average Results and and discussion As in Experiment Experiment 11 both accuracy and and response response time were analyzed. Unlike Unlike in Experiment Experiment I1,, perceptual perceptual accuracy was estimated by adjusting adjusting the probability probability that that a given given word word was was chosen given given that the word word had had been presented presented (hit rate) rate) by the probability probability that the word was chosen given that no word had been presented presented (false alarm rate) within a signal detection framework. In what what follows, hit rate rate and false alarm alarm rate rate will be briefly
224 224
Chapter 5
described before an an adjusted adjusted accuracy accuracy measure is examined in greater detail. described Response time will will be be reported at at the the end. Response Hit rates. Hits rates, rates, shown shown in in the the upper upper part part of of Table Table 3, were submitted Hit to to an an ANOVA ANOVA with with one one between-subject between-subject variable variable (word (word length) and and three three within-subject variables (the presence/absence of of an expectation, affective versus neutral words, and and the valence (positive versus negative) of of an affective word). The effect for expectation proved significant The main effect p<.005, and minE'(l,68)=4.33, (/71(1,25)=7.82 (FI ( 1 ,25)=7.82,, p<.01, p<.OI, F2(1,78)=9.69 F2( 1,78)=9.69,, p<.005, minF'( 1 ,68)=4.33, p<.05) p<.05) indicating overall that that the hit hit rate was lower when the expectation was was present present (.70) than when it was absent absent (.74). This main effect, however, was qualified by by aa significant interaction between expectation and and target affectivity (F1(1,25)=4.25, (FI ( 1,25)=4.25, p<.05, p<.05, and F2(1,78)=4.27, affeetivity p<.05). The The F2( 1,78)=4.27, p<.05). detrimental effect effect of of expectation was was much more pronouncexl pronounced for the neutral words (.68 versus .75) than for for the affective words (.72 versus .74). False alarm rates. False alarm rates are given in the center of of Table 3. 3. In the present procexture, procedure, a false alarm rate for a neutral choice was equal to one minus the corresponding false alarm rate for an affective choice within any given condition. Hence, only false alarm rates for affective choices were analyzeA analyzed in an ANOVA ANOVA with one between subject variable (word length) and two within subject variables (the presence/absence of of an expectation and the of an affective word in aa pair). Only word length turned out to be valence of p<.005, and (Fl ( l,25)=7.24, p<.01, p<.OI, F2(1,78)=9.15, F2( 1 ,78)=9. l 5, p<.005, significant (F1(1,25)=7.24, minF'(1,67)=4.04, p<.05). There was a bias to choose a neutral word when 1 ,67)=4.04, p<.05). minF'( words were long (the false alarm rate for an affective .45)� when affeetive choice= choice=.45); words were short, however, this trend was reversed (the false alarm rate for an affective affeetive choice=.53). These These biases did not not depend on the valence (positive of the affect involved, somewhat contrary to what might be versus negative) of of Freudian defense. predicted by the notion of Accuracy. The results for the hit hit rates, described above, can be seen as a combination of of the accuracy with which a target stimulus was perceived and the bias for or against an affective 966). Hence, to affeetive choice (Green & Swets, 11966). obtain an unbiased estimate of of perceptual accuracy, hit rate must be adjusted for the corresponding response bias, especially when, as in the present case, response bias varied systematically across conditions. Following Pollack and Norman ((1964), 1964), Grier ((1971) 1 97 1 ) has suggested that the average area under a (nmning through the points [0, 0], hypothetical operating characteristic curve (running [x, y], and [I, [1, 1] 1] within a unit square, with false alarm rates on the x axis and y axis) can be used as an unbiased estimate of the perceptual hit rates on the y sensitivity (A (,4 �. 9. As recommended by Grier, the following formula was used to
S. S. Kitayama Kitayama
225 225
Table 33.. Mean Mean accuraCles accuracies (and (and standard standard deviations) deviations) In in perceptual perceptual Table identification as as a a function function of of word word affect, affect, word word length, length, and and expectation expectation identification (Experiment (Experiment 2). 2). Expectation Absent
Present
Word
Target Aff ectivity Affectivity
Length Valence
Affective
Neutral
Affective
Neutral
Hit Rate Long Words Words (N=I4) (N= 14)
Short Words 3) Words (N=1 (N= 13)
Positive
.73 (. 15) (.15)
.68 ((.15) .15)
.73 (. 1 7) (.17)
.75 ((.16) . 1 6)
Negative
.71 ((.18) . 1 8)
.69 (. 14) (.14)
.71 (.20) .71
.74 (. 1 7) (.17)
Combined
.72
.69
.72
.74
Positive
.74 (.16) (. 16)
.67 ((.. 119) 9)
1) .78 (.2 (.21)
.72 (.2 1) (.21)
Negative
.72 (.20)
.67 (.2 1) (.21)
.74 (.2 1) (.21)
.78 (.22)
Combined
.73
.67
.76
.75
Positive
.44 (. 14) .44 (.14)
.56 14) .56 (.(.14)
.44 (.09)
..56 56 (.09)
Negative
.47 (. 111) 1)
.53 ((.11) .1 1 )
.45 (.09)
.55 (.09)
Combined
.46
.54
.44
..56 56
Positive
..52 52 (. 115) 5)
.48 ((.15) .15)
(.14) ..57 57 (. 14)
(.14) .43 (. 14)
Negative
(.10) ..50 50 (. 10)
.50 (. 1 0) (.10)
1 2) .51 (. (.12)
.49 ((.12) . 1 2)
Combined
..51 51
.49
.54
.46
False Alarm Rate
Words (N=14) (N= 14) Long Words
Words (N= (N= 13) Short Words 1 3)
Sensitivity ?4 (,4') ? Long Words Words (N=14) (N= 14)
Short Words Words (N= (N= 13) 1 3)
Positive
14) .72 (. (.14)
.60 (. 16) .60(.16)
.72 ( . 1 3) .72(.13)
.67(.14) .67 (. 14)
Negative
.69 (. 14)
.63 ((.. 116) 6)
(. 16) .69 (.16)
.66 (. 1 5) .66(.15)
Combined
.70
.62
..71 71
.66
Positive
.68 (. 115) 5)
.64 (. 1 9) (.19)
1 9) .68 (.(.19)
(.18) .71 (. 1 8)
Negative
.68 (. 115) 5)
(.19) .64 (. 1 9)
.68 (.(.17) 17)
.72 (. 1 7) .72(.17)
Combined
.68
.64
.68
.71
226
Chapter 5
compute should theoretically compute this this index index of of perceptual perceptual sensitivity sensitivity (A ~� when, as should theoretically be be the the case, case, aa hit hit rate rate (y) was was equal equal to to or or greater greater than than the the corresponding corresponding false false alarm alarm rate rate (x): A' =
1 -
2
+ +
(y -- x)(1 x)(l +y-x) + Y x) (y -
------
4y( 1 -x) - x) 4y(1
Practically, however, aa hit hit rate rate can be smaller Practically, smaller than than the the corresponding alarm rate to the the variability variability inherent any empirical data especially false alarm rate due to inherent in any data especially when sensitivity is quite the comparable average area area under when quite low. In this this case, the under an imaginary imaginary "operating "operating characteristic characteristic curve" that through the three an that runs through I ], was I , 1], points, computed by the formula: points, [0, 0], [x, y], y], and and [ 1, was computed
A ' ==- -
1 -
2
+
(x - y)(l + (x-y)(1 + xx--yy) ) ------
4x(1 --y) y)
A sensitivity index (A (A)� was was computed in two different ways. First, First, it was computed over relevant word word pairs for each subject subject.. Analysis was . Analysis on these A of aa given effect over subjects (as indexed by A's's would suggest suggest the reliability of FI). F1). A ' wwas a s also calculated calculated over subjects separately separately for each word pair. Analysis on the the latter would would suggest the reliability of of the the effect over word ). In the part of 3 are presented the pairs (as indexed by F F2). lower of Table 2 pertinent means. A A'' is not a mere arithmetic mean of of individual choice data. Hence, even though FI F 1 and F F 22 can be computed in the ways described in the text, the corresponding minF' minF' cannot be defined based on the corresponding FI F 1 and F F 2. attempt was made to compute minF's. minF's. For the same 2 . Therefore, no attempt reason, the condition means of of A A's's differed slightly depending on whether A A's's were originally computed over word pairs or over subjects. As may be expected, however, the difference was was quite negligible. negligible. An ANOVA performed on the sensitivity index (A (A')) showed significant main effects for expectation (FI ( I ,25)=7.03, p<.02, and F (F 1(1,25)=7.03, F2(1,78)=8.18, 2( 1,78)=8. l 8, I p<.005), ,25)=4.49, p<.05, p<.005), and the affectivity of of a target (FI( (F1(1,25)=4.49, p<.05, and (1,78)=5.25, p<.03). These main effects, however, must be evaluated in F2(1,78)=5.25, F 2 view of an interaction between word length and target affectivity of I ), and an interaction (FI (1 ,25)=3 .59, p<.07, (/71(1,25)=3.59, p<.07, and F F2(1,78)=8.06, p<.O1), 2( l ,78)=8.06, p<.O between expectation and target affectivity (FI( 1,25)=3 .91, p<.06, and (F1(1,25)=3.91, { 1 ,78)=3 .5 1 , p<.07). No other effect approached statistical significance. F2(1,78)=3.51, F 2 Among others, there was no significant effect involving the valence (positive
Kitayama S. Kitayama
227 227
0.72 [']
Affectiv
1
Neutral
0.70 -
0.68 -
IO
0.66
--
0.64
"
I,m IO
0.62 -
0.60 -
Word Length/Expectation
Figure 4. Perceptual of word Perceptual accuracy accuracy (the sensitivity index A )� as a function of word length, and expectation affect, word expectation (Experiment 2).
or or negative) of of an affective word in each pair. We We thus found again that the perceptual of affect does not not depend on the valence of of the affect. perceptual influence of Figure 4. The The most crucial crucial prediction of of The relevant means are displayed in Figure the amplification amplification model is that that affective enhancement would be most likely in long-word/expectation-present condition whereas affective affectivr impairment the long-wordlexpectation-present short-word/expectation-absent condition. In would be most likely in the short-wordlexpectation-absent support support of of this analysis, a significant significant affective affectivr enhancement effect was obtained in the long-wordlexpectation-present long-word/expectation-present condition (t} (t1(49)=2.84, p<.01, obtained (49)=2.84, p<.O I, and t2( 1 56)=2. 80, p<.OI). t2(156)=2.80, p<.01). This enhancement effect, however, was reversed, short-word/expectation-absent condition. The albeit non-significantly, in the short-wordlexpectation-absent simple interaction interaction representing this pattern proved significant (t}(25)=2.95, (t1(25)=2.95, p<.01, p<.O I , and t2(78)=2.55, p<.05). In the remaining two conditions, a marginal of affective affectiw enhancement was evident (ps > ..10). effect of 1 0). Consistent with the amplification model, however, these affective enhancement effects were no long-word/expectation-present stronger than the comparable effect in the long-wordlexpectation-present condition.
228 228
Chapter 5
addition, as as in in Experiment Experiment 11,, an increase in in word word length length diminished diminished In addition, an increase the This effect the general general performance performance level. level. This effect can can be be attributed attributed to to aa greater greater choice choice condition. As difficulty difficulty in in the the long long word word condition condition than in in the the short short word word condition. As predicted, less for predicted, however, however, this this decline decline was was less for affective affective words words than for for neutral neutral words. Further, in decreased the words. Further, in Experiment Experiment 22 expectation expectation decreased the performance performance for for neutral neutral words. words. No No such such effect effect was was found found in in Experiment Experiment 11.. In In another another recent recent experiment that only relatively short words relatively short words (Kitayama, (Kitayama, 1990), a a similar similar experiment that used used only detrimental effect effect of of expectation expectation was found only only for for neutral neutral words words of of an an detrimental was found extremely extremely high high frequency: frequency: it it was was not not found found for for word word frequencies frequencies comparable comparable those used in the the present present studies studies (see Figure 2). 2). All All in in all, such aa those used in (see Figure all, while while such detrimental ionally, the detrimental effect effect of of expectation expectation seems seems to to occur occur occas occasionally, the exact exact condition for occurrence is is currently currently unclear. detrimental effect condition for its its occurrence unclear. This This detrimental effect is is consistent possibility, suggested consistent with with the the possibility, suggested by by the the previous previous research, research, that that explicit formulation formulation of of aa "hypothesis" "hypothesis" or or expectation expectation can can impair impair perceptual perceptual explicit Such a identification identification (e.g., (e.g., Bruner Bruner & Potter, Potter, 1964; Lawrence Lawrence & Coles, Coles, 1954). 1954). Such a "hypothesis" only one is "hypothesis" may may direct direct one's one's attention attention to to only one type type of of information information that that is potentially potentially available available (e.g., (e.g., orthographic, orthographic, phonemic, phonemic, semantic, semantic, etc., etc., in in perceptual identification). Information perceptual identification). Information on on the the ignored ignored dimensions dimensions may may then then be be rmance in unavailable, unavailable, leading leading to to poorer poorer perfo performance in the the presence presence of of an an expectation expectation than in in its its absence. absence. Thus, Thus, as as predicted, performance is in the the than predicted, performance is generally generally worse worse in presence of of an an expectation expectation than than in in its absence, but this detrimental detrimental effect effect of of presence its absence, but this expectation was was significantly significantly less less for for neutral neutral words. words. expectation for affcctivc affective words words than for time. Mean response response times times are arc given given in in Table Table 4. Data Data from from the the Response time. trials in in which a target target was was actually actually prcscntexi were analyzed. analyzed. Unlike Unlike in in the the trials which a presented were first experiment, no no significant significant effect effect involving involving affect affect was was observed. observed. first experiment, Nevertheless, it it is is noteworthy noteworthy that that aa pattern pattern that that closely closely resembles resembles that that of of Nevertheless, Experiment 1 (Table 2) is discernible in Table 4. Relevant to the present Experiment 1 (Table 2) is discernible in Table Relevant to the present hypothesis are arc the the response times when when aa target target stimulus stimulus was was shown shown (i.e., (i.e., hypothesis response times affcctivc and and neutral neutral target target types). types). When When combined combined over over the the two two valence valence affective categories (positive and negative) of affective affcctivc words, words, response time was was categories (positive and negative) of response time shorter for affective affcctivc than for for neutral neutral words words in in the the long-wordlexpectation long-word/expectationshorter for present condition condition (affective (affcctivr enhancement; enhancement; 585 ms versus versus 6615 ms). This This effect effect present 1 5 ms). was reversed both in in the the long-wordlexpectation-absent long-word/cxpoctation-absent condition condition (affective (affe~ivc was reversed both impairment; 11115 ms versus versus 11100 ms)) and and in in the the short-wordlexpectation short-word/expectation1 15 ms 100 ms impairment; present versus 6 1 3 ms); finally, when present condition condition (638 versus 613 ms); fmaUy, when there there was was no no expectation and and the the words words were were short, short, the the reversal reversal was was somewhat somewhat stronger stronger expectation ms versus ms). It is not clear why this pattern, which which was was ((1047 1 047 ms versus 985 ms). is not clear why this pattern, statistically significant significant in in Experiment failed to attain attain statistical statistical significance significance statistically Experiment 1, failed in Experiment Experiment 2. 2. It It seems seems to to fair fair to to suggest, suggest, consistent consistent with with Kitayama Kitayama ((1991), 1 99 1), in
S. S
Kitayama
229
Table 4. Mean Mean response response time in milliseconds milliseconds (and (and standard standard deviations) as a a Table function of of word word length, expectation, expectation, and target target type type (Experiment 2). function Expectation Absent
Present
Word
Target Target Type Type
Length Valence
Affective Neutral Neutral Affective
None None
Affective Affective Neutral Neutral
None None
Long Long Words Words Positive Positive
62 5562 (212) (2 12)
627 (263)
676 (212)
11107 1 07 (351) (35 1)
11122 122 (325)
11318 318 (351) (35 1)
Negative
607 (240)
603 (287)
689 (239)
11123 123 (394)
1077 1077 (317) (3 17)
11.325 325 (420)
585
615 6 15
683
11115 1 15
11100 100
11321 321
Short Words Words Positive Positive
65 6511 (28 1) (281)
556 556 (213) (2 1 3)
627 (211) (21 1)
11068 068 (384)
994 (337)
11318 318 (333)
Negative
628 (302)
6 656 56 (380)
8311 83 (299)
1031 1031 (256)
979 (343)
1401 1401 (378)
638
6 13 613
743
1047 1047
985 985
11366 366
Combined
Combined
that that although although perceptibility perceptibility can can be indexed by both both accuracy accuracy and response response time, the former is a more reliable and and sensitive measure measure than than the latter latter under under current conditions conditions of of impoverished impoverished stimulus presentation. presentation. the current Comparison between Experiments J1 and and 2. The The above results results replicated replicated the general pattern pattern obtained obtained in Experiment Experiment 11.. In both experiments, experiments, affective enhancement enhancement was was observed observed when long words were expected; expected; whereas whereas affective impairment impairment was was observed observed when short words were unexpected. unexpected. Further, Further, in both experiments experiments the the influence of of affect in the remaining two two conditions (i.e., when when long words words were unexpected, unexpected, and and when short words words were expected) expected) was was no stronger stronger than than in the above above two conditions conditions (i.e., when long words were expected, expected, and and when when short words were unexpected). unexpected). Thus, Thus, these converge with the past 990) these two two experiments experiments converge past evidence (e.g., Kitayama, Kitayama, 11990) to that both to suggest that both word word length length and and expectation expectation are crucial in determining of stimulus stimulus affect affect on perception. perception. Both an an increase increase in word word length the influence influence of and the presence of an expectation render affective enhancement more and of expectation render enhancement more likely, and impairment impairment less likely. and
Chapter 5 Chapter
230 230
The The Amplification Amplification Model Model Evaluated Evaluated
Summary oft ofthe present research Summary he present It was was hypothesized that that affect affect elicited by preattentive processing affect inherent in a a stimulus may amplifies attentive processing. Accordingly, affect either impair or enhance enhance the veridical perception perception of either of the stimulus itself depending depending on the accuracy with which attentive processing is directed directed to the relevant perceptual code. When it is difficult to to locate aa perceptual code corresponding to to an an impinging stimulus so that that attentive processing is misdirected to an irrelevant perceptual code, affect magnifies invalid to an perceptual perceptual information, leading to to affective impairment (i.e., poorer perception of of affective than than neutral stimuli). If, however, the difficulty of of locating the relevant code is alleviated by some means so that that attentive locating processing is correctly directed to the relevant code, then the affect-caused of attentive processing should lead to affective enhancement amplification of (i.e., better perception for for affective than for for neutral stimuli). In perceptual identification experiments like the the present ones, an extremely impoverished target word is presented. Under these conditions, it should be difficult difficult to determine exactly which perceptual code corresponds to an impinging stimulus. Hence, unless the difficulty in locating the relevant perceptual code is alleviated by some means, affective impairment should be that should found. The two experiments reported here examined two variables that aid in determining the relevant code. First, a valid expectation, or priming, of the stimulus. should activate the relevant code prior to the presentation of an increase in word length should conduce to unequivocal activation Second, an of the relevant perceptual code, especially when high stimulus contrast is of used. Under these conditions, initial, letter-level activations should impose more constraints on the identity of of the word. In general congruence with this analysis, in both experiments perception affective was better for aff ective than neutral words (affective enhancement) when long words were used and an expectation was present, but it was worse for the former than the latter (affective impairment) when short words were used and of there was no expectation. In the remaining two conditions where only one of the two factors aids in locating the relevant code (long words/unexpected, and short words/expected), the influence of aff affect ect was no stronger than in the above two conditions where these two factors were either both present or both absent. Response times tended to corroborate the pattern obtained for the
S.
Ki tayama Kitayama
23 2311
for accuracy measures. Finally, virtually the identical effects were observed f or both both positive and negative affective words words..
Methodological comments
Confounding Conf ounding variables. These findings were documented here using a of occurrence of of these large number of of stimulus words. Further, the frequency of at a level comparable to previous studies (the words was carefully controlled at of the Kitayama, 11990, low-frequency condition of 990, 11991, 99 1 , studies). This of past past methodological point is not trivial, because the generalizability of otten suspect due to the small number of of stimuli tested. It is now findings was often difficult to dismiss the findings as mere artifacts due to certain idiosyncratic stimuli or a confounding between affect and word frequency. Response bias. There is always room to suspect that allegedly perceptual effects might be due to methodological artifacts operating at the post-perceptual, response end. For example, recognition threshold often used in the perceptual defense literature in the 11950s 950s did not eliminate the of such response bias (Eriksen, 11963; possibility of 963 ; Goldiamond, 11958). 958). In view of this, it should be emphasized that the present experiments used two of bias.. different forced choice formats both designed to control for response bias the past past studies, studies, response time was analyzed along with Also, unlike in the of results across the two accuracy measures. We found a similar pattern of forced choice methods, with both accuracy and response time, making it that the observed influence of of affect is due to response artifacts. unlikely that Evaluation All in all, along with the studies reviewed earlier, which demonstrate that other variables such as stimulus contrast and word frequency can interact
of the word, the present with word affect to determine the perceptibility of research has provided clear evidence for the amplification model of of affect affectcognition interaction. Affective enhancement enhancement becomes more likely and impairment, less likely when the difficulty in locating a relevant perceptual code is alleviated or, more specifically, when (i) stimulus contrast increases,
exposure time increases within a range of of relatively short times, (iii) word (ii) exposure frequency increases, (iv) word length increases, or (v) a valid expectation is made available. Finally, the affective influence is mostly the same for both
positive and negative affect.
232 232
Chapter 5
The research guided by by the the amplification amplification model, model, described described so so far, far, has has The research guided raised aa number number of of important important issues. issues. The The rest rest of of this this paper paper will be devoted devoted to to raised will be a topics. First, First, the a discussion discussion of of four four such such topics. the key key idea idea in in the the amplification amplification model model is that is that attention attention mediates me~atos an an interaction interaction between between affect affect and and cognition. cognition. It It is is important, important, then, then, to examine examine this this idea idea within within aa broader broader context context of of attention attention research. past three research. We We will will review review major major theories theories in in the the past three decades decades and and explicate the the current current model's model's relation with each each of of them. them. Second, Second, although although the the explicate relation with present research research has perception, the present has focused focused exclusively exclusively on on immediate immediate perception, the amplification model may amplification model may prove prove applicable applicable to to other, other, non-perceptual non-perceptual tasks. tasks. Relevant empirical empirical evidence evidence will will be be reviewed. reviewed. Third, Third, we we are are now now in in aa far far Relevant better position than was better position was possible possible before before to to critically critically evaluate evaluate the the notion notion of of psychological both in psychological defense defense both in perceptual perceptual identification identification and and in in real real life. life. We We will thus thus examine examine the the current current theoretical status of of Freudian Freudian defense defense in in view view of of will theoretical status the Finally, some will be the current current findings. findings. Finally, some directions directions for for future future research research will be discussed. discussed. Relations with with Extant Extant Theories Theories of of Attention Relations
Early versus late selection Since 1 95 8), theories Since Broadbent Broadbont ((1958), theories of of attention attention have have been been classified classified according according to to the the location location of of "information "information filter", filter", whereby whereby sensory sensory input input is is selected selected for for further further processing. processing. Placing Placing the the filter filter early early in in the the processing, processing, Broadbent selection is Broadbcnt argued argued that that attentional attcntional selection is based based on on crude crude sensory sensory or or external as voic~ voice tone external cues cues such such as tone or or location. location. Consistent Consistent with with this this early early selection selection model, model, selection selection is is quite quite easy easy when when distinct distinct physical, physical, external external cues cues are are readily readily available. available. Nevertheless, Nevertheless, the the model model did did not not seem seem to to account account for for some observations. For some other, other, quite quite common common observations. For example, example, attention attention is is often oi~en captured names) presented captured by by aa significant significant stimulus stimulus (e.g., (e.g., one's one's own own names) presented in in aa channel 959; channel that that is is being being ignored ignored (the (the "cocktail "cocktail party" party" effect; effect; e.g., e.g., Moray, Moray, 11959; Treisman, 964). Several Treisman, 11964). Several researchers researchers therefore therefore proposed proposed late late selection selection models models (e.g., 963; Marcel, 983b; Norman, 968; Shallice, (e.g., Deutsch Deutsch & Deutsch, Deutsch, 11963; Marcel, 11983b, Norman, 11968; Shallic~, 11987), 987), whereby whereby all all sensory sensory inputs inputs are are thought thought to to receive receive full full processing processing before before one one ofthem of them is is selected selected for for attention attention and and thus thus for for conscious conscious awareness. awareness. The The amplification amplification model model is is congruent congruent with with the the late late selection selection models. models. In In fact, fact, if if it it is is only only the the early early selection selection model model that that were wore valid, valid, the the phenomena phenomena of of affective affective enhancement enhancement and and impairment impairment would would seem seem quite quite puzzling, puzzling, for for they they imply based on imply selection selection of of sensory sensory data data based on affective affective or or semantic semantic information. information. Indeed, Indeed, Broadbent Broadbent himself himself supplied supplied one one of of the the best best early early demonstrations demonstrations of of
S. S.
Kitayama
233 233
affective impairment impairment (Broadbent (Broadbent & Gregory, Gregory, 11967; see also also Broadbent, Broadbent, affective 967; see Having unexpectedly unexpectedly found found aa reliable reliable affective affective impairment impairment effect, effect, he he 11975). 975). Having remarked: ""...we have to to suppose suppose that that emotional emotional words words do do not not behave behave like like remarked: . . . we have words of of low low probability probability...there appears actually actually to to be be aa failure failure of of the the words . . .there appears stimulus information reformation to to reach reach the the perceptual perceptual mechanism mechanism...Frankly we were were stimulus . . . Frankly we not expecting expecting these these results, results, and and have have described described [the [the findings] findings] at at an an informal informal not level as as showing showing that that 'Freud was right right after after all' (Broadbent & Gregory, Gregory, 11967, 'Freud was all' (Broadbent 967, p. 583)." It would would seem seem more more reasonable, reasonable, then, then, to to suppose suppose that that the the early early and and the the It late selection selection models models of of attention attention specify specify quite quite different different and and separate separate aspects aspects late of attentional attentional selection. selection. Whereas Whereas the the early early selection selection model model is is concerned concerned of primarily with with selection selection based based on on external external or or physical the late late selection selection primarily physical cues, cues, the models focus focus on on selection selection based based on on internally internally generated generated features features such such as as the the models activation strength strength of of relevant relevant or or irrelevant irrelevant internal internal representations representations (Norman, (Norman, activation Both selective functions may may co-exist in the the human human brain brain and and 11968). 968). Both selective functions co-exist in supplement each each other other to to coordinate coordinate an an action action (such as conscious conscious thought thought or or supplement (such as overt action) action) with with respect respect to to an an input input (Allport, (Allport, 11989). As proposed proposed by by the the overt 989). As early selection selection model, model, information information can can be be selected selected for for conscious awareness in in early conscious awareness terms of of physical or external external cues. cues. Yet, Yet, as as proposed by the the late late selection selection terms physical or proposed by models, information information unselected unselected on on aa physical or external external basis basis may may still still models, physical or receive considerable considerable preattentive preattentive processing processing and and can can provide provide aa basis basis for for receive subsequent attentional attentional selection. selection. subsequent A neuroscience supports supports this view. A recent recent development development in in cognitive cognitive neuroscience this view. Drawing tomography (PET) (PET) as as from Drawing on on data data from from positron positron emission emission tomography as well well as from patients distinct patients with with brain brain injuries, injuries, Posner Posner and and his his colleagues colleagues have have suggested suggested distinct anatomical structures for operations (Posner, Peterson, anatomical structures for each each of of these these selection selection operations (Posner, Peterson, Fox, that require require attentional selection Fox, & Raichle, Raichle, 1988). 1 988). Specifically, Specifically, tasks tasks that attentional selection based orienting to external cue) greatly based on on spatial spatial information information (e.g., (e.g., orienting to an an external cue) seem seem greatly impaired in patients with parietal whereas in tasks that that require parietal damage, whereas attentional attentional selection based on internal information (e.g., (e.g. , generating uses [pound] for an object Stroop color color naming task), more [pound] for an object [hammer], [hammer], and and aa Stroop naming task), more anterior anterior areas areas including including the the anterior anterior cingulate cingulate gyrus gyrus and and the the surrounding surrounding regions display the the strongest strongest PET PET activation. Thus, Thus, aa system subserving selection based based on on external external information (called the posterior attention attention system) appears appears quite quite distinct distinct in both both anatomical anatomical structure structure and and function from aa system serving serving selection selection based based on on internal internal information information (called (called the the anterior anterior attention attention system; see Posner Posner & & Peterson, Peterson, 1990, 1 990, for for aa review).
Chapter 5
234 234
and attentive processing processing Preattentive and
Along with with prior prior theories theories of of late late selection selection (e.g., Deutsch Deutsch & & Deutsch, Deutsch, 1963; 1 963; Marcel, Marcel, 1983a; 1 983a; Norman, Nonnan, 1968), 1 968), the the amplification amplification model model elaborates elaborates on implicated in selection based on internally generated the mechanisms implicated information. infonnation. Specifically, Specifically, it rests on aa formulation fonnulation in which the development of a conscious percept is decomposed into two relatively separate separate component component of a conscious percept two processes (preattentive (preattentive processing processing and attentive attentive processing), first made processes explicit explicit by Posner Posner and and Snyder Snyder (1975) ( 1 975) and subsequently subsequently elaborated on by aa number of of researchers researchers (e.g., Neely, 1977). 1 977). number To this formulation To this fonnulation the amplification amplification model has has added a a possibility possibility that that affect, affect, caused by preattentive preattentive processing, influences subsequent subsequent processing. Because Because considerable considerable evidence now now suggests that that preattentive preattentive processing processing is sufficient to produce some affect, this addition seems logical. Indeed, Posner (1978) already made a similar suggestion suggestion over ten years years ago. In his ( 1 978) had already ofMind, he remarked (p. 238): seminal Chronometric Exploration of ..it of aa ..9 .it may be possible that feedback from the emotional arousal of given word can influence the the conscious processing of that word of that itself. This provides aa mechanism whereby an an emotional word can of our our attention to itself. influence the direction of Around 1 973, p. 79) also made a Around the same time, Broadbent ((1973, a similar comment: erent kind of ... Consider then a diff different of perceptual system, in which which... . . . the output ...the output of of (a perceptual) perceptual) mechanism feeds back to the mechanism itself . . .the itself in such a way as to change its parameters ..... (preconscious) identification of of aa word as nasty then produces an emotional state which changes the strategy of of (further) intake of of information. information.... . This in tum turn might change the (conscious) identification of of the the stimulus. [words in parentheses were added by the present author to clarify the point.] present According to ect is limited to to Posner, the effect of of preconscious aff affect conscious processing. The present research, however, has broadened the range of of such an effect, providing evidence that preattentive affect can also amplify attentive (yet (yet preconscious) processing required required to develop a
S.
Kitayama
235 235
conscious conscious percept, percept, thus thus confinning confirming Broadbent's Broadbent's conjecture. conjecture. The The entire entire mechanism suggested by mechanism suggested by Posner Posner can can be be preconscious. preconscious.
Resource theory o fattention of Another Another key key assumption assumption of of the the amplification amplification model model is is that that attention attention is is aa set of of mental mental capacities capacities that that can can be be increased increased (or (or amplified). amplified). This This idea idea can can be be set traced traced back back to to Freudian F reudian notion notion of of attentional attentional cathexis, cathexis, which which was was further further developed developed by by more more recent recent ego-psychologists, ego-psychologists, most most notably notably by by Rapaport Rapaport (e.g., (e.g., 11967). 967) . Subsequently the idea Subsequently the idea has has provided provided impetus impetus to to the the resource resource theory theory of of attention 1 973) and attention by by Kahneman Kahneman ((1973) and to to aa number number of of experiments experiments on on attention attention (see e.g., Dark, 11985, 985, for (see e.g., Johnson Johnson & Dark, for aa review). review). Nevertheless, Nevertheless, there there are are some some obvious difficulties in (Allport, 11989; 989; Navon, 984). obvious difficulties in the the resource resource theory theory (Allport, Navon, 11984). Specifically, Specifically, the the notion notion of of aa single single set set of of generic generic processing processing resources resources is is very very likely a albeit a likely a too-simplistic, too-simplistic, albeit a useful useful and and heuristic, heuristic, fiction. fiction. Future Future theoretical theoretical work this area must specify work in in this area must specify the the exact. exact, mechanism mechanism or or computation computation involved involved in attentive attentive processing processing (see e.g., LaBerge Treisman, 11988, (see e.g., LaBerge & Brown, Brown, 11989, 989, Treisman, 988, in for for analyses analyses for for spatial spatial attention) attention) and and delineate delineate what what is is to to happen happen to to this this computation computation when when attentive attentive processing processing is is "amplified". "amplified". The The present present data data should constrain constrain such such an an endeavor endeavor of of further further theorizing. theorizing. should
Amplification of of Attention in Other Domains Amplification Other tasks With enhancement and With the the amplification amplification model, model, both both affective affeetive enhancement and impairment in in perceptual perceptual identification identification can can be be seen seen as as aa variant of the the general general impainnent variant of capacity of of affect to amplify amplify perceptual perceptual and and cognitive cognitive processes. processes. This This capacity affect to perceptual influence of of affect affect was was first proposed by by Easterbrook Easterbrook ((1959), who perceptual influence first proposed 1 959), who suggested that that affect affect limits limits the the range of cues that can can be be utilized utilized by by the the suggested range of cues that perceiver. perceiver. Thus, Thus, affect affect may may enhance enhance ("amplify") ("amplify") the the perception perception of of central, central, focal features. features. Recent Recent studies studies by by Christianson Christianson and and colleagues colleagues (Christianson focal (Christianson & Loftus, 99 1 ; Christianson, 99 1 ) have Loflus, 11991; Christianson, Loftus, Loftus, Hoffman, Hoffman, & Loftus, Loftus, 11991) have demonstrated that that detailed detailed information information at at the the center of aa briefly briefly shown shown demonstrated center of emotional scene significantly better better recalled recalled than than the the same same detail from aa emotional scene is is significantly detail from comparable comparable neutral neutral scene. scene. Furthermore, Furthermore, there there is is some, some, albeit albeit mixed, mixed, evidence evidence that that enhanced enhanced perception perception of of aa central central emotional emotional stimulus stimulus is is accompanied accompanied by by an an impainnent impairment of of the the perception perception of of peripheral, peripheral, nonfocal nonfoeal features. features. Although Although Christianson et al. al. ((1991) find any any evidence, evidence, Erdelyi Erdelyi and and Appelbaum Appelbaum Christianson 1 99 1) failed failed to to find
236 236
Chapter 5
(1973) that a highly emotional emotional symbol located located at the center center of aa ( 1 973) did show that tachistoscopic tachistoscopic display decreases decreases incidental incidental memory of peripheral peripheral stimuli. stimuli. Likewise, Likewise, Kitayama (1996) ( 1996) finds finds that that emotional emotional intonation intonation of a speech speech can disrupt the the attentive attentive processing and, and, thus, thus, subsequent memory, of the attendant verbal content. content. The same theme is evident in the literature on arousal and memory. Eysenck (1976) ( 1976) has concluded that "high levels literature, M. Eyscnck Reviewing this literature, of arousal have the effect of biasing the subject's subject's search process toward readily accessible sources sources of stored information information more than is the case with lower levels levels of arousal." A similar point has been raised more recently by Paulhus and Levitt (1987), ( 1987), who showed that a para-fovcal para-foveal affcctive affective stimulus stimulus dominant, thinking intensifies "ego-enhancing, "ego-enhancing,"" presumably personally dominant, tendencies. tendencies. A more elaborate model of arousal and memory has been proposed by Humphmys Rovcllo ((1984). Humphreys and Revelle 1984). Consistent with previous theorizing by Walker ((1958), 1 958), they have suggested that arousal produced by stimulus affect interferes interferes with short-term retention of the stimulus (because arousal amplifies competing responses responses along with a correct response) while improving long-term memory trace (because it helps sustain the "transfer" "transfer" of the correct response to to long term memory). In support of the model, model, several studies have found that in a paired associate learning task, task, recall of affective affective materials is poorer than recall of neutral materials immediately after learning; but after a long delay the opposite pattern emerges (e.g., Kleinsmith & relationship between the Kaplan, 11963; 963; Walker & Tarte, 11963). 963). The exact relationship amplification model and the Humphreys-Revelle model has yet to be specified, although both have in common the assumption that affect amplifies some relevant cognitive or perceptual operation(s). operation(s). Amplification still another Amplification due to affect has been demonstrated in still measure of perceptual sensitivity. Worthington ((1969) 1969) presented two words threshold, and had subjects subjects in succession, one at a time, time, below a recognition threshold, judge which was brighter. It It.was . was found that taboo words were judged as brighter than neutral words even though the subjects subjects were not aware of the identities of the words. According to the amplification model, model, preattentively preattentively induced affect amplified the attentive processing of the word itself, producing a more vivid image for taboo words than for neutral words. Further support for this analysis comes from a second experiment, in which an affective and a simultaneously. Under this condition, the neutral stimulus were shown Simultaneously. disappeared. Perhaps, Perhaps, attentive processing foregoing effect of affect entirely disappeared. itself, another amplified by an affective word was applied to either the word itself, (neutral) word, or both of them, erasing any difference in the vividness of the
S. S.
Kitayama
237 237
two images. images. Finally, Finally, the the literature literature of of "sensory "sensory interaction" interaction" in in the the 11940s and two 940s and 11950s 950s tended to demonstrate tended to demonstrate that that aa simultaneously simultaneously presented presented extraneous extraneous stimulus threshold of stimulus lowers lowers the the detection detection threshold of aa focal focal target target (Zajonc (Zajonc & Dorfinan, Dorfman, see Dixon, Dixon, 11980, for aa review). review). This This literature literature is is also also consistent consistent with with 11964; 964; see 980, for the present present analysis, analysis, insofar insofar as as the the presentation presentation of of an an extraneous extraneous stimulus, stimulus, by by the itself, itself, may may evoke evoke some some arousal. arousal. All in in all, all, the the amplification amplification of of perceptual perceptual and and cognitive cognitive processes processes by by All affect affect seems seems quite quite pervasive pervasive across across aa wide wide variety variety of of experimental experimental tasks. tasks. Affect can can amplify amplify both both behavioral behavioral and and cognitive cognitive responses responses that that are are Affect "dominant. model, stimulus "dominant."" According According to to the the present present model, stimulus affect affect can can amplify amplify attentive attentive processing, processing, which which is is directed directed to to aa strongly strongly activated activated code, code, that that is, is, "dominant" perceptual perceptual information. information. Thus Thus there there is is clear clear correspondence correspondence "dominant" between the current model model and and prior (mostly behavioral) behavioral) theories theories on on between the current prior (mostly amplifying functions of affect. affect. The The amplification amplification model adds to to the the previous previous amplifying functions of model adds literature literature the the possibility possibility that that amplification amplification by by affect affect may may occur occur even even in in aa very very early, would seem further specifying early, preconscious preconscious stage stage of of processing. processing. It It would seem that that further specifying the the nature nature of of this this amplification amplification may may be be aa key key to to understanding understanding the the interaction interaction between cognition that occurs preconsciously. between affect affect and and cognition that occurs preconsciously. Valence ect Valence of of aff affect
Although the the present present research research has has found found no no evidence evidence for for any any effect effect of of the the Although valence of of affective affective stimuli, stimuli, this by no no means means implies implies that that valence not valence this by valence is is not activated preconsciously. preconsciously. In In fact, fact, existing existing evidence suggests that that positive positive and and activated evidence suggests negative affect can can be be evoked evoked by by undetectable undetectable stimuli stimuli (Greenwald (Greenwald et et aI., al., negative affect Niedenthal, 11990; Zajonc et et aI., al., 11989). The present present study study does does suggest, suggest, 11989; 989; Niedenthal, 990; Zajonc 989). The however, that that preattentively preattentively elicited elicited affect affect is is likely to amplify amplify subsequent subsequent however, likely to processing regardless of of its its valence. valence. The The present present finding finding also leaves leaves open open the the processing possibility that that positive positive and and negative negative affect affect may may have have divergent divergent effects effects on on possibility attention in in tasks other than than immediate immediate perception. perception. Affect Affect examined examined in in the the attention tasks other present research research is is neither differentiated nor nor consciously consciously recognized. recognized. Affect Affect of of present neither differentiated this kind is shown to to amplify amplify attention attention regardless regardless of of valence. valence. More More elaborate elaborate and differentiated differentiated forms forms of of affect, affect, however, however, may may entail entail quite quite different different and processing consequences. processmg As proposed proposed by by appraisal appraisal theories theories of of emotion emotion (Frijda, (Frijda, 11986; Smith & As 986; Smith Ellsworth, 11986; Scherer, 1984), 1984), the the processing processing of of affect affect is is Ellsworth, 986; Roseman, 11984; 984; Scherer, likely to to proceed proceed from from the the detection detection of of global global valence valence or or significance significance to to the the likely identification of differentiated affective affective states in terms of discrete discrete identification of more more differentiated states in terms of emotion categories. Thus, Thus, consistent consistent with an earlier earlier formulation formulation by by Schachter Schachter emotion categories. with an
238 238
Chapter Chapter 5 5
and Singer Singer ((1962), undifferentiated affect is combined with a variety 1962), initially undifferentiated of cognitions cognitions about about the the surrounding surrounding situation situation ("appraisals") ("appraisals") to fform a full fullof OnD a fledged emotional state. It may may well be be the the case case that that once valence valence is clearly
differentiated, perhaps, perhaps, in conscious conscious awareness, awareness, it recruits recruits characteristic characteristic differentiated, response 984� Lang, response patterns patterns (Davidson, (Davidson, 11984; Lang, Bradley, & Cuthbert, Cuthbert, 1990). 1990). Specifically, Specifically, negative negative affect affect may may mobilize
coping
responses, responses,
causing causing
additional additional attention attention to be allocated allocated to to the stimulus event, whereas whereas positive positive
affect may may halt halt any any such such active active attentional attentional and and motor motor processes (Tucker (Tucker & affect Williamson, 11984; Taylor, 11991, integrative review of of this Williamson, 984� see Taylor, 99 1 , ffor or an integrative literature). Hence, Hence, literature).
an emotional event has become become conscious, conscious, more more once an
attention may may be allocated allocated to to the the ffocal stimulus only if if the the event is negative. attention ocal stimulus Several studies studies have have examined examined allocation allocation of of attention attention to clearly
perceptible stimuli 980), perf onnance perceptible stimuli by by measuring measuring gaze gaze time (e.g., Fiske, 11980), performance in a 986), speed in a secondary secondary task task (e.g., MacLeod, MacLeod, Mathews, Mathews, & Tata, Tara, 11986), lexical judgment judgment task task (Matthews, (Matthews, Pitcaithly, & Mann, Mann, 1995), 1995), and and interf interference erence a Stroop Stroop color color naming task task (e.g., Pratto Pratto & John, 11991). have in a 991). These studies have ound that ummon more attention consistently ffound that negative negative stimuli ssummon attention than than either ound f or a positive or neutral neutral stimuli. Analogous Analogous effects have also been ffound for a negative negative mood mood state. Derryberry Derryberry and and his colleagues colleagues (Derryberry, Brandt, Bran&, & used a a probe probe reaction reaction time task task and and ffound that reaction reaction time time to aa Reed, 11991) 99 1 ) used ound that central probe probe is brief briefer after failure failure ffeedbacks concurrent tasks tasks (designed central er after eedbacks in a concurrent induce aa negative negative mood) than than after after success feedbacks feedbacks (designed (designed to induce induce aa to induce positive mood). Interestingly, however, there there was was no comparable comparable effect of of positive feedback on reaction reaction time time to to aa peripheral peripheral probe. Thus, consistent consistent with with the the feedback studies studies by Christianson Christianson et al. ((1991) earlier, a a negative mood 199 1) reviewed earlier, amplifies the the processing processing of of central central stimuli, but but not not peripheral peripheral ones 11.. amplifies Perceptual Perceptual Defense and Vigilance? Vigilance?
Def ense in perception Defense perception Traditionally, the notion notion of of Freudian Freudian defense was was postulated postulated to account account Traditionally, for the influence of of affect affect on perception perception (e.g., Blum 11954, f or the 954; Brunet, Bruner, 11957, 957; 11 Evidence Evidence indicates indicates that in a positive positive mood, mood, attention attention tends tends to be more more broadly broadlydistributed
space or semantic semantic region region (Derryberry (Derryberryet al., al., 1991; 1991;lsen, Isen, 11987; Isen & & Daubman, Daubman, 987; Isen in either space terms of the spotlight metaphor, a positive moodappears appears to "broaden" "broaden" the scope scope 11984). 984). In terms spotlight metaphor, positive mood attention. The influence influence of aff affect more structural structural aspect aspect of attention attention deserves deserves of attention. ect on this more further research 984). research (Tucker (Tucker & & Williamson, Williamson, 11984).
S.
Kitayama
239 239
Dixon, 11980; 980; Erdelyi, 11974; 974; McGinnies, 11947). 947). Unfortunately, however,
of defense have have never been specified enough to advance exact mechanisms of clear predictions. In contrast, the amplification model describes these effects as a result of an interaction between affective and cognitive pathways commonly involved in any ordinary processes of perceiving and thinking.
the amplification model model provides a coherent explanation As noted above, the for an otherwise anomalous pattern of the past studies conducted under the Freudian tradition. Furthermore, it successfully predicts the key outcomes of
experiments. The model can also comfortably accommodate the the present experiments. evidence that the influence of stimulus affect on immediate conscious perception is mostly the same for positive and negative affect (Broadbent &
Gregory, 11967; 1990, 199 1991) 967; Kitayama, 1990, 1)
- evidence incongruous with the of Freudian def defense. ense. It would seem that attentive processing can be notion of of an impinging stimulus has amplified once the significance or interest value of been detected through preattentive processing. Future research must now be of this preattentive computation of of the directed toward the precise nature of of the stimulus (cf. of significance of (cf. LeDoux, 11989). 989). All in all, although effects of identification task have traditionally been studied under affect in a perceptual identification of perceptual defense/vigilance, defense/vigilance, the present research suggests that the rubric of they may actually have nothing to do with Freudian defense. Instead, the lies primarily in the ffact act that they provide significance of these effects lies valuable clues about the functional organization of of unconscious processes commonly involved in various forms of of perceiving and thinking. -
Defense real life Def ense in real life Although the the notion of of defense is perhaps irrelevant in understanding the influence of processes in the present experimental influence of affect affect on perceptual processes paradigm, we also acknowledge that that a variety of of psychological defenses defenses can be identified identified in real life, especially in many clinical settings (e.g., Schwartz, 1990; 1 990; Weinberger, 1990). 1 990). Yet, possible mechanisms by which these defensive functions are implemented implemented have remained remained as ill-specified ill-specified as they were over 100 1 00 years ago when Freud Freud (1895/1966, ( 1 89511 966, p. 370) observed the following: Everything that the nervous system that I call a a biological acquisition of of the is in my opinion represented by aa threat threat of of unpleasure..., unpleasure . . . , the effect of of which consists in the fact that that those neurones which lead to to a a release of unpleasure are not cathected. This is primary defense. of unpleasure are not cathected. How to a threat How primary defense, non-cathexis owing to threat of of unpleasure,
240
Chapter Chapter 5 is to be represented mechanically - this, I confess, I am unable to say.
From From the present theoretical perspective, however, we may suggest that the interactive process between affect and cognition captured in the funetion amplification model is one mechanism that can serve a defensive function under certain circumstances in real life, although this by no means precludes any other processes that that might be implicated especially when affect involved is very intense. Specifically, an idea said said to be "conflictual" in a psychodynamic sense (of. Shevrin, 1i 990) may often be characterized by two features. First, such an (cf. idea may be associated assoeiateA with an extremely intense affective response. Second, the idea may be relatively isolated from more innocuous scripts or schemas commonly activated in everyday life. Under these conditions, the amplification model would predict that even when a "conflictual" idea is preconsciously activated and a strong affect evoked, attentive processing is likely to be allocated to a more accessible script or schema. As a result, the perception or recognition of of the innocuous script or schema may be magnified; but the original "conflictual" "conflictuar' idea will stay nonconscious. The mechanism specified by the present model, then, seems to serve the purpose of of psychological defense in situations in which the above two features validly characterize a "conflictual" "eonflietual" idea, even though the mechanism itself is not specifically designed for defense. Future Future Research Research Directions Directions
The amplification model has suggested a reasonable answer to the The specific agenda set for the current research, but it has also generated an array of of open issues and questions. In particular, assessing the generality of of the model must receive first priority in future research. To begin with, we must pose questions about different forms of of cognitive performance: Does the nature of of the cognitive task determine the influence of of affect? For example, will affect have any influence on lexical judgment and, if so, can the the nature of of the the influence be described by the amplification model? What What about categorization or some other tasks requiring more deliberate thinking or explicit memory retrieval? Many similar questions can be asked for affect as well: Does the influence of orm oof fa ffect being induced? Is the of affect depend on the exact f form affect amplification model generalizable to other forms of of affect? For example, will
S. S.
Kitayama
24 2411
affect influence influence cognition cognition differently differently depending depending on on whether whether it it is is directly directly affect attached to to aa stimulus stimulus object object (e.g., (e.g., the the negative negative affect affect usually usually attached attached to to attached representations of of snakes) snakes) or or whether, whether, on on the the other other hand, hand, it it is is induced induced by by aa representations certain "irrelevant," "irrelevant," "incidental" "incidental" stimulus stimulus (e.g., (e.g., the the negative negative affect affect commonly commonly certain elicited by by aa scratching scratching sound)? sound)? What What about about more more enduring enduring affective affective states, states, or or elicited the moods, moods, of of the the perceiver? perceiver? Alternatively, Alternatively, would would affect affect implicit implicit in in cognitive cognitive the representations (e.g., (e.g., the the affect affect associated associated with with the the notion notion of of ANGER) ANGER) behave behave representations in some some different different ways ways from, from, say, say, the the affect affect more more directly directly expressed expressed by by in patterns of of sensory sensory signals (e.g., vocal vocal expression expression of of hostility; of. Scherer, Scherer, patterns signals (e.g., hostility; cf. Before further further generality generality of of the the amplification amplification model model is is claimed, claimed, each each 11986)? 986)? Before one of of these these empirical empirical issues issues must must be be addressed. addressed. one In the the current current research research we we have have strategically strategically focused focused on on word word perception, perception, In because it it seemed seemed the the best best task task for for the the purpose purpose of of formulating formulating aa process processbecause oriented theory theory of of an an interaction interaction between between affect affect and and cognition. cognition. Future Future oriented research may may apply the theoretical theoretical processes processes captured captured in in the the amplification amplification research apply the model to to perception perception in in social domains, such such as as facial facial perception, perception, in in which which model social domains, preattentive processing processing of of aa stimulus stimulus is is known known to to play play critical critical roles roles (Damasio (Damasio preattentive 982; Hansen 990). To et al., al., 11982; Hansen & Hansen, Hansen, 1989; Niedenthal, Niedenthal, 11990). To illustrate illustrate one one possible possible extension extension along along this this line, line, faces faces of of one's one's own own race race tend tend to to be be more more familiar and and their their representations, representations, more more differentiated differentiated than than foreign foreign faces. faces. familiar Accordingly, faces faces of of one's one's own own race race should should yield yield aa greater greater number number of of Accordingly, constraints on on preconscious preconscious processing, processing, hence hence making making for for more more unequivocal unequivocal constraints activation of the the relevant representation. From From the word activation of relevant representation. the current current results results on on word length, one that one's to recognize faintly seen face length, one could could predict predict that one's ability ability to recognize aa faintly seen face would depend both on the emotional expression of the face and its race (one's would depend both on the emotional expression of the face and its race (one's own foreign race) is, emotional should be be more own race race or or a a foreign race) -- that that is, emotional faces faces should more perceptible and and more more easily easily identified identified than than neutral neutral faces faces if if these these faces faces are of perceptible are of one's own own race, race, but but not not if if they they are are foreign foreign faces. Future work work along along this this line line one's faces. Future can contribute to research (e.g., (e.g., Schneider, Schneider, Hastorf, Hastorf, & can contribute to basic basic social social perception perception research Ellsworth, 1979), well as more applied applied domains domains such such as eye-witness 1 979), as as well as more as eye-witness Ellsworth, Loftus, testimony (Brigham, (Brigham, Maass, Maass, Martinez, Martinez, & Whittenberger, Whittenberger, 1983; Loftus, testimony 1979). 1 979). The amplification more general the The amplification model model also also has has aa more general implication implication for for the current With aa few current social social cognition cognition literature. literature. With few notable notable exceptions exceptions (e.g., (e.g., Isen, Isen, 1987), 1 987), this this literature literature has has examined examined an an interaction interaction between between affect affect and and cognition cognition within within the the confine confine of of associative associative network network models models of of memory memory (e.g., (e.g., Bower, be argued 1 98 1 ; Lang, Lang, 1984). 1 984). Yet, it it may may be argued that that evocation evocation of of affect affect Bower, 1981; changes contents of changes not not only only the the contents of thought thought (as (as predicted predicted by by these these models), models), but but also 1 987; Tucker Tucker & also the the structure structure or or organization organization of of thought thought (e.g., (e.g., Isen, Isen, 1987;
242
Chapter 5
Williamson, that thinking Williamson, 1984). 1 984). Thus, Thus, it seems seems likely that thinking about a a negatively stereotyped ethnic ethnic group group not only raises negative connotations, but but also leads leads stereotyped to Burnstein, 1989). to more rigid and and narrower narrower thought thought processes processes (Kitayama & Bumstein, The The structure structure of of thought thought may may in turn turn depend on on certain functional characteristics of of attentive of the characteristics attentive processing. The present analysis of interdependence between between affect attention may serve as aa useful point of interdependence affect and and attention of departure for further elucidating some important questions and issues in, for example, stereotyping and and prejudice. prejudice.
Concluding Remarks paper has described a model of of affect-cognition affect-cognition In sum, the the present present paper interaction in perception, presented experimental evidence, and discussed a variety of the model. model. We We believe of empirical and theoretical issues raised by the that ect can participate in that the current current research research revives revives the possibility that that aff affect very early, early, preconscious preconscious stages of of processing. With sophisticated cognitive very (e.g., Bower & Clapper, 1989), 1 989), neuroanatomical (e.g., Posner & Peterson, 1 990), and electrophysiological (e.g., Cacioppo, Martzke, Petty, & 1990), and electrophysiologieal Tassinary, 1988; et al., 1 988; Davidson et aI., 1990) 1 990) techniques, experimental analysis of of 987, 11990). 990). This easible (Kihlstrom, 11987, nonconscious processes is now quite quite ffeasible ar analysis emphasizes that the role of of nonconscious processes may may be ffar greater greater than is routinely supposed in cognitive and social social psychology. psychology. The current work suggests that aff ect participates in these nonconscious affect processes processes.. By elucidating the role of ect in early perceptual processing, the of aff affect present research has supplied a new theoretical perspective to some old perceptual def defense subception. We showed problems such as perceptual ense and vigilance or subception. that these phenomena cannot be dismissed solely on methodological methodological grounds as was often done in the past. However, we also raised a serious doubt on previous theorizing that tended to link these phenomena to Freudian def ense. defense. We argued, instead, that that the theoretical significance of of the phenomena of of defense to reveal the def ense and vigilance actually lies primarily in their potential to functional structure of the nonconscious processes commonly involved in virtually every every form form of of perceiving perceiving and and thinking. thinking. that the present attempt to integrate aff affective We believe that ective phenomena within the current cognitive framework has the potential of of enriching this framework itself, insofar as the latter is presently essentially "affect-free." Furthermore, granted that aff ective involvement or aff ective charge is one affective affective 984; Markus & defining feature of of social cognition (cf. (of. Fiske & Taylor, 11984;
S.
Kitayama Ki tayama
243
Zajonc, 1984), aa process-oriented process-oriented model model of of affect-cognition affect-cognition interaction, interaction, such such Zajonc, as the the one one presented presented in in the the current current work, work, will be be aa necessary necessary theoretical theoretical as element with with which which social cognitive cognitive processes processes are are analyzed analyzed with with aa greater greater element rigor and and precision precision than than is possible possible today. The present present approach approach can can be be distinguished distinguished from from aa more more traditional traditional view view The of affect affect as as aa final final product product of of what what Lazarus Lazarus (e.g., (e.g., 1982; Smith Smith & Lazarus, Lazarus, of has called called the the cognitive cognitive appraisal. appraisal. Echoing Echoing this this view view of of affect affect as as aa 1990) has consequence of of cognition, cognition, many many social social psychological psychological studies studies have have described described aa consequence variety of of affective affective states states as as aa function function of of prior prior cognition cognition (e.g., (e.g., Anderson, Anderson, variety Smith & Ellsworth, Ellsworth, 1986; Hastie Hastie et et aI., al., 11981; Higgms, 1987; Weiner, Weiner, 11981; 98 1 ; Smith 98 1 ; Higgins, Although this this program program of of research research has has revealed revealed aa great great deal deal about about an an 1982). Although interaction between between affect affect and cognition, with with an an important important exception exception of the interaction and cognition, of the literature on on the the effects effects of of mood mood on on information information processing processing (e.g., (e.g., Blaney, Blaney, literature Bimbaum, 1990), a 11986; 986; Bower, 11981; 98 1 ; Clark, 1982; Isen, 1984; Salovey & Birnbaum, causal influence influence of of affect affect on on cognitive cognitive processes processes is is frequently frequently ignored ignored (see (see causal Zajonc, 1980). Zajonc, In contrast, the current empirical findings findings and and theoretical theoretical analysis analysis seem seem In contrast, the current empirical of affect-cognition interaction as recursive, to invite a more dynamic notion of mutual, and and beginning beginning very very early early in in stimulus stimulus processing. processing. The The view view of of mutual, dynamic interaction interaction between between affect affect and and cognition cognition developed developed and and defended defended in in dynamic the present present paper paper is is quite quite congenial congenial to to recent recent connectionist connectionist theorizing theorizing in in the cognitive science science (e.g., (e.g., Rumelhart, Rumelhart, 1989). More More generally, generally, it it can can be be seen seen as as cognitive one instantiation instantiation of of the the modularity hypothesis hypothesis of of mind mind (e.g., (e.g., Fodor, 1983; one Marr, 1976; Shallice, ShaUice, 11988; see also Buck, 1987; LeDoux, 1987, for similar 988; views as as applied applied to to emotion), emotion), which which postulates postulates any any given given psychological psychological views function, say, say, conscious conscious perception, perception, as as an an emergent emergent property property of of aa lawful lawfial function, organization modules") processes (or "processing modules") organization of of elementary elementary component component processes (or "processing such as as preattentive preattentive processing, processing, activation activation of of affective affective circuits, circuits, and and attentive attentive such processing. Further clarifying this mutual, dynamic dynamic interaction interaction between between affect affect processing. Further clarifying this mutual, and cognition, and then specifying its phylogenic precursors as well and cognition, and then specifying its phylogenic precursors as well as as patterns of patterns of ontogenic ontogenic development development in in different different social, social, cultural cultural environments, environments, presents a challenge for the future. References
Abelson, P. ((1963). 1 963). Computer Abdson, R. R. P. Computer simulation simulation of of "hot "hot cognitions. cognitions."" In In S. S. Tomkins Tomkins & S. S. Mesick Mesick (Eds.), (Eds.), Computer simulation of of personality. New New York: Wiley.
244
Chapter 5.5
of words we are unable to Allport, A. ((1977). 1 977). On knowing the meaning of report: The eff ects of effects of visual masking. In SS.. Domic Dornic (Ed.), Attention and performance FI. Hillsdale, Hillsdale, NJ: Erlbaum. perf ormance VI. Allport, A 1 989). Visual attention. In M. I. Posner (Ed.), Foundations A. ((1989). Foundaaons of of cognitive science. Cambridge, MA: MIT Press. Anderson, 1 980). Foundations f inf ormation integration Anderson, N. ((1980). Foundations oof information integration theory. theory. New York: Academic Press. 1 983). Automatic ssemantic-activation emantic-activation and episodic memory Balota, D. A. ((1983). encoding. Journal of of Verbal Verbal Learning and Verbal Verbal Behavior, Behavior, 22, 88104. Bargh, J. A. ((1989). 1 989). Conditional Conditional automaticity: automatieity: Varieties of of automatic influence in social perception and cognition. In J. S. Uleman & J. A A. (Eds.), Unintended thought. New York: Guilford Press. Bargh (Eds.), A, Bond, R 1 986). The Bargh, J. A., R. N., Lombardi, W. J., & Tota, M. E. ((1986). additive nature of of chronic and temporary sources of of construct accessibility. Journal o f Personality and Social Psychology, 50, 869of Psychology, .50, 878. R. M. ((1984). of names and faces in Bauer, R. 1984). Autonomic recognition of prosopagnosia: prosopagnosia: A neuropsychological neuropsychological application of of the guilty knowledge test. Neurophysiolog;a, Neurophysiologia, 22, 456-469. Blaney, P. H 1 986). Affect H. ((1986). Affect and memory: A review. review. Psychological Bulletin, 99, 229-246. G. S. ((1954). of psychoanalytic theory with Blum, G. 1 954). An experimental reunion of perceptual vigilance and defense. f Abnormal and Social defense. Journal oof Psychology, Psychology, 49, 94-98 94-98.. Bootzin, R 1 965). Evidence for perceptual def ense R.. R, R., & Natsoulas, T T.. ((1965). defense uncontaminated by response bias. Journal o f Personality and Social of Psychology, -468. Psychology, 5, 461 461-468. Bower, G. 198 1). Mood and 29G. H H. ((1981). and memory. American Psychologist, Psychologist, 36, 1129148 148.. Bower, G. H. 1 990). Awareness, H. ((1990). Awareness, the unconscious, unconscious, and repression: An An experimental experimental psychologist's perspective. In J. L Singer Singer (Ed.), Repression and dissociation: Implications f or personality for personality theory, theory, psychopathology, and health. health. Chicago: The University of of Chicago Press. Bower, G. H 989). Experimental methods in cognitive H.,, & Clapper, J. P. (1 (1989). science. science. In M. I. Posner (Ed.), Foundations Foundations of of cognitive cognitive science. science. Cambridge, MA: MIT Press. Braun, J., & Sagi, D. ((1990). of attention. Braun, 1 990). Vision outside the focus of Perception and Psychophysics, Psychophysics, 48, 45-58.
S. Kitayama Kitayama S.
245 245
Brigham, 1.J. C., C., Maass, Maass, A., A., Martinez, Martinez, D., D., & & Whittenberger, Whittenberger, G. G. ((1983). The Brigham, 1 983). The Basic and and Applied Applied Social Social effect of of arousal arousal on on facial facial recognition. recognition. Basic effect 279-293.. Psychology, 4, 4, 279-293 Psychology, Broadbent, D D.. EE.. ((1958). Perception and and communication. communication. London: London: Broadbent, 1 958). Perception Pergamon. Broadbent, 1 973). In def ense of Broadbent, D. D. E. E. ((1973). defense of empirical empirical psychology. psychology. London: London: The The Camelot Press. Broadbent, D. D. E., & & Gregory, M. M. ((1967). Perception of of emotionally emotionally toned toned Broadbent, 1 967). Perception Nature, 215, 58 581-584. 1-584. words. Nature, Bruner, 1. J. ((1957). readiness. Psychological Psychological Review, Review, 64, 64, 123123Bruner, 1 957). On perceptual readiness. 152. 152. Science, 1144, 424-425.. J., & Potter ((1964). Bruner, 1., 1 964). Science, 44, 424-425 Buck, R R. ((1985). Prime theory: theory: An integrated integrated view view of motivation motivation and Buck, 1 985). Prime Review, 92, 389-4 389-413. 13. emotion. Psychological Review, Buck, R R. ((1988). motivation and emotion emotion (2nd (2nd ed.). ed.). New York: York: Buck, 1 988). Human motivation Wiley. J. T, T., Martzke, 1. J. S., Petty, Petty, R R. E., & Tassinary, L. G. ((1988). Cacioppo, 1. 1 988). of facial EMG index index emotions during an interview: interview: From Specific forms of continuous flow flow hypothesis hypothesis of affect-laden affect-laden processing. Darwin to the continuous of Personality and Social Psychology, Psychology, 54, 592-604. Journal of Carr, 1 990). Semantic Carr, T T. H., & Dagenbach, Dagenbach, D. ((1990). Semantic priming priming and and repetition repetition priming words: Evidence Evidence for a center-surround attentional priming from masked words: mechanism in perceptual recognition. recognition. Journal of of Experimental Psychology: Learning, Learning, Memory, and Cognition, Cognition, 116, 6, 341-350. 34 1 -350. Carr, T. R. D., & Parmelee, T H., McCauley, C., Sperber, Sperber, R Parmelee, C. M. (1982). ( 1 982). Words, priming: On semantic activation, activation, conscious conscious Words, pictures, pictures, and and priming: On semantic identification, and the automaticity of of information processing. Journal of of
Experimental Psychology: Psychology: Human Human Perception Perception and Performance, Performance, 8, 757-777. Chapman, C. R., of perception by R, & Feather, Feather, B. W. (1972). ( 1 972). Modification Modification of classical conditioning procedure. Journal of of Experimental Psychology, 93, 338-342. s.-A., & Loftus, E. F. (1991). Remembering emotional events: Christianson, S.-A., and Emotion, 5, 81-108. The fate of 8 1 - 108. of detailed information. Cognition and Christianson, S.-A., o g a n , H., & Loftus, G. R. S. -A ., Lottus, Loftus, E. F., H Hoffman, R (1991). ( 199 1). Eye fixations and memory of Experimental memory for emotional emotional events. events. Journal of Psychology: Learning, Memory, and 1 7, 693-701. 693-70 1 . and Cognition, 17, ( 1 973). The language-as-fixed-effect fallacy. Journal of of Verbal Clark, H. H. (1973). Learning and Verbal Behavior, 12, 335-359. and
246
Chapter 5
Clark, 1982). Affect and cognition: 7th annual Clark, M. S., S., & Fiske, Fiske, SS.. T. T. ((1982). cognition: The 117th Carnegie symposium on cognition. Hillsdale, NJ: Erlbaum. Erlbaum. Corteen, B. ((1972). 1972). Autonomic Corteen, R R. S., S., & Wood, Wood, B. Autonomic responses responses to to shock shockassociated associated words words in in an an unattended unattended channel. channel. Journal of of Experimental Psychology, 94, 308-3 308-313. 13. Psychology, Crick, 1984). The Crick, F F.. ((1984). The function function of of the the thalamic thalamic reticular reticular complex: complex: The The search search light f the National Academy oof fSciences, light hypothesis. hypothesis. The Proceedings oof 81, 4586-4590. Damasio, R, & Damasio, H., & Van 1 982). Damasio, A. A. R., Damasio, H., Van Hoesen, Hoesen, G. G. W. W. ((1982). Prosopagnosia: Prosopagnosia: Anatomical Anatomical basis basis and and behavioral behavioral mechanisms. mechanisms. Neurology, 32, 33 1 -341 . 331-341. Davidson, 1 984). Affect, Davidson, R R. JJ.. ((1984). Affect, cognition, cognition, and and hemispheric hemispheric specialization. specialization. In In C. Emotion, cognition, cognition, and C. E. E. Izard, Izard, 1. J. Kagan, Kagan, & R R. Zajonc Zajone (Eds.), (Eds.), Emotion, York: Cambridge University Press. behavior. New New York: Cambridge University Press. Davidson, P., Saron, Davidson, R. R. 1., J., Ekman, Ekman, P., Saron, C. C. D., D., Senulis, Senulis, 1. J. A., A., & Friesen, Friesen, W. W. V. V. ((1990). 1990). Approach-withdrawal Approach-withdrawal and and cerebral cerebral asymmetry: asymmetry: Emotional Emotional expression brain physiology. f Personality expression and and brain physiology. Journal oof Personality and Social 1. Psychology, Psychology, 58, 330-34 330-341. Dawson, 982). Electrodermal Dawson, M M.. E., E., & Schell, Schell, A A. M. (1 (1982). Electrodermal responses responses to to attended attended f and and nonattended nonattended significant significant stimuli during dichotic dichotic listening. listening. Journal oof
Experimental Psychology: formance, 8, Psychology: Human Human Perception Perception and Per Performance, 3315-324. 15-324. Derryberry, 1., & Reed, 1 99 1). Motivational f focusing ocusing of of Derryberry, D., D., Brandt, Brandt, J., Reed, M. A A. ((1991). Unpublished manuscript, manuscript, Oregon Oregon State State University, University, Corvalis, Corvalis, attention. Unpublished OR. OR. Deutsch, J. A., & Deutsch, Deutsch, D. D. ((1963). Attention: Some Some theoretical theoretical 1 963). Attention: Deutsch, 1. A, considerations. Review, 70, 80-90. considerations. Psychological Review, Dixon, N. F. ((1980). 1 980). Preconscious processing. New York: Wiley. Dorfinan, D. D. ((1967). Recognition of of taboo taboo words words as as aa function function of of aa priori priori 1 967). Recognition Dorfinan, probability. ofPersonality and Social Psychology, 10. probability. Journal of Psychology, 7, 11-10. Dorfinan, 1 965). Recognition Dorfinan, D. D. D., D., Grossberg, Grossberg, J. J. M., & Kroeker, Kroeker, L. L. ((1965). Recognition of of taboo fPersonality and taboo stimuli stimuli as as aa function of of exposure exposure time. time. Journal oof Social Psychology, Psychology, 2, 552-562. Easterbrook, J. J. A. A. ((1959). The effect effect of of emotion emotion on on cue cue utilization utilization and and the the Easterbrook, 1959). The 83-20 1 . Review, 66, 1183-201. organization of of behavior. behavior. Psychological Review, Egeth, Egeth, H H.. (1977). Attention Attention and and preattention. preattention. In In G. G. H. H. Bower Bower (Ed.), The psychology oof motivation. New New York: York: Academic Academic Press. Press. flearning and motivation. look: Perceptual defense and Erdelyi, M. H. (1974). A new look at the new look: Review, 81 81,, 1-25. 1-25. vigilance. Psychological Review, vigilance.
S. S.
Kitayama
247
Erdelyi, 1 985). Psychoanalysis: Erdelyi, M. H H. ((1985). Psychoanalysis: Freud's Freud's cognitive cognitive psychology. psychology. New New York: W. H York: W. H. Freeman Freeman and and company. company. Erdelyi, M. H., Appelbaum, G. A. ((1973). Cognitive masking: masking: The The H, & Appelbaum, 1 973). Cognitive Erdelyi, of an emotional stimulus upon the perception of of disruption effect of 1. contiguous contiguous neutral items. Bulletin Bulletin of of the Psychonomic SOCiety, Society, J, 1, 59-6 59-61. Eriksen, C C.. W. W. ((1963). Perception and and personality. personality. In In 1J.. M M.. Wepman Wepman & R R. Eriksen, 1 963). Perception W. W. Heine Heine (Eds.), (Eds.), Concepts of of personality. Chicago: Chicago: Aldine. Aldine. 1 967). The biological Eysenck, biological basis basis of of personality. personality. Springfield: Springfield: Eysenck, H H. ((1967). Thomas. Thomas. Eysenck, M. W. W. ((1976). Arousal, learning, learning, and and memory. memory. Psychological Psychological 1 976). Arousal, Bulletin, Bulletin, 83, 389-404. Fazio, 1 986). Fazio, R. R. H., H., Sanbonmatsu, Sanbonmatsu, D. M., M., Powell, Powell, M. C., C., & Kardes, Kardes, F. R R. ((1986). On On automatic automatic activation activation of of attitudes. attitudes. Journal of of Personality and Social Psychology, Psychology, 50, 229-238 229-238.. Field, T T. M. M. ((1985). Neonatal perception perception of of people: people: Maturation Maturation and and Field, 1 985). Neonatal Social individual differences. differences. In In T T. M. M. Field Field & N. N. Fox Fox (Eds.), (Eds.), Social individual perception in inf ants. Norwood, infants. Norwood, NJ: Ablex. Ablex. Fiske, 1 980). Attention Fiske, SS.. T. T. ((1980). Attention and and weight weight in in person person perception: perception: The The impact impact of of negative of Personality Personality and Social Social negative and and extreme behavior. behavior. Journal of Psychology, Psychology, 38, 889-906. Fiske, S. T., Taylor, S. E. E. ((1984). cognition. Reading, Reading, MA: MA: Fiske, T, & Taylor, 1 984). Social cognition. Addison-Wesley. Addison-Wesley. of mind. mind Cambridge, MA: MIT Press. 1. A. ((1983). 1 983). The modularity of Fodor, J. Fowler, C. A., Wolford, G., Slade, R., R, & Tassinary, L. ((1981). 1 98 1 ). Lexical access with with and and without without awareness. awareness. Journal of of Experimental access 341-362. Psychology: General 1110, 1 0, 34 1-362. Psychology: General, J. T T. ((1954). defense. Journal of of Experimental Freeman, 1. 1 954). Set or perceptual defense. Psychology, 48, 282-288. Psychology, S.. ((1895/1966). J. Strachey (Ed. Freud, S 1 895/1 966). Project for scientific psychology. In 1. edition of of the complete complete psychological works and Trans.), The standard edition of Sigmund Freud (Vol. (Vol. 1). London: London: Hogarth Hogarth Press. Press. (Unpublished (Unpublished of manuscript, manuscript, originally originally entitled entitled "Psychology "Psychology for for neurologists," neurologists," completed completed in 11895). 895). emotions. Cambridge: Cambridge University Press. 1 986). The emotions. Frijda, N. ((1986). Goldiamond, I. ((1958). 1 958). Indicators Indicators of Goldiamond, I. of perception: perception: I.I. Subliminal Subliminal perception, perception, subception, subception, unconscious perception: An analysis analysis in in terms terms of of psychophysical indicator indicator methodology. methodology. Psychological Bulletin, psychophysical Bulletin, 55, 373411. 4 11.
248
Chapter 5
Green, 1966). Signal Detection Green, D. M., M., & Swets, Swets, J. ((1966). Detection Theory Theory and psychophysics. New New York: York: Wiley. Wiley. R, & Lui, 1 989). Unconscious Greenwald, Greenwald, A. G., G., Klinger, Klinger, M. R., Lui, T. T. J. ((1989). Unconscious Memory and Cognition, 7, processing of dichopticaIly masked words. Cognition, 117, processing of diehoptieally masked words. 35-47. Grier, J. (1971). Nonparametric indexes indexes for sensitivity and bias: Computing formulas. formulas. Psychological Bulletin, Bullean, 75, 424-429. Hansen, H, & Hansen, R D. ((1988). 1 988). Finding Hansen, C. H., Hansen, R. Finding the the face face in in the the crowd: crowd: An anger ofPersonality and Social Psychology, anger superiority superiority effect. effect. Journal of Psychology, 54, 9 1 7-924. 917-924. Hastie, R, Ostrom, Hastie, R., Ostrom, T. T. M., M., Ebbesen, Ebbesen, E. B., B., Wyer, Wyer, R R. S., S., Hamilton, Hamilton, D. L., L., & Carlston, f social Person memory: memory: The cognitive cognitive basis basis oof Carlston, D. E. (Eds.), (Eds.), Person perception. Hillsdale, NJ: Erlbaum. Erlbaum. 1 884). Uber Helmholtz, Vortrage und Helmholtz, H H. v. ((1884). Ober das Sehen Sehen des des Menschen. Menschen. Vortrage , 365-396. Reden, Reden, 11,365-396. Higgins, 1 987). Self-discrepancy: Higgins, E. E. T. ((1987). Self-discrepancy: A A theory theory of of relating relating self self and and affect. affect. Psychological ReView, 1 9-340. Review, 94, 3319-340. Higgins, E. T., & Bargh, J. A. ((1987). 1987). Social cognition and social perception. Annual Review o fPsychology, of Psychology, 38, 369- 425 425.. activation without conscious identification in Holender, D. ((1986). 1 986). Semantic activation listening, parafoveal vision, and visual masking: masking: A survey and dichotic listening, Sciences, 9, 11-66. appraisal. Behavioral and Brain SCiences, -66. Humphreys, 1 984). Personality, Humphreys, M. M. S., & Revelle, Revelle, W. W. ((1984). Personality, motivation, motivation, and and individual differences performance: A theory of the relationship between individual 91,, 1153-184. Review, 91 53-184. and information processing. Psychological Review, Isen, A. M. ((1984). of affect in cognition. cognition. In 1 984). Toward understanding the role of R. S. Wyer & T. K. SruU of social cognition cognition (Vol. R Srull (Eds.), Handbook of 3). Hillsdale, Hillsdale, NJ: Lawrence Lawrence Erlbaum. Erlbaum. Isen, A. M. ((1987). 1 987). Positive affect, cognitive processes, and social behavior. In L. Berkowitz Berkowitz (Ed.), (Ed.), Advances in Experimental Experimental Social Psychology (Vol. (Vol. 20). New New York: York: Academic Press. Press. influence of affect on Isen, A. M., & Daubman, K. A. ((1984). 1984). The influence categorization. categorization. Journal of of Personality and Social Psychology, Psychology, 47, 11206-1217. 206-12 17. Isen, A. M., Shalker, T. E., Clark, M., & Karp, L. ((1978). 1978). Affect, accessibility of material in memory, and behavior: behavior: A cognitive cognitive loop. accessibility Journal of ofPersonality and Social Psychology, Psychology, 36, 1-12. 1-12.
S.
Kitayama
249
Jacoby, L. L. ((1983). Perceptual enhancement: enhancement: Persistent Persistent effects effects of of an an Jacoby, 1 983). Perceptual experience. experience. Journal of of Experimental Psychology: Psychology: Learning, Learning, Memory, and Cognition, 9, 221-38. 1 -3 8 . Johnston, 1 985). Selective f Review oof Johnston, W. W. A., A., & Dark, Dark, V. 1. J. ((1985). Selective attention. attention. Annual Review Psychology, 3 7, 43-75 43-75.. Kahneman, 1 973). Attention and eff ort. Englewood Kahneman, D. ((1973). effort. Englewood Cliffs, Cliffs, NJ: Prentice PrenticeHall. Kakar, 1 978). The inner world: A psychoanalytic fchildhood and psychoanalytic study oof Kakar, S. ((1978). society in India. Delhi: Delhi: Oxford Oxford University Press. Press. 1 987). The Science, 238, 238, 1445-1452. Kihlstrom ((1987). The cognitive cognitive unconscious. unconscious. Science, Kihlstrom, 1 990). The Kihlstrom, 1. J. F. F. ((1990). The psychological psychological unconscious. unconscious. In In L. L. A. A. Pervin Pervin (Ed.), Handbook of (Ed.), of personality: Theory and research. research. New New YorkILondon: York~ondon: The The Guilford Guilford Press. Press. 1 989). Aff ect and cognition: requency of Kitayama, S. ((1989). Kitayama, S. Affect cognition: The role role of of f frequency of occurrence in the perception o f aff ectively toned words. of affectively words. Unpublished Unpublished manuscript, manuscript, University University of of Oregon, Oregon, Eugene, Eugene, OR. OR. Kitayama, SS.. (1990). Interaction between between affect affect and and cognition cognition in in word word Kitayama, ( 1 990). Interaction of Personality and Social Psychology, 58, 209-2 209-217. perception. Journal of 17. Kitayama, 1 99 1). Impairment Kitayama, S. S. ((1991). Impairment of of perception perception by by positive positive and and negative negative Emotion, 5, 255-274. affect. Cognition and Emotion, of emotional emotional speech: speech: Improvement and Kitayama, S. ((1996). 1 996). Remembrance of of impairment of of incidental incidental verbal memory by by emotional voice. Journal of impairment emotional voice. Experimental Social Psychology, 32, 289-308. Experimental Bumstein, E. ((1989). between opinion opinion and Kitayama, S., & Burnstein, 1 989). The relation between memory: Distinguishing between between associative density and structural structural memory: J. Bassili (Ed.), (Ed.), On-line centrality. In 1. On-line cognition in person perception. Hillsdale, NJ: Erlbaum. Kitayama, S., S., & Howard, Howard, SS.. ((1994). 1994). Affective Affective regulation regulation of of perception perception and and comprehension. In In P. P. M. Niedenthal Niedenthal & S. S. Kitayama (Eds.), (Eds.), The heart' heart'ss comprehension. eye: Emotional influences in perception and attention. attention. New York: York: eye: Academic Press. Kleinsmith, L. 1J.,. , & Kaplan, S. ((1963). Paired-associate learning learning as a 1963). Paired-associate Kleinsmith, function of of arousal arousal and and interpolated interpolated interval. interval. Journal of of Experimental function Psychology, 65, 1190-193. 90- 1 93 . Kucera, 1 967). Computational analysis of Kucera, H., H., & Francis, Francis, W. W. N N.. ((1967). of present presentday American English. Brown University Press. Kunst-Wilson, W. W. R., R., & Zajonc, Zajonc, R. R. B. B. ((1980). Affective discrimination discrimination of of 1 980). Affective Kunst-Wilson, 207, 557-558. stimuli that cannot be recognized. Science, Science, 207,
250
Chapter 5
LaBerge, D. ((1983). 1 983). Spatial extent of of attention to letters and words. Journal o ormance, 9, off Experimental Experimental Psychology: Psychology: Human Perception and Perf Performance, 371-379. LaBerge, D., & Brown, V. ((1989). 1989). Theory of of attentional operations operations in shape identification. Psychological 0 1 - 1 24. Psychological Review, Review, 96, 1101-124. LaBerge, D., & Samuels, 1. 973). Toward a theory of J. (1 (1973). of automatic information information processing in reading. Cognitive Psychology, 6, 292-323 292-323.. Lacey, J. 1. I., & Lacey, 1 968). Verification and extension of the Laeey, B. C C.. ((1968). principle of f of autonomic response sterotypy. American Journal oof Psychology, 71, 50-73. W., Lewinger, N., & Adamson, 1. Lacey, O. O. W., J. E. (1953). Foreknowledge as a factor affecting perceptual defense and alertness. Journal of of 69-1 74. Experimental Experimental Psychology, 45, 1169-174. 1 984). Cognition in emotion: emotion: Concept and action. In C. Izard, 1. Lang, P. ((1984). J. Kagan, & R. B. Zajonc Zajone (Eds.), Emotion, Emotion, cognition, cognition, and behavior. behavior. New York: Cambridge. 1 990). Emotion, attention, and Lang, P., Bradley, M. M., & Cuthbert, B. N. ((1990). the startle reflex. Psychological Psychological ReView, Review, 97, 377-395 377-395.. Lawrence, D. H., & Cole, G. R. (1954). Accuracy of of recognition with f Experimental alternatives before and after the stimulus. Journal oof Experimental 14. Psychology, 47, 208-2 208-214. Lazarus, R. S. ((1982). 1 982). Thoughts on the relations between emotion and cognition. American Psychologist, 3 7, 1019-1024. MeCleary, R. A. ((1951). discrimination Lazarus, R. S., & McCleary, 1 95 1). Autonomic discrimination without awareness: A study of of subception. Psychological Review, Review, 58, 1113-122. 1 3- 1 22. of Physiology, LeDoux, 1. J. E. ((1987). 1 987). Emotion. In F. Plum (Ed.), Handbook o f PhYSiology, leDoux,
Section 1. I. The nervous system. system. Vol. Vol. V: Higher functions of of the brain. Bethesda: American Physiological Society. LeDoux, 1. J. E. ((1989). interactions in the brain. leDoux, 1 989). Cognitive-emotional interactions Cognition and Emotion, 3, 267-289. 267-289. information processing. New York: Lewicki, P. ((1986). 1986). Nonconscious social inf ormation processing. Academic Press. Lindsley, D. B. ((1951). 1 95 1). Emotion. In S f S.. SS.. Stevens (Ed.), Handbook oof experimental psychology. psychology. New York: Wiley. Loffus, E. F. ((1979). testimony. Cambridge, MA: Harvard Loftus, 1979). Eyewitness testimony. University Press. McGinnies, E. M. ((1949). defense. McGinnies, 1 949). Emotionality and perceptual defense. Psychological 1 -482. Psychological Review, 56, 47 471-482.
S. S.
Kitayama Ki tayama
2511 25
MacLeod, C., Mathews, A., A, & Tata, P. ((1986). 1 986). Attentional bias in emotional of Psychology, 95, 115-20. disorder. Journal o fAbnormal Psychology, 5-20. Marcel, A 1. ((1983a). l983a). Conscious and unconscious perception: Experiments A. J. on visual masking and word recognition. Cognitive Cogmtive Psychology, Psychology, 15, 197-237. Marcel, A (l983b). Conscious and unconscious perception: An approach A. 1. J. (1983b). to the relation between phenomenal experience and perceptual processes. Cognitive Psychology, Psychology, 15, 238-300. Markus, H. R., & Kitayama, S. ((1991). 199 1). Culture and the self: self: Implications for cognition, emotion, and and motivation. Psychological Review, Review, 98, 224-253 224-253.. Markus, H . , & Zajonc, R 1 985). The cognitive perspective in social Markus, H., R.. B. ((1985). psychology. In G. Lindzey, & E. Aronson (Eds.), Handbook o f social of , Vol. 2). New York: Random House. psychology (3rd ed. r Marr, 1 976). Early processing of Marr, D. ((1976). of visual information. Philosophical Transactions of fLondon G, of the Royal Society Society oof G, 275, 275, 483-524. Mathews, A., A, & Wertheimer, M. (1958). A "pure" measure of of perceptual defense uncontaminated uncontaminated by response suppression. Journal of of Abnormal Psychology, 57, 373-376. and Social Psychology, Matthews, G., Pitcaithly, D., & Mann, R. E. ((1995). 1 995). Mood, neuroticism, and Cognitive Therapy Therapy and Research, Research, 119, of affective words. Cognitive the encoding of 9, 563-587. Perception and Merikle, P. M. ((1982). 1 982). Unconscious processing revisited. Perception Psychophysics, 31, 31,298-301. Psychophysics, 298-30 1 . J. G. ((1965). of "perceptual defense": A Minard, 1. 1 965). Response-bias interpretation of selective review and and an evaluation of of recent research. Psychological selective 74-88.. Review, 72, 74-88 Review, McClelland, 1. J. L L.,. , & Rumelhart, D. E E.. ((1981). 1 9 8 1 ) . An interactive activation model of of context effects in letter perception: Part 11.. An account of of basic Review, 88, 375-407. findings. Psychological ReView, Moray, N. ((1959). 1 959). Attention in dichotic listening: Affective cues and the infl uence of Quarterly Journal oof f Experimental Experimental influence of instructions. Quarterly Psychology, 111, Psychology, 1 , 56-60. Mozer, M. C. ((1988). 1 988). A connectionist model of of selective attention in visual f the Annual Meeting oof fthe Cognitive of Cognitive Science Science perception. Proceedings o Society, 1195-201. SOCiety, 95-20 1 . Murphy, S ective primacy and affi ctive Affecave affective S.. T., & Zajonc, R. B. (1990). Aff priming of erences. Unpublished manuscript, The University of of pref preferences. of Michigan.
252
Chapter 5
((1965). 1 965). Converging operations f or perceptual def ense. for defense. Psychological Bulletin, Bulletin, 64, 393-40 393-401.1 . 1984). Resources - a theoretical Navon, D D.. ((1984). th~retical soupstone. soupstonr Psychological Psychological Review, Review, 91, 2216-234. 1 6-234. Neely, 1977). Semantic N~ly, 1. J. ((1977). Somantie priming and retrieval from lexical loxical memory: Roles Natsoulas,
T.
of of inhibitionless inhibitionloss spreading activation and limited-capacity attention. Journal of of Experimental Psychology: Psychology: General, General, 106, 106, 226-254. Newell, 1 989). Symbolic architectures Newr A, A., Rosenbaum, P. SS.,. , & Laird, 1. J. E. ((1989). or cognitive for cognition. In M. I. Posner (Ed.), Foundations Foundaaonsf for cognitive science. science. Cambridge, MA: MIT Press. 990). Implicit perception of ective information. Niedenthal, Niedenthal, P. M. (1 (1990). of aff affoctive Journal of of Experimental Social Psychology, Psychology, 26, 505-527. 1968). Toward a theory of No rman, D. A Norman, A. ((1968). of memory and attention. Psychological ReView, Review, 75, 522-536. 522-536. Ohman, 1 985). Face the beast and ffear ear the face: Animal and social fears Ohman, A A. ((1985). as prototypes f or evolutionary analyses of Psychophysiology, for of emotion. emotion. Psychophysiology, 23, 1123-145. 23-145. of affective affectivr meaning systems. Osgood, C. E. ((1962). 1962). Studies on the generality of American Psychologist, 7, 10-28. Psychologist, 117, Paulhus, D., & Levit, 1987). Desirable responding triggered by affect: Lovit, K. ((1987). Automatic egotism? Journal of of Personality Personality and Social Psychology, Psychology, 52, 245-259. Norman, A. ((1964). nonparametrie analysis of of recognition Pollak, I., & No rman, D. A 1964). A nonparametric Psychonomic Science, Science, 11,, 125-126. experiments. Psychonomic Posner, M. I. ((1978). Chronometric explorations explorations oof mind. New York: Posner, 1 978). Chronometric f mind. Oxford University Press. Oxf ord University of attention. Quarterly Quarterly Journal Journal oof 1 980). Orienting of f Posner, M. I. ((1980). Experimental Psychology, 32, 3-25 . Psychology, 3-25. Posner, 1989). Foundations Posner, M. I. (Ed.) ((1989). Foundaaons of of cognitive cognitive science. science. Cambridge, MA: MIT Press. Petorson, SS.. E. ((1990). The attention system of of the human 1 990). The Posner, M. I., & Peterson, Review of of Neuroscience, Neuroscience, 13, 25-42. brain. Annual Review Posner, M. I., Peterson, 1 988). Potorson, S. E., Fox, P. T., & Raichie, Raichle, M. E. ((1988). Science, 240, 240, of cognitive operations operations in the human brain. SCience, Localization of 1627-173 1. 1627-1731. Posnor, M. I., & Snyder, C C.. R. R. ((1975). Posner, 1 975). Attention and cognitive control. In ormation processing R. Solso (Ed.), Inf Informaaon processing and cognition: cognition: The Loyola Loyola S ymposium. Hillsdale, HiUsdalr NJ: Erlbaum. Symposium.
S.
Kitayama Ki tayama
2 53 253
Postman, L., Bronson, W. C., c., & Gropper, G. L. ((1953). 1 95 3). Is there a mechanism f Abnormal and Social mechanism of of perceptual defense? Journal oof 15-224. Psychology, 48, 2215-224. Pratto, F., & John, O.P. ((1991) 1 99 1 ) Automatic vigilance: The attention-grabbing power of negative social information. Journal of of of Personality and Social Psychology, 61, 3380-391. 80-39 l . Rapaport, 1 967). The theory of Rapaport, D D.. ((1967). of attention cathexis: cathcxis: An economic and structural attempt at the explanation of of cognitive processes. In M M.. M M.. Gill (Ed.), The collected papers of of David Rapaport. New York: Basic Books. Rcvcllc, W., & Loftus, LoRus, D. A. ((1990). Revelle, 1990). Individual differences and arousal: Implications for the study of of mood and memory. Cognition and Emotion, 4, 209-237 209-237.. Roland, A 1 988). In search o f self A.. ((1988). of self in India and Japan: Toward a cross crosscultural psychology. Princeton: Princeton University Press. Roscman, I. ((1984). of emotion: A structural theory. Roseman, 1 984). Cognitive determinants of In P. Shaver (Ed.), Review o f personality and social psychology: of emotions, relationships, and health. Beverly Hills, CA: Sage. Rumclhart, D. E. ((1989). of mind: A connectionist conncctiomst Rumelhart, 1 989). The architecture of approach. In M M.. I. Posner Posncr (Ed.), (Ed.), Foundations of of cognitive cogmave science. Cambridge, MA: MIT Press. Russell, 1. ( 1 980). A circumplex f Personality J. A. (1980). circumplcx model of of affect. Journal oof 16 1- 1 1 78 . and Social Psychology, 39, 11161-1178. Sagi, D 1 985). "Where" and "what" in vision. Science, 228, D.,. , & Julesz, Julcsz, B. ((1985). 1217-1219. 12 1 7- 1 2 1 9. 1 968). A different look at perceptual defense for Sales, B. D., & Haber, Habcr, R. ((1968). taboo 56- 1160. taboo words. Perception & Psychophysics, 3, 115660. Salovey, 1 990). Influence of Salovcy, P., & Birnbaum Bimbaum ((1990). of mood on health-relevant cognitions. Journal of 7, 539-55 1. of Personality and Social Psychology, 557, 539-551. Schachter, 1 962). Cognitive, social and physiological Schachtcr, SS.,. , & Singer, 1. J. E. ((1962). Review, 69, 379-399. determinants determinants of of emotional state. Psychological Review, Scherer, 1 984). On the nature and function of Schcrcr, K. R. ((1984). of emotions: A component process approach. In K. R. Schcrcr Scherer & P. Ekman (Eds.), Approaches to emotion. Hillsdale, Hillsdalc, NJ: Erlbaum. 986). Vocal affect expression: A review and a model for Scherer, Schcrcr, K. R. (1 (1986). 43-1 65 . future research. Psychological Bulletin, 99, 1143-165. Schneider, D 1979). Person D.. 1., J., Hastorf, A A.. H., & Ellsworth, EUsworth, PP.. C. ((1979). perception (2nd ed.). exi.). Reading, MA: Addison-Wesley.
254
Chapter 5
Schneider, W. ((1988). Micro experimental experimental laboratory: laboratory: An integrated system 1 988). Micro for IBM PC compatibles. compatibles. Behavior Research Methods, Instruments, Instruments, & Computers, 20, 206-207. Schwartz, G. E. (1990). ( 1990). Psychobiology Psychobiology of repression and health: health: A systems systems approach. In J. 1. L Singer Singer (Ed.), (Ed.), Repression and dissociation: or personality theory, Implications f for theory, psychopathology, and health. Chicago: The University University of Chicago Press. Shallice, T. ((1988). 1 988). From neuropsychology neuropsychology to mental structure. structure. Cambridge, Cambridge, England: Cambridge University Press. England: Shevrin, 1972). Average Shevrin, H., Smith, Smith, W. H., & Fritzler, Fritzler, D. E. ((1972). Average evoked response response and verbal verbal correlates correlates of unconscious unconscious mental mental processes. processes. Psychophysiology, 8, 1149-162. 49-1 62. Shevrin, 1 990). Subliminal Shevrin, H. ((1990). Subliminal perception perception and repression. repression. In 1. J. L Singer Singer (Ed.), Repression and dissociation: Implications or personality theory, ImpBcatiom f for Chicago psychopathology, psychopathology, and health. health. Chicago: Chicago: The University University of Chicago Press. Smith, c., 1987). Patterns of appraisal and emotion C., & Ellsworth, Ellsworth, P. C. ((1987). emotion related f Personality related to taking an exam exam.. Journal oof Personality and Social Psychology, 52, 475-488. azarus, R. S. ((1990). 1 990). Emotion Lazarus, Emotion and adaptation. In L. A. Pervin Smith, C., & L personality: (Ed.), Handbook of of personality: Theory and research. research. New York: York: The Guilford Press. Spence, K. W. ((1956). 1 956). Behavior theory and and conditioning. New Haven, Haven, CT: Yale University University Press. f psychophysiology. Sternbach, 1966). Principles o Stembach, R. A. ((1966). of psychophysiology. New York: York: Academic Press. 1. T. Tassinary, L. G., Orr, SS.. P., Wolford, Wolford, G., Napps, S. E., & Lanzetta Lanzetta,, J. (1 984). The role (1984). role of of awareness awareness in affective affective information information processing: processing: An exploration of the Zajonc hypothesis. f the Psychonomic hypothesis. Bulletin oof 1. SOCiety, Society, 22, 489-49 489-491. Taylor, S 199 1). Asymmetrical S.. E. ((1991). Asynunetrieal effects effects of positive positive and negative negative events: events: The mobilization-minimization mobilization-minimization hypothesis. J 0, hypothesis. Psychological Psychological Bulletin, Bulletin, J110, 67-85 . 67-85. Tomkins, 1962). Affect, imagery and consciousness. Tomldns, ((1962). consciousness. Vol. Vol. J1:: The positive posiave a ff ects. New York: Springer. affects. York: Tomkins, (1 980). Affect as amplification: (1980). amplification: Some Some modification modification in theory. theory. In R. Plutchik rman (Eds.), Plutchik & H. Kelle Kellerman (Eds.), Emotion: Theory, Theory, research, research, and experience. New York: York: Academic Academic Press.
S.
Kitayama
255 255
Tranel, 1 985). Knowledge Tranel, D., D., & & Damasio, Damasio, A. A. R. R. ((1985). Knowledge without without awareness: awareness: An autonomic Science, 228, 228, autonomic index index of of facial facial recognition recognition by by prosopagnosics. prosopagnosics. Science, 14531 454. 1453-1454. Treisman, A. A. M. ((1964). Verbal cues, cues, language, language, and and meaning meaning in in selective selective Treisman, 1964). Verbal attention. fPsychology, 19. Psychology, 77, 206-2 206-219. attention. American Journal oof Treisman, A. A. M. M. ((1988). Features and and objects: objects: The The fourteenth fourteenth Bartlett Bartlett Treisman, 1988). Features Memorial Quarterly Journal of of Experimental Experimental Psychology, Psychology, Memorial Lecture. Lecture. Quarterly 40A, -237. 201-237. 40,4, 201 1984). Asymmetric Tucker, D. M., & & Williamson, Williamson, P. P. A. A. ((1984). Asymmetric neural neural control control 85-2 1 5 . systems Review, 91, 1185-215. systems in in human human self-regulation. self-regulation. Psychological ReView, Uleman, 1 989). Unintended Uleman, 1. J. SS.,. , & & Bargh, Bargh, 1. J. A A.. ((1989). Unintended thought. thought. New New York: York: Guilford Guilford Press. Press. Van 1 968). Repression sensitization: Sensitivity Van Egeren, Egeren, L. ((1968). Repression and and sensitization: Sensitivity and and recognition of Experimental Research Research in Personality, recognition criteria. Journal of -8 . 3, 11-8. Walker, 1 958). Action Walker, E E.. L L.. ((1958). Action decrement decrement and and its its relation relation to to learning. learning. Psychological Review, 29-1 42. Review, 65, 1129-142. Walker, 1 963). Memory Walker, E. E. L., L., & & Tarte, Tarte, R. R. D. D. ((1963). Memory storage storage as as aa function function of of arousal f arousal and and time time with with homogeneous homogeneous and and heterogeneous heterogeneous lists. lists. Journal oof Verbal Learning and Verbal Behavior, 90-1 93 . Behavior, 71 71,, 1190-193. Weinberger, D. A A.. ((1990). The construct construct validity validity of of the the repressive repressive coping coping Weinberger, 1 990). The style. Singer (Ed.), Repression and dissociation: Implications style. In J. L L Singer (Ed.), Repression Chicago: The The for theory, psychopathology, psychopathology, and health. health. Chicago: f or personality personality theory, University of of Chicago Press. of causal ascriptions. In M. Weiner, B. ((1982). 1982). The emotional consequences of S ect and cognition. S.. Clark Clark. & & S. S. T. T. Fiske Fiske (Eds.), (Eds.), Aff Affect cognitmn. Hillsdale, Hillsdale, NJ: NJ: Erlbaum. Wike, E. L., & & Church, Church, J. Comments on on Clark's Clark's "The "The language language1. D. ((1976). 1 976). Comments & Verbal Verbal Learning, as-fixed-effect fallacy." faUaey." Journal oof f Verbal Behavior & as-fixed-effect 15, 249-255. Worthington, 1 969). Paired Worthington, A. G. ((1969). Paired comparison comparison scaling sealing of of brightness brightness judgments" A A method method for for the the measurement measurement of of perceptual perceptual defense. defense. British judgments: Journal of 363-368. of Psychology, 60, 363-368. R. M., & & Dodson, Dodson, 1. J. D. (1908). The relation relation of of strength strength of of stimulus stimulus ( 1 908). The Yerkes, R. f Comparative to rapidity rapidity of of habit habit formation. formation. Journal oof Comparative Neurology and to Psychology, 18, 459-482. Psychology, of Zajonc, R. R. B. B. ((1962). Response suppression suppression in in perceptual perceptual defense. defense. Journal of Zajonc, 1 962). Response Psychology, 64, 206-2 206-214. Experimental Psychology, 14 . Zajonc, R R.. B. B. ((1965). Social facilitation. facilitation. Science, Science, 149, 149, 269-274. 269-274. Zajonc, 1 965). Social
256 256
Chapter 5
Preferences need no inferences. Zajonc, R. B. ((1980). 1 980). Feeling and thinking: Preferences American Psychologist, Psychologist, 35, 1151-175. 5 1 -1 75 . Zajonc, R. B., & Dorfman, 1 964). Perception, drive, and behavior Dorfinan, D D.. D. ((1964). theory. Psychological Review, Review, 71, 273-290. 71,273-290. Zajonc, R. B., Murphy, S. T., & Ingelhart, M. ((1989). 1 989). Feeling and facial eff'erence: efference: Implications of of the vascular theory of of emotion. Psychological Review, 1 6. Review, 96, 395-4 395-416. Author Author Notes Notes
This research was supported by National Institute of of Mental Health Grant l1R01 RO l MH501 17-01 . This chapter was completed while the author was MH50117-01. a fellow at the Center for Advanced Study in the Behavioral Sciences.
S.
257 257
Kitayama Kitayama
Appendix A.. Word Appendix A W o r d pairs p a i r s used u s e d in in Experiment E x p e r i m e n t 11.. Long words words Positive Positive
FASCINATING FASCINATING -- OUTSTANDING OUTSTANDING ATTRACTIVE -- SATISFYING ATTRACTIVE SATISFYING EXCELLENCE EXCELLENCE -- DELIGHTFUL DELIGHTFUL FANTASTIC FANTASTIC -- WONDERFUL WONDERFUL HAPPINESS HAPPINESS -- HONEYMOON HONEYMOON
COMFORTABLE -- MAGNIFICENT COMFORTABLE MAGNIFICENT D E D I C A T I O N -- ENTHUSIASM ENTHUSIASM DEDICATION FAVORABLE F A V O R A B L E -- PROMINENT PROMINENT BRILLIANT B R I L L I A N T -- DESIRABLE DESIRABLE �VELOUS MARVELOUS -- INTEGRITY INTEGRITY
Negative Negative
DESTRUCTION D E S T R U C T I O N -- THREATENING THREATENING INADEQUATE INADEQUATE -- CONSPIRACY CONSPIRACY ISOLATION CRITICISM ISOLATION -- CRITICISM SUFFERING S U F F E R I N G -- DESTROYED DESTROYED TREMBLING EXECUTION T R E M B L I N G - EXECUTION VIOLENCE V I O L E N C E -- WEAKNESS WEAKNESS
ALIENATION A L I E N A T I O N -- DISTURBING DISTURBING
CONCEPTION C O N C E P T I O N -- IMPRESSION IMPRESSION MECHANISM MECHANISM -- APPARATUS APPARATUS A L TERNA T E R N ATIVE T I V E -- MEASUREMENT AL MEASUREMENT APPLICABLE INHABITATANT APPLICABLE -- IN ITATANT RESIDENTS -- FORMATION RESIDENTS FORMATION
INTENTIONS INTENTIONS -- OBmCTIVES OBJECTIVES DIMENSION D I M E N S I O N -- MAGNTI1JDE MAGNITUDE PROBABILITY P R O B A B I L I T Y -- TRANSLATION TRANSLATION ARRANGEMENT -- MANUFACTURE MANUFACTURE ARRANGEMENT VARIABLES -- PERTINENT PERTINENT VARIABLES
Neutral Neutral
RESENTMENT RESENTMENT -- CORRUPTION CORRUPTION CONDEMNED PRISONERS CONDEMNED -- PRISONERS DANGEROUS EMERGENCY DANGEROUS -- EMERGENCY REVOLUTION R E V O L U T I O N - SLAUGHTER SLAUGHTER
Short words words Short Positive Positive
C O M E D Y -- WISDOM WISDOM COMEDY SMILE -- CHARM SMILE CHARM G L O R Y -- PRIZE GLORY PRIZE
P R O U D -- EAGER EAGER PROUD
L U C K Y -- HUMOR HUMOR LUCKY T R U S T -- ENJOY ENJOY TRUST JOKE CASH JOKE -- CASH
Negative Negative
C A N C E R -- INJURY INJURY CANCER D E V I L -- PANIC PANIC DEVIL FAULT -- GUILT FAULT GUILT SNAKE -- CRIME CRIME SNAKE
DAMAGE -- HATRED HATRED DAMAGE ERROR ERROR -- ANGRY ANGRY WASTE -- STORM STORM WASTE H U N G -- UGLY UGLY HUNG
V I C T I M -- TERROR TERROR VICTIM F A L S E -- BLAME BLAME FALSE F O O L -- PALE PALE FOOL SHAME -- WORSE WORSE SHAME
Neutral Neutral
B O R D E R -- �GIN MARGIN BORDER T R A C K -- ROUTE ROUTE TRACK
S T A M P -- LABEL LABEL STAMP SWITCH -- CUSTOM SWITCH CUSTOM
HABIT TREND HABIT -- TREND
LOCATE -- DETECT LOCATE DETECT S P A R E -- EXTRA EXTRA SPARE C H A I R -- TRACE TRACE CHAIR
SHEET -- FENCE SHEET FENCE
TREND PAUSE PAUSE -- TREND
T A L E N T -- MATURE MATURE TALENT FUNNY AWARD FUNNY -- AWARD
STONE S T O N E -- PANEL PANEL W I R E -- BONE BONE WIRE
258 258
Chapter Chapter 5 5
A p p e n d i x B. W o r d pairs p a i r s used u s e d in in Experiment E x p e r i m e n t 11.. Appendix B. Word
Long words Positive Positive -Neutral Neutral
FASCINAIqNG -- ALTERNATIVE ALTERNATIVE FASCINATING COMFORTABLE COMFORTABLE -- RESIDENTIAL RESIDENTIAL A T T R A C T I V E -- MECHANICAL MECHANICAL ATIRACTIVE DEDICATION DEDICATION -- OBJECTIVES OBJECTIVES EXCELLENCE EXCELLENCE -- APPLICABLE APPLICABLE FAVORABLE FAVORABLE -- SCHEDULED SCHEDULED FANTASTIC FANTASTIC -- DIMENSION DIMENSION B R I L L I A N T -- RECORDING RECORDING BRILLIANT H A P P I N E S S -- V VARIABLES HAPPINESS ARlABLES MARVELOUS M A R V E L O U S -- PERTINENT PERTINENT
OUTSTANDING OUTSTANDING -- REPRESENTED REPRESENTED MAGNIFICENT M A G N I H C E N T -- SURROUNDING SURROUNDING SATISFYING -- OCCASIONAL OCCASIONAL SATISFYING ENTIIUSIASM E N T H U S I A S M -- INTENTIONS INTENTIONS DELIGHTFUL -- INHABITANT INHABITANT DELIGHfFUL HAPPINESS HAPPINESS -- ASSEMBLED ASSEMBLED W O N D E R F U L -- MAGNITUDE MAGNITUDE WONDERFUL D E S I R A B L E -- FORMATION FORMATION DESIRABLE H O N E Y M O O N -- COMPONENT COMPONENT HONEYMOON I N T E G R I T Y -- RESIDENTS RESIDENTS INTEGRITY
Negative -N e g a t i v e N e u t r a l Neutral
DESTRUCTION -- ALIERNATIVE ALTERNATIVE DESTRUCTION ALIENATION -- OBJECTIVES OBJECTIVES ALIENATION I N A D E Q U A T E -- MECHANICAL INADEQUATE MECHANICAL RESENTMENT -- APPLICABLE APPLICABLE RESENTMENT DESTRUCTION D E S T R U C T I O N -- RESIDENTIAL RESIDENTIAL ISOLATION -- ASSEMBLED ISOLATION ASSEMBLED C O N D E M N E D -- SCHEDULED SCHEDULED CONDEMNED SUFFERING S U F F E R I N G -- FORMATION FORMATION DANGEROUS PERTINENT DANGEROUS -- PERTINENT TREMBLING RECORDING T R E M B L I N G - RECORDING
T H R E A T E N I N G -- REPRESENTED REPRESENTED THREATENING DISTURBING OCCASIONAL DISTURBING -- OCCASIONAL C O N S P I R A C Y -- INTENTIONS INTENTIONS CONSPIRACY CORRUPTION C O R R U P T I O N -- SURROUNDING SURROUNDING S L A U G H T E R -- MAGNITUDE MAGNHZIDE SLAUGHIER CRITICISM C R I T I C I S M -- DIMENSION DIMENSION PRISONERS PRISONERS -- RESIDENTS RESIDENTS DESTROYED ARlABLES D E S T R O Y E D -- V VARIABLES EMERGENCY E M E R G E N C Y -- COMPONENT COMPONENT EXECUTION E X E C U T I O N -- INHABITANT INHABITANT
Short words Positive Positive -Neutral Neutral
COMEDY C O M E D Y -- BORDER BORDER MATURE MATURE -- DETECT DETECT SMILE S M I L E -- TRACK TRACK AWARD AWARD -- STONE STONE GLORY GLORY -- CHAIR CHAIR EAGER EAGER -- PANEL PANEL
WISDOM W I S D O M -- MARGIN MARGIN LUCKY L U C K Y -- STAMP STAMP CHARM C H A R M -- ROUIE ROWI~ TRUST TRUST -- HABIT HABIT PRIZE P R I Z E -- TRACE TRACE JOKE J O K E -- WIRE WIRE
TALENT T A L E N T -- LOCAIE LOCATE HUMOR H U M O R -- LABEL LABEL FUNNY F U N N Y -- SPARE SPARE ENJOY E N J O Y -- TREND TREND PROUD P R O U D -- EXTRA EXTRA CASH C A S H -- BONE BONE
Negative -N e g a t i v e Neutral N e u t r a l
CANCER C A N C E R -- BORDER BORDER DAMAGE D A M A G E - LOCATE LOCATE IERROR CUSTOM TERROR -- C USTOM ANGRY A N G R Y -- ROUIE ROUTE FAULT HABIT FAULT -- HABIT STORM S T O R M -- TRACE TRACE WORSE WORSE -- STONE STONE FOOL FOOL -- SEED SEED
INJURY I N J U R Y -- MARGIN MARGIN HATRED DEIECT H A T R E D -- DETECT PANIC PANIC -- LABEL LABEL FALSE FALSE -- SPARE SPARE GUILT G U I L T -- TREND TREND SNAKE S N A K E -- FENCE FENCE HUNG H U N G -- WIRE WIRE PALE PALE -- GEAR GEAR
DEVIL DEVIL -- STAMP STAMP VICTIM V I C T I M -- SWITCH SWITCH ERROR E R R O R -- TRACK TRACK BLAME BLAME -- EXTRA EXTRA WASIE WASTE -- SHEET SHEET CRIME C R I M E -- PANEL PANEL UGLY U G L Y -- BONE BONE
Personality and and Emotion Emotion- G. G. Matthews Matthews(Editor) (Editor) Cognitive Science Perspectives on Personality -
© 9 1997 1997 Elsevier Science B.V. B.V. All rights rights reserved. reserved.
CHAPTER CHAPTER
6
in Emotion-Antecedent Emotion-Antecedent Appraisal Levels of Processing in M. van Reekum Reelcum and and Klaus R. Scherer Carien M
Given the widely different scope of phenomena that are currently described by the word "emotion" - ranging from fleeting preferences to existential life crises - it seems advisable to propose a brief working definition of the psychological construct to be dealt with in this chapter. Following a of Soberer ((1993a), suggestion by Scherer 1993a), we define an emotion episode as a sequence of interrelated, synchronized changes in the state of of all organismic of of an external or internal stimulus subsystems in response to the evaluation of event as relevant to central concerns of of the organism. As highlighted in the of an event with respect to its relevance definition, the evaluation or appraisal of of to the individual is seen as responsible for the elicitation and differentiation of of such "emotion "emotionemotion. This chapter is concerned with the nature of constitutive" appraisals and the role of of individual differences. Lazarus ((1968) first 1 968) and Arnold ((1960) 1 960) were the fi rst to elaborate on the Lazarus of "appraisal" in an effort to explain and differentiate between the notion of positive and negative emotional consequences of of an event. Following their pioneering work, aa large number of emotion" have number of of "appraisal theories of been developed in attempt to to predict the elicitation and differentiation of in an attempt of emotion on the basis of appraisal criteria (De Rivera, 1977; basis of of aa detailed set of 1 977; Frijda, 1986; 1 986; Oatley & Johnson-Laird, 11987; 987; Lazarus, Lazarus, 1991; 1 99 1 ; Roseman, 1984, 1 984, 1991; 1 99 1 ; Soberer, Scherer, 1982, 1 982, 1984, 1 984, 1986, 11988; 988; Smith & Ellsworth, 1985; 1 985; Solomon, 1976; All of of these authors consider the 1 976; Weiner, 1982, 1 982, 1986). All appraisal of potentially appraisal process process generally as aa subjective, cognitive evaluation of emotion-eliciting events. In spite spite of of divergent disciplinary disciplinary and and historical historical traditions traditions of of the authors authors involved, one finds aa high degree of of convergence with respect to the nature nature of of the the appraisal appraisal dimensions or or criteria criteria postulated postulated by different theories theories (see Lazarus Tetlock, 1989; Lazarus & & Smith, 1988; 1 988; Manstead Manstead & & Tetlock, 1 989; Reisenzein Reisenzein & & Hofmann, Hofinann, 1990, 1 990, 1993; 1 993; Roseman, Roseman, Spindel, Spindel, & & Jose, 1990; 1 990; Scherer, 1988). 1 988) . These include the perception perception of of aa change change in the the environment that that captures captures the subject's subject's attention or attention (novelty and and expectancy), the perceived pleasantness pleasantness or unpleasantness of the unpleasantness of of the stimulus stimulus or or event (valence), the importance importance of or event to to one's one's goals or or concerns (relevance and goal stimulus or
Chapter Chapter 6
260
conduciveness conduciveness or motive consistency), consistency), the notion of of who or what caused the event (agency or responsibility), the estimated ability
to deal with the event
and its consequences consequences (perceived (perceived control, power or coping potential), and the evaluation of of one's own actions in relation to moral standards or social norms (legitimacy), and one's -ideal. At the very least, this convergence on a one's self self-ideal. standard set of of appraisal criteria suggests high face validity of of the underlying assumptions. In addition, addition, a number of of empirical studies have provided or the idea that a small number of support ffor of relatively general appraisal dimensions erentiation of jor emotions. A dimensions or criteria can explain the diff differentiation of ma major
of diff different variety of erent paradigms has been used to study the relationship between particular particular configurations of of appraisal results and the nature of of the ensuing emotional
reaction. reaction.
The
results
generally
confirm
the
fundamental
assumptions of of appraisal theories as well as specific predictions about
In general, appraisal profiles allow the 40 to 50% of of the emotion episodes studied (see Scherer, Scherer, 1997 ffor or an overview of of this research).
emotion-specific appraisal profiles. profiles. discrimination of
Critique of of Appraisal Notions Despite substantial empirical evidence demonstrating the explanatory erentiation, the appraisal approach value of of appraisal theory in emotion diff differentiation, has been heavily criticized with respect to both its presumed emphasis on 984; cognitive, jonc, 1980, cognitive, i.e. i.e. conscious or voluntary, processes (e.g. (e.g. Za Zajonc, 1980, 11984; to study appraisal Berkowitz, 11994), 994), and the research methods generally used to reports of of presumed appraisal in emotion (particularly the emphasis on verbal reports experiences recalled from memory). The critique of of processes in emotion experiences of experimental exaggerated cognitivism is in large part based on the kinds of paradigms used in this research tradition: 1) asking subjects to recall specific 1) specific emotional experiences experiences and of antecedent antecexlent evaluation processes questioning them about the outcome of Folkman & L Lazarus, azarus, 1988; Frijda, Kuipers, & ter (Ellsworth & Smith, 1988; Folkman Scherer, 1988; Mauro, Sato, & Tucker, 1992; 1992; 989; Gehm & Scherer, Schure, 11989; Reisenzein & Hofinann er, in press; Roseman, Hofinann,, 1993; Reisenzein Reisenzein & Spielhof Spielhofer, 990; Smith & Ellsworth, 1985; Tesser, 1990); Spindel, & Jose, 11990; 2) obtaining judgments on appraisal processes after after naturally occurring or emotion-producing events such as examinations or emotions induced experimentally (Folhnan (Folkman & Lazarus, 11985; 985; Smith, 11989; 989; Smith & Ellsworth, 1987);
C.M. van Reekum and KR. K.R. Scherer eM
26 2611
the appraisal implications of the 3) having emotion words judged as to the 987; Frijda, 1987; Parkinson & underlying concepts (Conway & Bekerian, 11987; Schutzela.ars, 11986/87), and; Lea, 11991; 99 1 ; Smolenaars & Schutzelaars, 986/87), and; that have been systematically 4) using vignettes or scenarios that manipulated with respect to appraisal relevant dimensions and asking reactions that that they - or a fictitious other subjects to indicate the emotional reactions might experience in this situation (McGraw, 1987; Roseman, 1984; Russel & Weiner, & Li, 11989; McAuley, 11986; 986; Smith & Lazarus, 11993; 993; Stipek, Weiner, 989; Weiner, Amirkhan, Folkes, & Verette, 1987; Weiner, Graham, & Chandler, 11982; 982; Weiner, Russel, & Lerman, 1979). of these cases, participants are required to engage in conscious, In all of complex inference or imagination processes, followed by verbalization. Clearly, all of of these processes require a fairly high level of conceptual processing. It has been argued (Parkinson & Manstead, 11992, 992, 1993; Parkinson, 11997) participants are unlikely to be able to to report upon 997) that participants antecedent appraisal processes which mostly occur outside awareness. antecedent Furthermore, these critics point out that the very process of of imagining an upon cognitive interpretation and emotional event depends to a large extent upon circumstances, participants participants are likely to construct a memory. In such circumstances, rationale for their their emotional response, a situation which is possibly made rationale participants with ready made dimensions of of appraisal. worse by providing the participants problems parallel parallel those encountered by cognitive Cognitive psychologists and These problems knowledge engineers when attempting to extract extract information from experts about about problem-solving and and decision-making strategies which are often overtrained that experts consistently overtrained and therefore therefore implicit. It has been shown that report having used aa logical strategy which bears report having bears only limited resemblance to the one they actually actually used (see Berry, 1987, for an excellent summary of of this literature). that the empirical evidence literature). Furthermore, Furthermore, Frijda Frijda (1993) ( 1 993) has argued that supporting appraisal, than describing emotion antecedent antecedent processing, processing, appraisal, rather rather than has to date been limited to describing the the content content of of an emotion or the meaning of of the respective labels. Specifically, post-hoc analyses of of an emotional emotional event are are susceptible susceptible to to logical inferences inferences of of why an emotion arose, arose, but may but may not include an an accurate accurate description description of of the the actual actual antecedent antecedent evaluation evaluation which occurs occurs in part part outside of of awareness. awareness. While While these these criticisms criticisms are are well taken, taken, they reflect reflect the difficulties difficulties faced faced any attempt attempt to to observe and measure measure cognitive processes. Since thoughts by any cognitive processes. are are neither directly directly observable observable nor nor measurable, measurable, the only available available access access so far is to far to request request individuals individuals to to verbalize verbalize whatever whatever parts of of their their mental processes processes are are available available to to consciousness consciousness and and can be encoded in language.
262 262
Chapter 6
Cognitive Cognitive psychologists psychologists have have used used ingenious ingenious experimental experimental paradigms paradigms using using systematic variations variations of of exposure exposure conditions conditions and and reaction reaction times times to to study study systematic cognitive processes processes without without recourse to to conscious conscious verbal verbal report. However, However, cognitive they have have mostly studied formal formal aspects of of cognitive cognitive operations operations often often using using they emantic structures. formalizexl rule rule systems systems such such as as syntactic syntactic and and ssemantic structures. highly formalized Even in in these these highly highly constrained constrained cases, cases, the the use use of of reaction reaction time time as as aa "royal "royal Even road" road" to to understanding understanding cognitive cognitive processing processing has been been called into into question question since its its interpretation interpretation is often often highly highly conjectural conjectural (see Scherer, Scherer, 11992). since 992). Appraisal theorists theorists face face the the additional additional complication complication that that the the evaluation evaluation and and Appraisal inference processes studied are are largely largely independent of of language language or or other other inference formally structured structured systems and and often often occur outside outside of of awareness. awareness. Thus, Thus, the the formally issue issue here here is not not to to study study formal formal aspects aspects of of the the respective operations, operations, but but the the assessment of of the the content of the appraisal appraisal processes. processes. Clearly, Clearly, there there are, are, so assessment of the far, few few alternatives to to studying studying these these processes processes other other than than through through the the use of far, of verbal report. report. In In addition, addition, it it is probably probably fair fair to to state state that that the the critics critics of of appraisal appraisal theory theory have have so so far far failed failed to to suggest convincing convincing alternatives to to conceptualizing conceptualizing and and measuring measuring the the elicitation elieitation and and differentiation differentiation of of the entire entire gamut of of emotional emotional states. states. The The theoretical theoretical statements statements concerning concerning the the gamut structural structural components components of of the the appraisal appraisal profiles underlying underlying specific emotional emotional states states and and the the empirical empirical evidence evidence obtained through through verbal verbal report report so so far far have have laid solid solid foundations further work area. laid foundations for for further work in in this this area. In summary, summary, the the reliance upon upon questionnaire questionnaire data, data, which which by by its its very very In nature nature reflects reflects high-level, high-level, conscious processes, processes, combined combined with with the the large large number relatively sophisticated sophisticated appraisal appraisal dimensions dimensions typically posited, has number of of relatively typically posited, has led many critics critics to erroneously deduce deduce that led many to erroneously that appraisal appraisal processes processes are are necessarily necessarily deliberate, deliberate, conscious conscious and and thus thus "cognitivistic". "eognitivistie". In addition, addition, many many of the appraisal of appraisal theories are are worded worded in in such such aa way way as as to to reinforce reinforce this this misconception. misconception. However, However, some some appraisal appraisal theorists theorists have have pointed pointed out out quite quite explicitly conscious or explicitly that that appraisal appraisal processes processes are are not not necessarily necessarily conscious or voluntary voluntary (e.g. Frijda, 11993; 993; Lazarus, 984; Seherer, Scherer, 11984b). 984b). In (e.g. Frijda, Lazarus, 11984; In fact, fact, many many of of the the criticisms criticisms discussed discussed above above seem seem to to stem stem from from the the inherent inherent ambiguity ambiguity associated associated with with the the word word "cOgnition", "cognition", and and the the lack lack of of aa consensual consensual definition definition thereof. What is evidently needed in this situation is not a debate concerning thereof. What is evidently needed in this situation is not a debate concerning the the semantics semantics of of the the word word "cognition" "cognition" but but rather rather aa recognition recognition that that different different types types or levels of of processing processing are are implicated in in emotion-antecedent emotion-antecedent appraisal. appraisal. In 1 987) have In this this spirit, spirit, Leventhal Leventhal and and Scherer Scherer ((1987) have suggested suggested the the adoption adoption of of aa "levels "levels of of processing" processing" approach approach to to more more precisely precisely specify specify the the mechanisms mechanisms underlying underlying appraisal, appraisal, trying trying to to correct correct the the misconception misconception that that appraisal appraisal is predominantly predominantly aa high-level, conscious, conscious, or or controlled controlled cognitive cognitive activity. activity.
263
C.M. van Reekum and and K.R. Scherer CM
in Appraisal Levels of Processing in Soberer ((1987), of Leventhal and Scherer 1 987), in trying to demonstrate the futility of "cognition-vs.-emotion" approaches; on "cognition-vs .-emotion" debates, integrated their respective approaches; the one hand Leventhal's ((1979, 1 979, 11984) 984) perceptual-motor theory, suggesting
constructive activity of of a multi-component, that emotions are produced by the constructive hierarchical processing system; and, on the other hand, Scherer's ((1982, hierarchical 1 982, 11984a) 984a) component process theory, arguing for emotion elicitation and
differentiation through through a sequence of of stimulus evaluation or appraisal appraisal differentiation "checks" (SECs). Specifically, they suggested that appraisal can occur at
different three diff erent levels - sensory-motor, schematic, and conceptual. Table 11 reproduces the central figure from Leventhal and Scherer ((1987, 1 987, p. 117) 7) of appraisal occurring at at different levels of of suggesting a preliminary model of processing. processmg. Table l1.. Levels Levels of of processing processing for Stimulus Evaluation Evaluation Checks Checks.. Table ,=
Novelty
Conceptual Conceptual Level
Expectations: Expectations: cause/effect, cause/effect, probability probability estimates estimates
Pleasantness Pleasantness Goal/need Conducive Conducive Recalled, Recalled, anticipated, anticipated, or derived positivepositive-
Conscious Conscious goals, goals, plans
Coping Coping Potential Potential L
,
Norm/self Nonnlself ComEatibili!}: Cgmpatib!lity
Problem Problem solving solving ability ability
Self ideal, ideal, moral moral evaluation evaluation
negative evaluations
Schematic Level
Familiarity: Learned Acquired Learned Familiarity: Acquired schemata preferences/ schemata preferences/ needs, needs, matching aversions motives aversions
Body Body schemata
Self/social l Self/socia schemata schemata
SensorySensorymotor motor Level
Innate Basic needs Sudden, intense preferences/ intense preferences! stimulation aversion aversion stimulation
Available Available energy
(Empathic (Empathic adaptation7) adaptation?)
Note. Reproduced Reproduced from from Leventhal Leventhal and Scherer, Scherer, 1987, 1 987, p. 17.
Evaluation of events at Evaluation of at the the sensory-motor level mostly involves processing by innate, unconditioned, hard-wired feature by innate, unconditioned, hard-wired feature detectors, giving rise to The underlying assumption to reflex-like, often often uncontrolled uncontrolled reactions. The assumption is that, that,
264
Chapter 6
the sensory-motor mechanisms were tuned in such such a through evolution, the manner as to quickly recognize and respond to intrinsically harmful or
particular orienting beneficial stimuli. The rudimentary adaptive responses, in particular and defense reactions, are generated generated without the interference of of more
adaptive responses. controlled processes to ensure rapid adaptive Affective responses exclusively elicited through sensory-motor processing are likely to to be short-lived and are often of of a reflex-like nature as 990). One can ask to what in the ease case of of startle (Lang, Bradley, & Cuthbert, 11990). "true" emotions in the sense of of the extent such response patterns constitute ''true'' definition suggested above, i.e. involving changes in all organismic a synchronized fashion (see also Leventhal & Scherer, Soberer, 1987). subsystems in a Even though this might be be expected in some cases, eases, most adult emotions observed in daily lif lifee are unlikely to be exclusively accounted accounted ffor sensoryor by sensory motor processing. The The schematic level of of processing processing is based on structures structures acquired during past learning experiences of of an organism, which have often ot~en been called the past "schemata". 1932) described the schema as "schemata". Bartlett Bartlett ((1932) as the the active, organized setting within which new experiences are influenced by those previous reactions and experiences that that are connected by some common aspect. These schemata or the accommodation to schemata are expected expected to to be of of central central importance ffor or assimilation of of objects and events (see Piaget, 1937/1955). With respect to 1 987, schemata created in emotional encounters, Leventhal and Scherer ((1987, 0) suggested that that they "are concrete representations representations in memory of of p. 110) specific perceptual, perceptual, motor (expressive, approach-avoidance tendencies, and autonomic reactions), and subjective eelings each of subjective ffeelings of which were components of of the reactions during specific emotional episodes". Newly experienced events or stimulus patterns can lead to the activation activation of of schemata schemata which an throughout its life, and responses are likely to be organism has acquired throughout generated largely in accordance accordance with the reaction components of of the existing schema. The conceptual conceptual level of of processing is characterized characterized by more abstract, abstract, active, reflective processes, as opposed to the template template matching-like processes processes which were specific to the first two levels. This level comprises capacities capacities to to infer generalities from from and reason about events on the the basis of of propositionally propositionally organized memory structures, structures, which which are developed by comparisons between past past experiences. Processes such as anticipating or problem-solving are or the conceptual level. While inherently more are typical typical ffor effortful effortful and and less "automatic" "automatic" than than the preceding preeexting two levels, Leventhal and Scherer ((I1 987) expect this type of of processing to become, with maturation of of
265 265
C.M. van van Reekum Reekum and and KR. K.R. Scherer Scherer CM
the individual individual and and perf performance experience, "increasingly "increasingly rapid, rapid, automatic, automatic, the onnance experience,
and ''lazy and mindless' mindless' (Langer, (Langer, Blank, Blank, & & Chanowitz, Chanowitz, 11978)". In other other and lazy and 978}". In words, conceptual processing of frequently frequently encountered encountered situations situations is is thought thought to to
become become
schematic schematic
and and
thus thus
conceptual conceptual
"content" "content"
can can
become become
"schematized". Leventhal and and Scherer ((1987, argue that: that: "schematized". 1 987, pp.. 22) argue Indeed, we we expect expect most most if if not not all all emotional emotional processing processing to to be be Indeed, at the the middle, middle, or or schematic schematic level. level. Experience Experience with with specific specific initiated at and events develops schemata which which encapsulate persons, objects, and (Hebb, 11949) their history in perceptual memories or "identities" (Hebb, 949) that then organize current emotional experience (Bruner, 11957; 957;
Yates, 11985). 985). of the the criticisms that is frequently raised with respect One of
to Leventhal
Scherer'ss proposal concerns the assumption that all checks can can occur at and Scherer' all levels. It may seem odd, indeed, to claim that a painful stimulus is
individual's evaluated as obstructive to an individual 's "central concern" (Berkowitz, Lazarus-Zajonc personal communication). As in the classic Lazarus-Za jonc debate, the problem seems to be one of of semantics. Most of of the terms used in psychology unfortunately unf ortunately have a strong cognitivistic or mentalistic bent, obscuring the that the underlying processes can be very primitive indeed. The fact that fact that a reaction to pain may be very automatic or reflexive does not contradict the a idea that these defensive reflexes are at the service of of very fundamental needs of of the organism such as being unharmed, dry, warm, and fed. The "central concerns" of of the the organism thus range all the way from very basic concerns" motivational constructs (such as basic needs) to the most elaborate goals and plans (of. 1 970, hierarchy of of motives). Unfortunately, the neglect (cf. Maslow's, Maslow's, 1970, of of motivation in psychology and and the diffmulties difficulties inherent in functional argumentation is responsible for for the relative lack of of sophistication in our treatment of treatment Leventhal and Scherer made aa of these motivational phenomena. Leventhal preliminary attempt the appraisal attempt to to label label the appraisal criteria in a a manner that that is appropriate to each level. For example, with respect to the appropriate to each level . For with the goal goal conduciveness conduciveness check, they they mention "basic "basic needs" for for the the sensorimotor level, "acquired "acquired needs, needs, motives" motives" for for the the schematic level, and and "conscious "conscious goals, plans" plans" for for the the conceptual conceptual level. This This exercise proved to to be be rather rather difficult, particularly particularly for for the the coping coping potential potential and and norm/self nonn/self compatibility checks (see Table Table 1), 1 ), and and further further efforts efforts are are required required to to elaborate elaborate on these these concepts. concepts . Leventhal and Scherer's proposition of a levels of processing Leventhal and Scherer' s proposition of a levels of processing model, even even though though preliminary preliminary and and highly highly conjectural, conjectural, as as emphasized emphasized by by the the
266
Chapter 6
the need to address different levels of of emotion-antecedent emotion-anteceAent authors, recognized the appraisal appraisal and argued for a new approach to studying appraisal. Although this early suggestion provided a potential mechanism to explain the more or less automatic elicitation and and differentiation of of emotion, the adoption of of multi multiappraisal theorists theorists remains rather limited to date. One of of level frameworks by appraisal the possible reasons might be that that most theorists found it difficult to go beyond the rather general suggestion by Leventhal and Scherer, given the uncharted nature of of the territory and the difficulty of of operationalizing the concepts for empirical study. Appraisal theory, although claiming to be a cognitive approach to emotion, has generally not contributed much to the discussion of of exactly which cognitive processes underlie appraisal. For instance, attentional processes (i.e. the allocation of of resources to a potentially of irrelevant information sources often ot~en significant stimulus and the inhibition of controlled by the general concerns of of the individual; see Williams, Mathews, & Macleod, 996) are not addressed by appraisal Macl_exxt, 11996) appraisal theorists theorists even though though they emphasize the evaluation of of a situation with respect to its relevance to one's goals. The The adoption of of more elaborate cognitive processing models would of appraisal theories in seem to be the next logical step in the development of of the processes underlying the elicitation elicitation order to better explain the nature of differentiation of of emotion and to thus address some of of the criticism of of and differentiation "classical" "classical" appraisal approaches approaches mentioned above. Hierarchical Process Process Notions Notions in Related Related Traditions Traditions Hierarchical
Contrary to the the relative neglect of of hierarchical process models by appraisal theorists, theorists, these notions have have received strong attention from theorists appraisal of social and researchers concerned with affective phenomena in the areas of cognition, clinical psychology, and neuropsychology. In envisaging a hierarchical, process-oriented model for appraisal, it is useful to review the hierarchical, might shed light on aspects of of emotion-antecedent approaches which might cognitive processing. There are a number of of different research traditions, of one another and characterized by developed quite independently of of the emotion process, that qualify for this emphasizing specific aspects of of the approaches that seem review. In what follows, we will outline some of of levels in emotion-antecedent emotion-antecextent processing. directly pertinent to the issue of
C.M. van van Reekum Reekum and and K K.R. CM R.
Scherer Scherer
267 267
affective Implicit perception and aff ective preferences Zajonc, one one of of the major major critics of of the the "exaggerated "exaggerated cognitivism" cognitivism" of of Zajonc, appraisal theories, theories, bases his his argument argument (Zajonc, (Zajonc, 11980, on research research appraisal 980, 11984) 984) on findings showing that affect is generated generated for stimuli which cannot cannot be be findings (e.g. Kunst-Wilson & & Zajonc, 11980; Murphy & & consciously perceived (e.g. 980; Murphy "affective primacy hypothesis", hypothesis", claiming claiming that affective Zajonc, 11993). 993). His "affective towards a stimulus stimulus can be elicited with with minimal minimal stimulus stimulus input and reactions towards virtually no cognitive processing, has since triggered much research in the (e.g. Niedenthal, 11990; domain of implicit perception and emotion (e.g. 990; Niedenthal & Kitayama, 11994). 994). For instance, Murphy and Zajonc ((1993) 1 993) carried out an experiment in which they established positive and negative valence for a an affective priming priming procedure. They previously neutral stimulus using an compared the effects of what they called "affective" and "cognitive" priming on the judgment of novel stimuli under extremely brief and longer expressing happiness and exposure duration of the primes. Pictures of faces �xpressing anger were used as primes, while Chinese ideographs, selected as being affectively neutral, were used as targets. Among other effects, they found that only the affective primes presented at a subliminal level level produced significant of "good" versus "bad" towards the targets. shifts in subjects' evaluations of These results were explained by claiming that "when affect is elicited at of conscious awareness, it is diffuse and its origin and address levels outside of are unspecified. unspecified. Because Because of of its diffuse quality, nonconscious affect can can "spill are over" onto unrelated things" (p. 736). One can argue that Zajonc's affective primacy hypothesis is concerned with the role of of consciousness or awareness rather than that that of of cognition, cognition, emphasizing again the importance importance of of definitional definitional rather of the empirical issues (see Leventhal & Scherer, 1987). 1 987). Furthermore, most of studies by Zajonc and followers followers concern affective preferences preferences and are thus not directly pertinent to the study of of emotion as defined above. Yet, the research in this tradition tradition addresses two two important issues, notably a) the fact that stimuli are often evaluated in an automatic manner, that that is, without much cognitive effort, and b) that that there may be an affective response disposition to certain stimuli, stimuli, such as facial expressions expressions of of certain emotions. Clinical Clinical approaches approaches to emotional emotional disorder disorder The study of of pathology pathology has has always been of of great great importance importance to to The study understanding the fundamental mechanisms of behavior, providing through understanding the fundamental of behavior, through highlighting malfunctioning malfunctioning and and its its effects effects - aa privileged privileged window window on the the
268 268
Chapter 66 Chapter
underlying underlying mechanisms. Not Not surprisingly, surprisingly, then, then, the study study of of emotional emotional disturbance can can greatly greatly contribute contribute to to the the understanding understanding of of normal normal emotion. disturbance Emotion and and attention. attention. Recently, 'Recently, clinical clinical research research on on emotion and and Emotion attention attention has has applied applied information-processing information-processing modds models to to the the question question of of the the role of of emotional states states and and traits traits in attontional attentional processes processes (sdection (selection and and intensive processing & Matthews, Matthews, 1994) 1 994) and sustaining sustaining processing of of stimuli, see Wells & thoughts (Teasdale (Teasdale & & Barnard, Barnard, 1993; 1 993; Wells & Matthews, Matthews, 1994). emotional thoughts Experimental research research in the the fidd field of of attentional attentional bias result of of an Experimental bias as aa result emotional disturbance has been disturbance has been inspired by paradigms devdopexi developed within the tradition of of cognitive psychology (e.g. the Stroop color naming task; Stroop, tradition 1935), 1 935), resulting resulting in interesting interesting adaptations adaptations of of these paradigms paradigms to problems problems of of emotional malfunctioning emotional malfunctioning (e.g. using using an an emotion version of of the the Stroop Stroop task). This body of work cannot here; for more complete This extensive extensive body of work cannot be reviewed here; of theoretical theoretical models and research results, the reader is referred to reviews of Teasdale Teasdale and Barnard Barnard (1993), ( 1 993), Wells and Matthews Matthews (1994) ( 1 994) and Williams, Mathews, Mac~ (1996). Mathews, and Macleod ( 1996). In brief, evidence so far shows that that disturbed patients often show an attentional bias towards emotionally disturbed information relevant to their concerns or worries. This effect effect can also be found in non-clinical groups trait emotion (e.g. trait-anxiety), or for groups with high trait participants following emotion induction, but normal participants but tends to be less pronounced than in clinical patients (Wells & Matthews, 11994) 994) 1I.. pronounced & can be presented subliminally, Furthermore, relevant information earl eliciting a comparable attentional bias (although the effect is small in magnitude). This emphasizes the automatic nature nature of of the detection detection of of 992; Mogg, Bradley et aI., 993; Mogg, relevance (see Macleod MaeLexxt & Hagan, Hagan, 11992; al., 11993; al., 11993; al., 11996). Kentish et aI., 993; all in Williams et aI., 996). Automatic, low-level processes, triggered by specific specific stimulus inputs and often interacting with higher-order controlled processes, are held responsible for this shift shill of of attention toward potentially relevant relevant information (e.g. Wells & Matthews, 11994). 994). , The fact that that attentional bias is found for persons high on certain emotional traits (especially trait anxiety) but not for normal individuals experiencing a particular emotional state (such as state anxiety), is explained by Wells and Matthews Matthews in terms of of unusual and extensive S-R learning experiences, which are supposedly anchored in low-level memory traces. 11 This Thisphenomenon, phenomenon,however, however,C8JUlot cannotbe replicated replicatediinn single singlestimulus stimulustasks, tasks, such suchas as lexical lexical 988) or decision decision tasks tasks with with emotionally emotionallyrelevant relevantwords words(see (see Mathews, Mathews, 11988) or homophone! homophone/ 989 for homograph Richards, & homographspelling spelling(e.g. (e.g. Mathews, Mathews, Richards, & Eysenck, Eysenck, 11989 for homophone homophonespelling; spelling; 992 for homograph 994). French French & & Richards, Richards, 11992 homographassociations; associations;in Wells Wells& & Matthews, Matthews, 11994).
C.M. CM
K.R. van Reekum and K. R. Scherer
269
Whether or or not not most most phobics phobics have have actually actually encountered encountered the the object object of of their their Whether phobia in in aa more more striking striking manner manner than than nonphobics nonphobics is is an an open open question question (see (see phobia Rachman, 11974, cited in in Frijda, Frijda, 11986). What seems seems clear, clear, however, however, is is that that Rachman, 974, cited 986). What through aa process process of of repeated repeated exposure, exposure, reflection, reflection, and and rehearsal, rehearsal, through associations between between objects objects related related to to the the phobia phobia and and the the resulting resulting emotional emotional associations response tend tend to to become become highly highly automatized automatized or or schematized. schematized. response Even though though strong strong attentional attentional bias bias effects effects have have been been shown shown mostly mostly in in Even clinical patient patient groups groups rather rather than than nonpatient nonpatient groups, groups, these these studies studies imply imply that that clinical information which which is is relevant to to the the organism organism is is given given processing processing priority. priority. information The fact fact that that personal personal worries worries or or concerns concerns play play aa significant significant part part in in the the The attentional bias bias effect effect fits fits with with one one of of the the basic basic tenets tenets of of appraisal appraisal theory, theory, attentional namely that that information information is is assessed assessed on on the the basis basis of of its its need, need, goal, goal, or or concern concern namely relevance. Appraisal Appraisal theory theory would would thus thus predict predict that that attentional attentional bias bias effects effects relevance. will be be more more pronounced in individuals with with affective affective disorders than than controls, controls, to the the extent extent that that they they appraise appraise certain certain stimuli stimuli as as being being more more concern concernto relevant. What current current appraisal theory fails fails to to specify are the the precise precise relevant. What appraisal theory specify are mechanisms whereby whereby the the appraisal appraisal of of aa stimulus stimulus as as highly highly relevant relevant regulates regulates mechanisms the organism's attention. O Ohman's "'preparedness"" model oof Ohman Ohman hman 's "preparedness f emotion-generation. centrally concerned with with models models of of fear, fear, panic, panic, and and anxiety anxiety ((1986, 1 986, 11988) 988) is centrally disorders. His on Pavlovian conditioning of of fear-relevant fear-rdevant disorders. His research, research, based based on Pavlovian conditioning stimuli, reveals reveals that that physiological physiological responses responses can can be be activated activated pre-attentively, pre-attentivdy, stimuli, and that learning can take place without the being and that emotional emotional learning can take place without the participant participant being aware of its research evidence, evidence, Ohman 0hman argues argues that that aware of its source. source. Based Based upon upon this this research evolution mechanisms in manner as evolution has has tuned tuned the the perceptual perceptual mechanisms in such such aa manner as to to promptly stimulus relevant the organism organism promptly generate generate aa response response as as soon soon as as aa stimulus relevant to to the is at aa very of the stimulus. According According to is perceived, perceived, even even at very low low level level of of analysis analysis of the stimulus. to this view, the perceptual and this perceptual systems detect relevant stimuli automatically and independently of the current soon as as aa relevant relevant stimulus stimulus of the current attentional focus. As soon is perceived, ongoing actions are are interrupted and and more more conscious, i.e. controlled, processing mechanisms are activated activated to to further further analyze analyze the situation's action is situation's significance significance before before directed directed action is taken. taken. This This "request" "request" for for controlled unspecific physiological controlled processing processing is is accompanied accompanied by by an an unspecific physiological response, response, preparing preparing the the organism organism for for subsequent subsequent emotional emotional responses. responses. The ( 1 986) model comprise comprise aa primary primary The controlled controlled processes processes in 0hman's Ohman's (1986) appraisal appraisal mode, mode, aa secondary secondary appraisal appraisal mode mode and and aa response response selection selection mode. mode. Primary the emotional Primary appraisal appraisal is is defined defined as as aa further further evaluation evaluation of of the emotional significance of of an event using any related information held in memory at at aa conscious or explicit level. the events conscious or explicit level. Secondary Secondary appraisal appraisal evaluates evaluates the events in in
270
Chapter 6
relation to to available available action action alternatives alternatives bef before a response response is is selected selected (response (response relation ore a selection selection mode). mode). This This sequence sequence of of modes modes results results in in physiological physiological activation, activation,
motor behavior, and, and, potentially, potentially, verbal verbal responses. responses. motor Research findings reported (e.g. Ohman, Research findings reported by by Ohman Ohman and and colleagues eoUcagues (e.g. 0hman,
1988) underscore pre-attentive activation underscore the the prc-attcntivc activation aspect aspect in in Ohman's 0hman's model, model, in in
contrast with with the the controlled controlled (appraisal) (appraisal) processes processes which which have have not not been been contrast thoroughly thoroughly addressed. addressed. The The research research paradigm paradigm generally generally used used in in this this work work
consisted of of conditioning conditioning of of emotionally emotionally relevant relevant and and irrelevant irrelevant pictures pictures to to aa consisted light electric electric shock, shock, ffollowed by subliminal subliminal presentation presentation of of the the conditioned conditionod light ollowed by pictures (using backward backward masking masking procedures) in in order order to to maximally maximally exclude controlled controlled or conscious processing processing of of the the pictures. pictures. Skin conductance conductance response response measures measures revealed revealed increases increases in in autonomic autonomic nervous nervous system system (ANS) (ANS) activity activity only ffor fear-relevant pictures, pictures, such as as spiders and and snakes (0hman, or 986) or only or fear-relevant (Ohman, 11986) angry facial facial expressions expressions (Ohman, (0hman, Dimberg, & Esteves, Estcves, 11989). No such such 989). No angry eff ects were ound with fear-irrelevant pictures, effects were ffound with the the fear-irrelevant pictures, such such as as mushrooms mushrooms or or happy ffacial expressions. Ohman Ohman and and Soares Soarcs ((1994) have been been able able to to extend extend happy acial expressions. 1994) have these findings these findings to to spider spider and and snake snake phobias phobias where where the the phobics phobics show show similar similar physiological responses responses when when presented presented either either optimally optimally (i.e., (i.e., clearly clearly visible) visible) physiological or subliminally subliminally with with the the object object of of their their phobia. phobia. or (ICS). an attempt attempt to to better better model model Interacting cognitive subsystems (lCS ). In an cognitive-affective relationships, Teasdale Teasdale and and Barnard Bamard ((1993; Bamard & cognitive-aff ective relationships, 1993; Barnard Teasdale, 199 1991) have presented presented their their framework framework of of interacting interacting cognitive cognitive 1) have subsystems (ICS). (ICS). Briefly, Briefly, ICS ICS consists consists of of 9 subsystems, subsystems, including including subsystems sensory/proprioceptive, eff effector, propositional, and and implicational implicational subsystems, subsystems, sensory/proprioceptive, ector, propositional, each of of which which handles handles aa diff different type of of inf information. Contained in in every every each erent type ormation. Contained subsystem is is aa memory memory ffor that system's system's previously previously processed processed inf information, subsystem or that ormation, processes to to update update and and access the the memory, memory, as as well as as processes processes to to transform transform information into aa representation representation (or (or coding) coding) suitable for for processing processing by by the the inf ormation into other subsystems. subsystems. For For aa full full description of of the the ICS ICS framework framework and and the the other application of of ICS ICS to to emotion, emotion, the the reader is is ref referred to Teasdale Teasdale and and Barnard Bamard application erred to ((1993). 1993). Of particular particular relevance relevance to to emotion emotion is is their their idea idea that that emotions emotions are are Of generated primarily primarily within the the implicational implicational subsystem. subsystem. The The implicational implicational subsystem directly directly receives receives and and integrates integrates inf information from many many of of the the other other ormation from subsystems, such as as the the sensory sensory and the the propositional propositional subsystems, subsystems, about about the the subsystems, occurring event. event. The The implicational implicational subsystem then then makes makes inferences inferences based based occurring upon these diverse diverse sources of of information by matching matching the the patterns patterns of of upon information by information with previously previously stored stored schemata, schemata, which which include include emotion-related, emotion-related, inf ormation with "affective themes" (e.g. (e.g. the the theme theme of of "potential "potential severe severe self self-related threat"; "aff ective themes" -related threat";
C.M. van Reekum and and K.R. Scherer CM
27 2711
Bamard & Teasdale, see Barnard
1991, 19), and generates emotional reactions 1 99 1 , p. 19), estations such as facial and bodily responses by including overt manif manifestations activating patterns in the eff ector subsystems. effector In ICS, it seems that the type of of emotion elicited is highly dependent erred inf ormation to stored inf ormation upon "matching" of of incoming or inf inferred information information about past emotional experiences act that experiences.. And despite the ffact that any final inferences occur in the implicational inferences implicational subsystem, ICS makes explicit the erent types of inf ormation, being processed at diff erent possibility that that diff different information, different "levels", contribute to this final inf erence. Also, Teasdale and Barnard's inference. Bamard's Bamard & Teasdale, 11991) affective ((1993, 1 993, Barnard 99 1 ) reference reference to the aff ective themes which are extracted from previous experiences or innately prepared responses is 1 99 1 ) notion of reminiscent reminiscent of of Lazarus' ((1991) of Core Relational Themes (CRT). Adopting an ICS approach, one could inf er that the implicational infer implieational system is the seat of of appraisal. The ICS is unique in emphasizing emphasizing the importance of of the implicational or emotion in that implieational system ffor that this subsystems "adds" "adds" emotional connotations to otherwise "cold" cognitions cognitions.. "Classic "Classic"" cognitive psychology psychology
tended to In general, mainstream cognitive processing models have tended information is represented in memory (e.g. (e.g. emphasize either the way in which information Tulving, Tulving,
information 11972; 972; Johnson, 11983), 983), or the way in which inf ormation is processed Shiffrin & Schneider, 11977). 11984; 984; Shiffrin 977). Both approaches are
(e.g. Leventhal, (e.g.
pertinent for our current topic. Emotion and and memory. Recent advances in the study of of memory have Emotion influenced work on emotion-related cognitive processing (see greatly influenced 992). For instance, Christianson, 11992). instance, the dissociation between implicit and 1 985) and Schacter explicit memory as suggested by Graf and Schacter ((1985) ((1987) 1 987) has been used by emotion psychologists in order to address and further explain the seemingly unconscious elicitation elicitation of of aff affect, ect, particularly in of emotional emotional disorders (Teasdale & Barnard, Bamard, 11993). 993). Similarly, the case of 1 98 1) influential work on the role of Bower's ((1981) of emotion on thinking and remembering remembering has been directly influenced by memory models in cognitive psychology.. psychology Of Of particular importance to the notion of of hierarchical processing is Johnson'ss (Johnson, 1983: 1983: Johnson & Multhaup, 11992) of a Multiple MultipleJohnson' 992) model of Modular memory system (MEM). Johnson proposes that memory Entry, Modular of a perceptual system (dealing with perceptual activities such as consists of seeing, heating hearing etc.), and a reflective system (which is actively involved in the
272
Chapter 6
creation and recording of of self-generated activities such as planning, comparing, imagining etc.). Based on this conceptualization conceptualization of of a multi-level memory system, Johnson 1 992) also made a first attempt to Johnson and Multhaup ((1992) elieitation of of emotion within this kind of of cognitive structure. structure. This describe the elicitation approach is briefly illustrated below. The goals and plans of of an organism, which have a central place in appraisal theories of of emotion, are considered as "agendas" by Johnson and control processing in different (sub)systems, and Multhaup. These agendas control they can can vary in complexity and in the degree to which they are perceptually (situationally) (situationally) controlled or reflectively (self-) controlled. For example, an activated perceptual agenda guides one as to where to look, whereas activated of self' self' (Johnson reflective agendas, providing the individual with a "sense of & Multhaup, 11992), 992), might give rise to one blaming others for an event. Emotions, according to Johnson and Multhaup, arise from processing within subsystems (whether or not controlled by agendas), accompanied by withinsubsystems autonomic and motor responses. All subsystems give rise to emotions, but there are differences as to which emotions arise from which systems. Some emotions, like fear, anger or joy, can arise from all subsystems. But the quality of of the emotion that arises from reflective processing will will be different from that arising from perceptual processing. As Johnson and Multhaup note, the reflective subsystems are responsible responsible for the occurrence and experience of of more complex emotions. Emotions arising from activity of the reflective subsystems unfold more slowly than emotions arising from the perceptual subsystems. To demonstrate that a multi-level process is involved in the generation generation of of an emotional state, and that the complexity of of an event can influence the automaticity automaticity of of some but not other aspects of of its evaluation, evaluation, Johnson and Multhaup investigated the acquisition and retention of acquisition of preferences preferences in amnesic patients in different situations. situations. They found that whilst amnesics developed smaller preferences preferences for imaginary men than controls (provided by biographical information depicting the men as "good" or "bad"), they showed no deficit in the development of of preferences preferences for melodies. melodies. Johnson and Multhaup concluded that that whereas preference acquisition acquisition of melodies depends largely on perceptual perceptual processes (which are intact in amnesics), the acquisition and especially especially the retention of evaluative impressions for people are mainly influenced by reflective processes (which are impaired in amnesics). While certainly pertinent to the topic under discussion, the experiments experiments reported by Johnson and Multhaup, as many others in this area, are concerned with preferences, preferences, not with emotion in the sense of the definition
CM C.M. van Reekum and KR. K.R. Scherer
273
given above. Similar dissociation effects for bona fide emotions remain to be demonstrated, even though it is tempting to conclude at this stage that at least two levels of of information processing can work in a dissociated manner. E ffort and Effort and control in cognitive processing. Cognitive models that specifically describe processing mechanisms are of of particular particular importance to of emotion-antecedent appraisal and might help to explain the the modeling of dynamics underlying emotion elicitation. The relative amount of of control or effort required for certain types of of processing has acquired great importance of (social) cognition. Distinctions such as "automatic" or in the study of "schematic" versus "controlled" or "reflective" processing provide explicit conceptualizations of of different processing levels. For example, Shiffrin and Schneider's ((1977) 1 977) notion of of "automatic processing" focuses on the aspect of effort required, and the constraints constraints imposed by the allocation of of limited of resources to a specific event or stimulus. Such concepts lend themselves to addressing the question of of how bottom-up processing of of sensory information and top-down processing of of memories of of past experiences can generate or modify an emotional response. and Schneider's notion by proposing an Logan ((1988) 1 988) extends Shiffrin and instance theory off automatization. This model provides a potentially useful the formation of of emotion schemata. In summary, theoretical construct for the of instances from memory rather Logan relates automaticity to the retrieval of than to resource limitations, as often is proposed in the literature on attention (e.g. Shiffrin & Schneider, 11977). 977). Each encounter encounter with a stimulus is encoded and stored as a specific trace in memory. The use of of a general algorithm or and replaced by retrieval from rule to perform a task is, after some practice, replaced of specific stimulus-response mappings. Thus, the more often similar memory of appropriate stimuli are encountered, the more easily one can retrieve an appropriate of those stimuli. response to anyone of This can be illustrated by the example of of a person finding himlherself him/herself in an entirely novel, highly obstructive situation. On the basis of of the previously discussed theories of of automatic versus controlled processing, one could infer that elaborate cognitive processing is required in order to appraise appraise the situation as goal obstructive and thus to generate an appropriate emotional response, for example anger. Instance theory, in contrast, predicts a somewhat different process. If assumes that emotional situations are If one assumes than comparable non-emotional events (see Bower, 11992; better remembered than 992; also presumably because because the attention is more centrally focused, see activate a trace of of Christianson, 11992), 992), the apparently novel situation might activate appraisal-response pairs from generally comparable situations. the previous appraisal-response
274
Chapter 6
of partial similarity of stimulus cues, an anger As a consequence, on the basis of response might be generated without elaborate controlled processing. Such an appraisal process is likely to occur in an implicit fashion, that is without the person being able to report exactly why he or she became irritated or angry. Although unsubstantiated, this example demonstrates how Logan and others' (e.g. Schneider, 1985; 988) description (e.g. 1985; for further references, see Logan, 11988) description of of the automatization of processes by practice might be useful in explaining how of appraisal becomes automatised. automatised. It should be noted that this account is a somewhat more elaborate model of what many authors have subsumed under genetic, schematic processing - the assimilation of novel instances to generic, prototypical patterns on the basis of of partial similarity. Brain mechanisms in emotion Multiple pathways in the neural processing oof f conditioned f ear fear responses. One of of the most important bodies of of neuropsychological neuropsyehological literature pertinent to multi-level notions of of appraisal is the work of of leDoux LcDoux (e.g. 11986, 986, 11989, 989, 11993, 993, 11994), 994), who traced two separate but interdependent routes responsible for emotional learning and emotional in the rat brain that are held responsible memory. He conditioned fear in rats by the application of a standard conditioning paradigm (presentation (presentation of of an auditory stimulus followed by an fear responses by producing electric shock), and studied the conditioned r lesions at systematically chosen locations and measuring the electrical activity of of neurons at different different sites in the rats' brains. This research shows that fear in the rat can be elicited through neural activity involving two pathways; subcortical subeortieal and cortical (or "low" and "high" roads respectively; see LeDoux, leDoux, 11996). 996). In both eases, cases, the amygdala plays a central role in the evocation of of fear responses to a conditioned auditory stimulus. The subcortical subcortieal pathway consists of of projections from the thalamus which pass rudimentary information information directly to the amygdala. The amygdala then activates the peripheral nervous system. The cortical pathway consists of of trajectories trajectories from the thalamus to the sensory (either visual, or in the case of of the research mentioned above, auditory) cortex, and then via the association cortices to the amygdala. Research from leDoux LeDoux and colleagues reveals that lesions in the auditory cortex do not interfere with the conditioned response to the sound, implying that the more direct, subcortical subeortieal route is sufficient to evoke behavioral fear responses. LeDoux leDoux (e.g. 1989, 1989, 1994) 1994) argues that the subcortical pathway processes information very quickly due to the fact that it only involves very few neural links, but only provides
CM van Reekum and C.M. and K.R. Scherer
275
coarse-grained, global information about the stimulus. stimulus. In contrast, signals processed via the "high road" provide much more detailed information but require much more processing time given that these pathways involve several links bbetween ~ o c n the thalamus and the amygdala. With respect to emotion in general, LeDoux ((1986) 1986) argues that subjective experience results from a conscious, higher-level of cognitive processing, but that every emotion is preceded precodexl by unconscious, low-level i.e. the "low emotional processing of the stimuli. The unconscious processes, i.e. road", "prepare" the organism for fight/flight behavior; the "high road" "prepare" modifies the responses, i.e. i.e. gives the reaction "direction", or inhibits the modifies processes triggered via the subcortical pathway in the case that they are maladaptive. While the research on fear conditioning conditioning in rats by LeDoux and his collaborators collaborators makes a strong case for the existence existence of different different levels of emotion-antecedent processing, it remains to be established whether the contribution of the amygdala to emotional processes in general is as important in humans as in rats (see Christianson, 1992, for some counter arguments). The subcortical pathways identified by LeDoux might be less instance, Halgren important in human emotional experiences experiences than in rats. For instance, ((1992) 1992) argues that sensory input to the amygdala via a direct thalamothalamo amygdaloid pathway is very weak or even non-existent in humans. Rather, studies studies measuring amygdala unit activity as well as stimulation of the amygdala reveal that the human amygdala functions within the cortical system, emphasizing cortico-amygdala communication in emotional electrical activity during word and face evaluation. Measurement of electrical recognition show high activation in the amygdala. Also, the amygdala seems to receive visceroscnsory studies show that viscerosensory input. The stimulation studies visceromotor and hormonal responses are produced by the amygdala as well as images, thoughts and feelings. Amygdala processes might, however, still precede the conscious evaluation (Halgren, 1992). and other brain structures. It is Interactions between the amygdala and possible that the direct pathway from the sensory cortex to the amygdala which LeDoux traced in the rat also plays a role in the emotion-antecedent sensory processing of information in humans. However, here it is likely to play only a relatively minor role, since the pathway seems to transmit mostly information about simple stimulus characteristics. In contrast, the hippocampal formation has been shown to be strongly involved in the processing of spatial and contextual information, that is, information concerning the relations relations between sensory stimuli and their embedding in
276
Chapter 6
specific spaces and places (Kolb & Whishaw, 11990). 990). Consequently, the of most human emotions complex, contextual information that is at the basis of is likely to be mediated via connections from the hippocampus to the 996).These connections might well reflect something amygdala (leDoux, (LcDoux, 11996).These like a schematic or automatic mode of of processing. There is strong evidence that the frontal lobes are also centrally involved in fear conditioning. conditioning. Lesions in the rat's medial prefrontal cortex prolong the extinction process of 996; see also of a fear conditioned stimulus (LeDoux, 11996; Kosslyn & Koenig, 11995). 995). leDoux 1996) describes a study showing that LcDoux ((1996) once the visual stimulus (a light) had been conditioned to shock, the lesioned rat rat kept responding fearfully to the light, whereas after a couple of of days the normal rat rat ceased showing any fear-related behavior. Apparently, the frontal of the amygdala. lobes in the rat monitor and regulate the outputs of In humans, the frontal lobes seem to be involved in the inhibition of of inappropriate (emotional) behavior. LeDoux ((1996) 1996) reports that frontal lobe damaged patients performing rule-directed tasks often have trouble changing from one rule to the other. For example, once they have discovered a certain rule, these patients continuously fall back on the use of of that that rule (e.g. once used to sorting on shape in a card-sorting solution, it is difficult for them to change to color-based sorting). D amasio ((1994) 1 994) illustrates the crucial role of Damasio of the frontal lobes in human emotional functioning with a number of of case studies, such as the classic case of 983), who produced a range of of Phineas Gage (see also Stuss & Benson, 11983), of anti-social and "careless" behaviors after atter having suffered severe injury to the frontal lobes. In recent work done with Bechara Beehara (e.g. Bechara, Damasio, D amasio, & Anderson, 11994; 994; see also D amasio, 11994), 994), this group found that Damasio, Damasio, humans with damage to the ventromedial prefrontal cortex show a more oblivious attitude towards a simulated gambling situation than controls. Based on these experiments, Damasio ((1994) 1 994) concludes that frontal lobe damaged patients are unable to develop affective responses which are suitable to a new situation, even though they have stable representations or factual knowledge of of future outcomes. He argues that the marking of of a positive or negative value is lacking, resulting in the inability to reject or accept a future outcome. D amasio takes this evidence as support for his "somatic marker" Damasio hypothesis, which states that that basic body-regulatory systems "prepare" cognitive processes determining what is considered to be good or bad for the organism. If If these impressions can be empirically confirmed, the frontal lobe can be considered as a crucial relay station in emotion-related processing in the sense of of affectively priming conceptual processes.
CM R. Scherer C.M. van Reekum and K. K.R.
277
Issues in Rewriting Appraisal Appraisal Theory
the neurosciences, neurosciences, as well as in social and clinical The recent work in the research, provides additional additional support support for the idea that multi-level cognition research, cognitive processing is implicated in the generation of of emotion. Among other issues, a multi-level framework could account for phenomena such as attention-deployment involved in the selection selection of of relevant information, the levels of of appraisal to particular particular events, the interaction of of allocation of higher levels lower, automatic automatic and higher, conceptual or controlled controlled processes, and the resulting activation of of bodily responses responses.. Together with the critique of of classical appraisal notions, notions, as outlined in the beginning of of the chapter, it of cognitive becomes imperative to integrate these advances in our knowledge of processing into appraisal theory. In what what follows, two central issues for the of multi-level multi-level appraisal appraisal theories will be discussed. discussed. development of Towards general process models oof fappraisal. Recently, Smith and collaborators (Smith, Griner, Kirby, & Scott, 11996) 996)
basic concepts suggested by have reported an ongoing effort to reformulate basic of an Leventhal and Scherer ((1987) 1 987) by constructing a preliminary version of appraisal process model model.. They emphasize the function of of emotion to motivate appraisal and regulate attention and argue that since emotion alerts the organism to
relevant information, the scanning or monitoring of the environment must attention; if if this evaluative scanning process were attentional, attentional, occur outside attention; then the demand upon attentional resources would be too high, restricting the resources available for ongoing processes. This implies that appraisal models
elicitation of of emotion as should consider non-attentive mechanisms for the elicitation well as the encoding encoding of of event event-specific of -specific information for the purposes of appropriately motivating the individual for action. In the model, the authors of processing, namely schematic and conceptual conceptual.. emphasize two modes of Schematic processing is described to be fast, automatic, and functioning in a parallel fashion, yet inflexible and and relatively concrete. The conceptual mode parallel on the other hand, is expected to function serially, to be slower, under voluntary control and therefore flexible, and relies upon semantically information. accessible information. As in the the case of of the automatic/controlled distinction made by Shiffrin (e.g. 11977) and Schneider (e.g. 977) and in general agreement with Leventhal and independent Scherer ((1987), 1 987), the two modes proposed by Smith et al. are not independent of one another, but are assumed to interact interact in several ways. Most important of
278
Chapter 6
in this context is the the creation of of schemata within the conceptual mode through the process of of learning from experience. Conversely, when existing schemata are sufficiently activated activated by incoming sensory data, the information they represent might become available to conceptual processing. appraisal information output from the conceptual and schematic The appraisal modes, in addition to perceptual perceptual information, is described to be monitored by a a so-called appraisal register. This register combines appraisal information from these three input modes, on the basis of of which a specific emotional output is generated. The register is not considered as an active processor which computes information, but rather rather as an information detector and response selector which combines the appraisal outcomes and initiates initiates the response accordingly. Attention is seen as being regulated by the registering of a subjective subjective feeling state, when sufficiently intense, into conscious of awareness. In this mann er, conceptual processes are called into play, in order manner, to further analyze the significance of of the stimulus and to refine the emotional response accordingly. While the model suggested by Smith et al. is still preliminary and so far lacks both detail and precision (in particular with respect to the notion of a passive "appraisal "appraisal register"), it constitutes an important effort to move appraisal theory from largely structural to dynamic process modeling. In addition, this effort is likely to result in greater convergence of of appraisal research with other branches of of emotion and cognition research. research
top-downprocessing Bottom-up vs. top-down dOails Apart from developing general process models that describe the details of information processing on different levels, the obvious interaction between of the levels needs to be theor~ieally theoretically conceptualized in a more stringent fashion. As has has become apparent in the discussion discussion above, the the issue of of bottom-up vs. of the central concerns in this respect (see top-down processing is one of Leventhal & Scherer, 987, pp. 2 1 -23). Soberer, 11987, 21-23). of leDoux, LcDoux, Ohman, 0hman, and D Damasio of pre-attentive pre-attemive amasio on the role of The work of of or amygdaloid processing together with general advances in the field of attention and emotion indicates that that a bottom-up process is involved in the evaluation of of the significance of of sensory stimuli. This process is responsible organism that that attention needs to be focused upon an for signaling to the organism order to further further specify the results results of of the rather coarse, external event in order unspecific analysis performed at the lower level. At the same time, the lower preparation for action, by activating activating the ANS AN S level initiates a rudimentary preparation
C.M. van van Reekum Reekum and and KR. K.R. Scherer Scherer CM
279
and and SNS SNS in in aa rather rather general, general, non-specifi non-specificc manner. manner. However, However, the the propositions propositions of of the the authors authors mentioned mentioned above, above, although although firmly firmly based based on on empirical empirical evidence, evidence, have have thus far far only only been been tested tested for for one one emotion, emotion, namely namely fear. fear. Research Research on on implicit implicit perception of of affective affective information, information, even even though though it it is is currently currently limited to to addressing addressing positive-negative positive-negative valence distinctions, distinctions, also suggests that that affective affective information information is is evaluated evaluated in aa highly automatised, automatised, largely pre preattentive attentive manner. manner. It It thus thus seems seems that that relevance relevance checking checking might might precede precede and and prepare, prepare, in in aa bottom-up bottom-up fashion, fashion, other other appraisal appraisal processes. processes. While there there is is good good reason reason to to assume assume bottom-up bottom-up processing processing of of the the While significance of an an event, event, there there is is evidence that that top-down top-down processes processes are are signifi cance of involved in in the the priming priming of of the relevance relevance "detectors" "detectors" as as well as as subsequent subsequent involved modification 1996) notes modification of the the emotional emotional response. response. LeDoux LeDoux ((1996) notes that that the the amygdala is not only only activated activated by bottom-up bottom-up processes processes evaluating the the significance incoming simple sensory stimulus, also receives significance of of an an incoming simple sensory stimulus, but but that that is is also receives information information from various other other parts parts of of the the cortex. cortex. In consequence, the the appraisal performed performed at at low low levels levels might might well well be be influenced influenced by by higher higher level level processes such as as the the monitoring monitoring of of the current current concerns and and goals goals of of the the processes individual. In In fact, fact, evidence in in the the field of attentional attentional bias bias (see above) above) individual. field of suggests that that the the individual actively actively inhibits any any information information not not relevant relevant to its current concerns or goals. Also, according to Kosslyn and Koenig ((1995), 1 995), of as being the interface between bottom-up and working memory is thought of top-down processing. LeDoux ((1996) 1 996) has suggested that the lateral prefrontal cortex, presumably presumably involved in in working working memory, plays aa role role in in the the selection selection of of relevant stimuli. Thus, whatever is active in working memory determines processing. which stimuli will be selected for further processing. doubt that in emotion-antecedent appraisal, both There can be little doubt bottom-up and and top-down top-down processes processes are are relevant. relevant. Low Low level level appraisal appraisal can can bottom-up prime higher-level higher-level appraisal appraisal by by bottom-up bottom-up processing processing of of potentially potentially relevant relevant stimuli (e.g. (e.g. by by attention attention focusing focusing or or activation activation spreading spreading to to related related information in long-term memory). Also, low-level appraisal might be primed by higher level level processes. Leventhal and Scherer ((1987, p.. 221) 1 987, p 1) suggest that of conceptual processing may sensitize classes of of schemata top-down effects of relevant to to situations that that are are related related at at aa more more abstract abstract or semantic semantic level. level. relevant of hearing a noise that fails to match the schemata They provide the example of of noises normally encountered upon entering one's home, with the result of of of priming sensory-motor reactions and strengthening the fear response. Frijda provides aa similar similar example: example: "Emotions indeed indeed rarely rarely have have the the one one((1993) 1 993) provides shot, immediate, immediate, and and fast fast character character of of the the paradigmatic paradigmatic case case of of being being startled startled shot, by a crackle in the solitary woods. And even there one has been walking
280
Chapter 6
around with all sort of of expectations against which the crackle is perceived" (p. 382). Thus, the implications of of being alone in the woods facilitate or enhance low level appraisal of of the crackle by allotting more than the usual of attention to the stimulus. share of Also, one's self self concept (which presumably relies upon episodic or 992) or explicit memory; see Schaeter, Schacter, 11996, 996, and Johnson & Multhaup, 11992) personality (which is based more upon implicit processes; Schacter, 1996) influence upon upon what information is selected and further processed have an influence and, more importantly, how this information is appraised. These personality dispositions can thus influence, in a top-down manner, what information will be attended to and how this information will be further processed and appraised. This important, and hitherto largely neglected, field of of appraisal next section. research is discussed in the next
Differences in Appraisal Processes Individual Differences One of of the the paradoxes in the the area of of appraisal research is the lack of of concern with individual individual differences in perceiving the same type of of event. Appraisal theory partly started with Lazarus' ((1968) 1 968) insistence on the "transactional" nature of of appraisal, linking the objective event and the subjective appraisal, strongly affected by the perceived coping ability of of the individual. individual. In consequence, one might have thought that appraisal theorists were particularly interested in individual difference factors that that can explain divergences in appraisal outcomes under similar objective conditions. Indeed, most appraisal theorists will stress that appraisal is highly subjective and depends on the individual's perception and evaluation of events rather than their objective characteristics, predicting that the resulting emotion will be determined by the subjective interpretation. Yet, there has been little effort to more systematically identify stable individual individual traits that might predispose persons to show systematic appraisal tendencies or even biases in the appraisal process. Among the exceptions has been Scherer's ((1987) 1 987) suggestion of conceptualizing different types of of emotional disorders on the basis of of appraisal malfunctioning malfunctioning (see also Kaiser & Scherer, in press). The underlying assumption is that although appraisal is subjective and may vary from individual to individual, it must remain - within certain limits appropriate to the objective situation (e.g. through reality testing) and to the coping potential that is commonly perceived to be within the individual's means. Violation of of these appraisal reality constraints, constraints, as one might call
C.M. CM
K.R. Scherer van Reekum and K.R.
28 2811
them, will lead to the resulting emotion being being considered as abnormal or them, disordered, at least by an individual's social environment, environment, if not by Seherer ((1987) that one particular him/herself. For example, Scherer 1987) has suggested that form of depression, helplessness, might be partly due to a consistent of one's coping potential. It is important to stress that an underestimation of individual who truly lacks the means to deal with a particularly difficult situation would be described as "dejected" whereas someone described as situation, particularly "depressed" is implicitly assumed to appraise the situation, one's coping potential, in an inappropriate, disturbed fashion. Scherer (in press) reviews some of the individual difference factors that are likely to systematically affect emotion-antecedent appraisal. The following overview summarizes some of the major suggestions. First, theoretically postulated individual differences in both formal, process-related characteristics and content-related differences will be reviewed. appraisal characteristics the personality characteristics that might underlie the tendency to Second, the are discussed, including pertinent findings show systematic appraisal biases are in the literature.
form Appraisal biases with respect to the f orm or process oof fappraisal One of of the most basic formal variables in appraisal is the speed of of the emotion-antecedent appraisal processes. It is possible that there are potential differences with respect to to general processing speed in the central nervous (CNS) as well as other other factors that that produce consistent differences in system (eNS) the speed of appraisal, resulting, for example, in differential onset of the of appraisal, of reaction patterns. A related factor is the relative predominance of of automatic vs. be consistently different across individuals controlled processing which might be (see below). Individuals may may also differ with respect to to the the thoroughness thoroughness or completeness of of the appraisal a particular event. For example, it is appraisal following a conceivable that one individual fairly rapidly arrives at at a specific appraisal that one appraisal result, whereas another result, whereas another one engages in repeated repeated re-appraisals re-appraisals before settling on an an emotion-specific action action tendency. These These formal or process aspects of of the degree of that is expended, the appraisal may be be related related to the the degree of cognitive cognitive effort effort that which in turn linked to which tum is linked to possible possible differences with respect to potential tendencies to to engage in relatively effortful controlled controlled processing on aa conceptual often described conceptual level (a (a tendency often described as "intellectualizing"). Another issue concerns the relative complexity of Another related related issue of the analysis within the appraisal, i.e. gross vs. more fine-grained appraisal, the appraisal, gross appraisal, aa dimension
282
Chapter 6
directly related to the width of of the categories used in inference and of the major criteria in classification. For example, causal attribution, one of appraisal, might be evaluated quite differently depending on whether broad or narrow categories are used for assigning responsibility and inferring intentionality. A factor related to the breadth of category might be the absolute or temporary (interim) character of of attributions (also linked to the readiness to engage in re-appraisals). Major individual differences could exist with respect to vigilance, i.e. i.e. the detection of of events that are marginally pertinent to an individual and the nature of of the attention deployment strategies used. Whereas a highly vigilant of individual would be likely to have a very low threshold for the activation of more elaborate appraisal processes following detection of of a potentially relevant stimulus, coupled with immediate direction of of attention to this new stimulus, another, a less vigilant individual might not even notice a comparable stimulus. This suggests that there may be differential levels of of reactivity to implicit or subliminal stimuli. This dimension could also be be linked to the degree of of feedforward feexlforward guidance, or bottom-up processing of of higher-level conceptual processing. Similarly, there may be systematic differences in the degree of of top-down control of of lower level processing, such as sensitization for schema formation or automatization. Thus, the issue of of levels of of processing might be of of fundamental importance for individual differences in appraisal, not only with respect to systematic preferences for certain processing levels but also relative to the kind and frequency of of interactions between the levels.
Appraisal Appraisal biases biases with respect to content Individuals may also differ with respect to tendencies to lean more towards one rather than another direction in appraising events with respect to certain dimensions, i.e. to attribute responsibility to oneself rather than others, or underestimate or overestimate one's power. Such biases concern the content of Scherer ((1987, 1 987; Kaiser & Scherer, of appraisal more than its form. Soberer Soberer, in press) has proposed a classification of of such content-related appraisal biases and their potential contributions to clinical syndromes. The main points of of this classification will be summarized below, bdow, using Scherer's stimulus evaluation check 1; for further chock (SEC) appraisal criteria (Table 1, details consult Schcrcr, Scherer, 1984a, 11988, 988, 1993b). With respect to to the Novelty check cheek one might expect differences, possibly of the ANS and CNS, of linked to fundamental characteristics of eNS, on speed of
CM C.M. van Reekum and and KR. K.R. Scherer
283
habituation and/or extent extent of of inhibition. inhibition. Slow Slow habituation habituation and and lack lack of habituation inhibition could result result in in oversensitization ovcrscnsitizationand overvigilance ovcrvigilance with respect respect to inhibition incoming stimulation. stimulation. If If extreme, extreme, this this might correspond to pathological pathological frequent and massive startle startleresponses. responses. nervousness or jumpiness, provoking frequent rapid habituation habituation or strong strong inhibition inhibition potential potential might Conversely, very rapid correspond to a more more sluggish sluggish appraisal appraisal system, system, requiring requiring fairly fairlystrong strong and stimulus characteristics characteristics to diagnose novelty novelty or change. change. Again, unusual stimulus clinicalsyndromes syndromes of stupor stupor or lethargy. lethargy. extreme forms might produce clinical On the Intrinsic Intrinsic Pleasantness Pleasantness check check one can imagine systematic systematic On differences with respect respect to hedonic hcdonic evaluation, evaluation,due to differences differencesin in the tuning differences of valence valence detectors, detectors, unusual learning learning experiences, experiences, or specific specific states. states. The The anhedonia anhcdonia often often found in in depressive depressive patients patients provides a clinically clinically relevant relevant example. Several subchecks of of the of of Goal Conduciveness appraisal can be the appraisal appraisal of of relevance or affected by biases. A central issue concerns the pertinence of of stimuli or events with respect to one's needs and goals. This pertinence of motivational may depend partly on individual differences in the intensity of of strong concerns striving - individuals with weak drive structure and lack of of events, whereas strong strongmay systematically underestimate the pertinence of willed individuals with intense motivational impetus might show a tendency to see relevance everywhere accompanied by excessive personalization. appraisal might also be due to lack of of ability to evaluate Biases in relevance appraisal consequences of of and establish links between events, resulting either in overassimilation and overgeneralization, or in lack of of concern or caring. The frontal lobe lesion syndromes described above are good examples for clinically relevant cases, yielding blandness and and lack of of ego-involvement. clinically Paranoia, the other extreme where any any event is Paranoia, obviously, is an example of of the personalized. personalized. Another Another aspect aspect of of the goal conduciveness check concerns the appraisal the estimation of of the the likelihood that appraisal of of outcome probabilities, probabilities, the that certain consequences will follow certain events. The optimism-pessimism personality dimension dimension provides an example for systematic biases in this context Carver, 1985). & Carver, 1 985). Systematic biases may may also occur occur with context (Scheier (Scheier & respect respect to to the the conduciveness judgment judgment itself, i.e. i.e. the estimation estimation to to which extent aa certain state state of of affairs affairs will help or or hinder hinder attainment attainment of of aa goal or or satisfaction individuals lacking satisfaction of of aa need. For For example, perfectionists perfectionists or individuals realism underestimate the realism may may well systematically underestimate the conduciveness of of events events or or states might be individuals considered to states of of affairs. The The opposite opposite might be true true for for individuals to be naive could be expected naive or or overly overly credulous. credulous. Major Major differences could expected for the Urgency appraisal, appraisal, depending on time perspective perspective and assessment assessment of of
284
Chapter 6
consequences, and and giving rise to either overreaction and panic or apathy or passivity. Some of of the best documented individual differences are found for the Coping potential appraisal, appraisal, particularly the Causal attribution subcheck. The extemaVinternal external/internal control personality dimension is a prime example for powerful biases with respect to causality or responsibility assessment. For example, overestimation of of self self agency has been at at the core core of of cognitive theories of of depression etiology (Abramson, Seligman, & Teasdale, 11978; 978; Teasdale, 11988). 988). Other potential biases include faulty ability or effort attribution. Similar biases may occur with the appraisal of of intention. Again, paranoia is a ease case in point where overattribution of of intention is frequent. Important differences between individuals can be expected for perceived Control, i.e. the degree to which certain events or consequences are under human control, independent of the amount of independent of of power that that is available to the person. Some individuals may have illusions of of control or may overestimate control potential whereas others may systematically denigrate the possibility of of control, the extreme case ease being the clinical syndrome of of hopelessness. The Power check is one of of the most important subchecks of of the Coping Potential Potential appraisal appraisal since its result often ot~en powerfully determines the nature nature of of the emotion that is elicited. In many cases, eases, the result of of the power appraisal appraisal may make the difference between fear and anger responses. Given the highly subjective nature of of power appraisals, and the many dimensions that that are involved (e.g. types of of power such as physical force, money, knowledge, ability to recruit help), this appraisal appraisal dimension is a prime candidate candidate for individual differences and biases. Such systematic misjudgments of one's own or other persons' power can earl be due to to a large number of of different personality characteristics characteristics and cognitive style variables (see below). Most important, is the person's self-image, including persistent tendencies to over overor underestimate one's power or ability to influence a given situation. The helplessness syndrome as described by Seligman and his collaborators (Abramson et at, 978; Peterson & Seligman, 11984) 984) is a classic example in al., 11978; the clinical literature. A final check related to Coping Potential is the appraisal of of Adjustment, the estimated capacity to accept and live with the consequences of of a given situation. Clearly, individuals tending toward fatalism, espousing more passive life philosophies, will appraise the respective situations very differently from people with more deterministic philosophies. The The Compatibility with Standards check concerns the relationship of of one's own or someone else's behavior to both external (norms, morals) and
C.M C.M. van van Reekum Reekum and and K.R. Scherer
285
internal internal standards. standards. Potential Potential sources sources for for systematic systematic biases biases might might be be undersocialization, undersocialization, normlessness, normlessness, amorality, amorality, or or asociality, asociality, all all tending tending toward toward minimizing behavior-norm discrepancies, thus reducing reducing guilt minimizing the the appraisal appraisal of of behavior-norm discrepancies, thus guilt potential and and producing, producing, in in extreme extreme cases, cases, shamelessness shamelessness and and antisocial antisocial potential behavior. behavior. Conversely, Conversely, high high moral moral and and ethical ethical standards standards would would tend tend to to augment augment perceived perceived discrepancies discrepancies and and might might lead lead to to exaggerated exaggerated feelings feelings of of shame shame or or guilt guilt which which could could become become part part of of aa neurotic neurotic syndrome. syndrome. With With respect respect to to internal internal standards, standards, the the major major individual individual difference difference variable variable would would be be the the definition and and the the strength strength of of the the ideal ideal self. definition self.
Potential sources f or individual diff erences in appraisal tendencies for differences The The preceding preceding two two sections sections have have presented presented hypotheses hypotheses as as to to possible possible systematic systematic differences differences between between individuals individuals with with respect respect to to the the form form and and content content of of the the appraisal appraisal process. process. In this this section, section, we we will will look look at at this this issue issue from from another another vantage vantage point point -- the the likelihood likelihood that that dimensions dimensions of of individual individual difference difference and and personality personality that that have have been been shown shown in in the the literature literature to to have have powerful powerful effects effects on on behavior behavior and and cognition cognition may may systematically systematically influence influence appraisal processes. processes. This This review review is is organized organized by by the the nature nature of of the the underlying underlying appraisal individual difference difference dimension, dimension, here here broadly broadly classified classified into into organismic organismic individual predispositions, styles, and personality traits. predispositions, cognitive cognitive styles, and personality traits. Organismic predisposition. At At the the very very lowest lowest level level of of individual individual Organismic differences factors factors are innate characteristics characteristics of of the the CNS and/or ANS AN S.. As As differences are innate CNS and/or mentioned above, above, rapidity rapidity of of habituation habituation and and efficiency efficiency of of inhibition inhibition are are mentioned among the the candidates in in this this category. category. There There is is reason reason to to assume that that among differences in in efficiency efficiency of of inhibition and and interference interference sensitivity sensitivity might might be be differences related to to working working memory memory capacity, capacity, cognitive cognitive strategy strategy use, and and knowledge knowledge related (Hamishfeger & Bjorklund, Bjorklund, 1994). 1994). Lord Lord and and Levy Levy ((1994) have attempted attempted to to (Harnishfeger 1994) have show that that connectionist-level connectionist-level mechanisms mechanisms related related to to activation activation and and inhibition inhibition show can model model the the information information selection selection needed needed to to provide provide continuity continuity in in thoughts thoughts can and may may explain explain individual individual differences differences in in selective selective attention attention capacities. capacities. and Another important important variable variable might might be be general general speed speed of of processing. processing. Hale Hale and and Another Jansen ((1994) have shown that that individual individual performance performance on on diverse diverse tasks tasks can can Jansen 1 994) have be predicted on on the the basis basis of of aa single processing-time processing-time coefficient. coefficient. Another Another be CNS variable variable of of interest interest for for differential differential questions questions is habitual cortical arousal arousal CNS (Matthews & Amelang, Amelang, 1993; 1993; Matthews Matthews & Harley, Harley, 1993). 1993). All All of of these these (Matthews cognitive mechanisms mechanisms are are likely likely to to be be directly directly involved involved in in appraisal appraisal cognitive processes, particularly particularly at at the the lower, lower, more more automatic automatic levels. levels. processes,
286
Chapter 6
Cognitive styles. Another Another classic classic area area of of individual individual differences on aa Cognitive differences on fairly of processing are persistent the styles fairly low low level level of processing are persistent cognitive cognitive styles. styles. Among Among the styles that likely to have bearing on appraisal processes processes are are holistic that are are likely have aa bearing on appraisal holistic vs. vs. analytic processing (Dunn, (Durra, Dunn, Dunn, Andrews, Languis, 11992), parallel vs. 992), parallel vs. analytic processing Andrews, & Languis, 995), field dependence (Messick, serial processing Fuste, & Vidal, serial processing (Tous, (Tous, Fuste, Vidal, 11995), field dependence (Messick, 1). 11994), 994), overgeneralization, overgeneralization, and and personalization personalization (Dritschel (Dritschel & Teasdale, Teasdale, 199 1991). The that the to The latter latter authors authors suggest suggest that the persistent persistent differences differences in in response response to emotional events by the the traits traits of of affect and neuroticism emotional events measured measured by affect intensity intensity and ncurotir may particular styles styles of cognitive processing, processing, aa relationship relationship may be be mex~ated mediated by by particular of cognitive that is conceivably influenced by in the the emotion-antecedent emotion-antecedent that is conceivably influenced by differences differences in appraisal appraisal processes. processes. Directly pertinent pertinent for for appraisal appraisal is is the certain ways Directly the tendency tendency to to use use certain ways of of categorizing linked to the the classic of cognitive categorizing events, events, linked classic issue issue of cognitive complexity complexity (Sommers, Schroder, Driver, Streufert, 11967; Tetlock, 11983; Harris, (Sommers, 11981; 98 1 ; Schroder, Driver, & Streufert, 967; Tetlock, 983; Harris, Press, Crockett Crockett,, & Delia, Delia, 1975). Among the possible variables are 1 975). Among the possible variables are 11981; 98 1 ; Press, wide vs. vs. narrow narrow category category use, use, rigidity, rigidity, or need for structure. For example, wide or need for structure. For example, Neuberg and Newson ((1993) 1 993) demonstrated demonstrated how individual differences differences in Neuberg and Newson how individual in the the desire for simple structure structure may may influence influence how how people people understand, experience, desire for simple understand, experience, and interact with with their their worlds. worlds. Participants high in and interact Participants high in Personal Personal Need Need for for Structure were especially to organize Structure were especially likely likely to organize social social and and nonsocial nonsocial information information in ways, stereotype others, and their research in less less complex complex ways, stereotype others, and complete complete their research requirements on time. time. Chronic motives may may obviously obviously requirements on Chronic information-processing information-processing motives strongly aspects of of the the appraisal strongly influence influence formal formal aspects appraisal process. process. A variable variable that much attention attention in that received received much in social social cognition cognition research research is is need for nee~ for cognition cognition (Cacioppo, (Cacioppo, Petty, Petty, Feinstein, Feinstein, & Jarvis, Jarvis, 1996). 1996). Waller Waller (1994) showed that low need for showed that low need for cognition-participants cognition-participants reported reported aa significantly greater greater reliance reliance on information (indicative (indicative of of significantly on normative normative information peripheral processing) processing) while in need for cognition reported aa peripheral while participants participants high high in need for cognition reported significantly greater effort (indicative (indicative of of central central significantly greater expenditure expenditure of of cognitive cognitive effort route data reported by Thompson, route processing). processing). Similarly, Similarly, data reported by Thompson, Chaiken, Chaiken, and and Hazlewood 1993) suggest need for Hazlewood ((1993) suggest that that (1) nee~ for cognition cognition involves involves intrinsic intrinsic motivation for for effortful effortful cognitive motivation cognitive processing, processing, (2) need need for for cognition cognition may may predict mainly in with minimal predict such such processing processing mainly in contexts contexts with minimal extrinsic extrinsic incentives incentives for and (3) control control motivation motivation may may be related both to to for processing, processing, and related causally causally both extrinsic undermining undermining effects effects and individual differences differences in need for extrinsic and to in&vidual in neexl for cognition. Clearly, Clearly, the in effortful effortful cognitive cognitive processing processing cognition. the disposition disposition to to engage engage in may likelihood of switching to higher levels levels of may greatly greatly determine determine the the likelihood of switching to higher of appraisal. Differences Differences on this dimension rather early. Berzonsky appraisal. on this dimension may may appear appear rather early. Berzonsky ((1993) 1 993) showed showed that adolescents with with an information-oriented, se1f-exploratory that adolescents an information-oriented, self-exploratory
CM CM. van van Reekum and KR. K.R. Scherer
287
identity style style were were significantly significantly more more motivated motivated to to engage engage in in active active identity information processing processing (need (need for for cognition) cognition) and and were were more more receptive receptive to to new new information ideas, personal personal feelings, feelings, and and actions actions (experiential (experiential openness), openness), even even though though this this ideas, might threaten threaten hard-core hard-core areas areas of of the the self, self, than than were were their their normative normative or or might diffuse/avoidant counterparts. counterparts. diffuse/avoidant Personality traits. Many Many classic classic personality personality traits traits such such as as extroversion, extroversion, Personality repression-sensitization, neuroticism, neuroticism, rigidity, rigidity, dysphoria, dysphoria, worrying, worrying, sensation sensationrepression-sensitization, seeking, or or openness, openness, can can be be seen seen as as pertinent pertinent to appraisal. appraisal. The The effects effects of of seeking, trait disposition disposition can can be be expected expected to to directly directly affect affect cognitive cognitive operations. operations. trait Sensation-seeking (Schroth (Schroth & Lund, Lurid, 11994) and openness openness (McCrae, (McCrae, 11993) 994) and 993) Sensation-seeking have also also been been shown shown to to correlate correlate with with cognitive cognitive processing processing preferences preferences that that have may play a major major role role in in appraisal, appraisal, such such as as creativity creativity and and intellectual intellectual may play a orientation. Pruzinsky Pruzinsky and and Borkovec Borkovec ((1990) found that that worriers worriers reported reported 1 990) found orientation. more negative negative daydreaming, daydreaming, greater greater difficulty difficulty with with attentional attentional control control and and more greater obsessional obsessional symptoms, symptoms, public public self-consciousness, self-consciousness, and anxiety. greater and social social anxiety. Worriers evidenced evidenceA significantly significantly more more negatively negatively affect-laden affect-laden cognitive cognitive Worriers intrusions during during relaxed relaxed wakefulness wakefulness and and aa focused focused attention attention task. task. Findings Findings intrusions reported by Lorig et al. al. ((1994) suggest that that repressors repressors exhibit exhibit an an absence absence of of reported by Lorig 1994) suggest cognitive activity activity when when faced faced with with the the recall recall of of negative negative memories. memories. cognitive Extroversion has been been frequently frequently studied studied with with respect respect to to cognitive cognitive Extroversion processing. Matthews Matthews and and his his collaborators collaborators have have been been examining examining relationships to to cognitive cognitive arousal (Matthews (Matthews & Amelang, Amelang, 11993; Matthews & relationships 993; Matthews Harley, Matthews, Davies Davies & Lees, Neurological differences Harley, 11993; 993; Matthews, Lees, 1990). 1 990). Neurological differences between extroverts and introverts, while hotly hotly debated, debated, continue continue to be between reported. Stenberg, Stenberg, Wendt Wendt and and Risberg Risberg (1993) from ( 1 993) confirmed findings findings from earlier studies of of cerebral blood flow (CBF) at rest showing higher blood flow flow in the the temporal temporal lobes for introverts than for extroverts, and and a negative negative correlation between extroversion and global CBF CBF among among women. Tobacyk, Driggers and Hourcade that the perceptual/cognitive Hourcade (1991) ( 1 99 1 ) suggest suggest that the perceptual/cognitive information processing processing preferences associated with with extroversion, intuition, and perceiving complement high self-monitoring, while while the the processing preferences associated with introversion, sensing, and judging sensing, judging complement preferences associated low self-monitoring. Of the form of Of direct relevance to to the the issue of of differences in the of the appraisal appraisal process process is the finding by Stelmack, Houlihan and McGarry-Roberts McGarry-Roberts (1993) ( 1 993) showing that that higher neuroticism scores were associated associated with faster faster P300 regarded as P300 latency, latency, aa measure measure that that is is regarded as an an index index of of stimulus stimulus evaluation evaluation time that that is independent independent of of response production. Contrary to what what one might expect on the basis of of this finding, finding, higher neuroticism scores were associated associated
288 288
Chapter 6
with slower reaction reaction time time (RT), (RT), aa measure measure that that is is also also regarded regarded as as an an index index of of with speed ormation processing. speed of of inf information processing. The The authors authors speculatively speculatively explain explain this this asynchronicity asynchronieity with with an an overly overly hasty hasty and and worried worried evaluation evaluation of of the the stimulus stimulus requiring ore response requiring further further processing processing bef before response initiation. initiation. This This explanation explanation fits fits
the notion notion of of oversensitization oversensitization to to changing stimulation mentioned mentioned above. above. the changing stimulation Furthermore, ective processes Furthermore, aff affective processes linked linked to to the the need neexl for for response response initiation initiation may slow down inhibit the may intervene intervene in in such such aa way way as as to to slow down or or inhibit the reaction. reaction. Apart Apart from from personality personality dispositions dispositions in in the the narrow narrow sense, sense, individual individual diff erences in also be specific attitudinal differences in appraisal appraisal could could also be due due to to specific attitudinal structures. structures.
For example, example, Zucker Zucker and and Weiner Weiner ((1993) showed that that conservatism conservatism correlated correlated For 1 993) showed positively positively with with aa belief belief in in
the the importance importance
of of individualistic individualistic
causes, causes,
controllability, controllability, blame, blame, and and anger, anger, and and negatively negatively with with perceptions perceptions of of the the importance importance of of societal societal causes, causes, pity, pity, and and intentions intentions to to help. help. Personal Personal help help was was emotionally are judgments emotionally determined, determined, whereas whereas welf welfare judgments were were directly directly related related to to attributions attributions of of responsibility responsibility and and political political ideology. ideology. Obviously, Obviously, these these and and
other types of of generalized generalized attitudes attitudes are are rather rather likely likely to to color color the the way way in in which which other types events events are are appraised. appraised. Another erence variables Another set set of of individual individual diff difference variables likely likely to to influence influence emotion-antecedent appraisal are emotion-antecedent appraisal are linked linked to to the the structure structure of of the the self, self, in in terms terms
of self or self image. For For example, example, individuals high on on self self-efficacy of self concept concept or self image. individuals high -efficacy (Bandura, 11982; Berry, 11989) are likely to to be be systematically systematically biased, across across aa (Bandura, 982; Berry, 989) are large range range of of situations, situations, towards towards auto-attributing auto-attributing higher higher internal internal control control and and large power those low on this dimension. In In addition, addition, they they are are likely likely to to power than than those low on this dimension. consistently check check outcomes outcomes against against their their internal internal standards. standards. Conversely, Conversely, consistently individuals with with more more fragile fragile selves selves might might have have the the opposite opposite tendencies, tendencies, individuals particularly in in social social contexts. contexts. Thus, Thus, Downey Downey and and Feldman Feldman ((1996) show that that particularly 1 996) show people who who are are sensitive sensitive to to social social rejection rejection tend to anxiously anxiously expect, expect, readily readily people tend to perceive, perceive, and and overreact overreact to to it. it.
Apart from from structural structural variables variables concerning concerning the the self, self, social social psychologists psychologists Apart have studied studied aa variety variety of of process process variables variables such such as as self-monitoring self-monitoring or or self selfhave handicapping. Again, Again, it it takes takes little little fantasy fantasy to to imagine imagine ways ways in in which which handicapping. individual diff differences on these these dimensions dimensions might might affect affect appraisal, appraisal, individual erences on
in in
particular with with respect respect to to assessing coping potential potential and and the the compatibility compatibility particular assessing coping with internal internal standards. standards. An An interesting interesting example example is is provided provided with
by Rhodewalt, Rhodewalt, by
Morf, Hazlett Hazlett and and Fairfield Fairfield ((1991) who examined examined the the eff effects of self self- Morf, 1 99 1 ) who ects of handicapping on on ability ability attributions attributions and and self-esteem. self-esteem. High High self self-handicapping handicapping -handicapping participants participants generally generally discounted discounted ability ability attributions attributions in in response response to to failure failure feedback. The The study study illustrates illustrates how how individual individual diff differences in motives motives to to feedback. erences in
eM C.M. van van Reekum Reekum and and K.R. K.R. Scherer Scherer
289
of engage in self-protective or self-enhancing behavior affects attribution of achievement and affect. In addition to these more more or less stable personality dispositions, momentary mood differences are are likely to systematically affect cognitive processes that are directly pertinent to emotion-antecedent emotion-antecedent appraisal (Basso, Scheffi, Schefft, & Hoffinann Hoffmann,, 1994; Isen, Niedenthal, & Cantor, 1992; Murray, Sujan, Hirt, & Sujan, 1990; Woodfield, Jones, & Martin, 11995). 995). This compilation of of potential individual difference dimensions which can be reasonably expected to affect appraisal processes shows the promise of of further theoretical efforts in this direction, accompanied by more focused empirical investigations. Such theoretical efforts will need to first systematize of the individual differences. One pertinent dimension is form the nature of versus content. While some of of the stable dispositions described seem to affect of the appraisal appraisal process as a whole, others seem to be specific to the nature of content such as the predominance of of certain schemata, certain values, or certain goals. A second pertinent dimension concerns differences with the "horizontal" organization of of differences by processing level - some of of the findings reported above seem to concern the way in appraisal functions on a particular level, for example the role of of cognitive style in category formation on the conceptual level. Finally, although there has been little direct evidence reported above, one may expect a "vertical" for this in the findings reported organization with respect to the levels, for example differences in the speed organization of "schematization" or "automatization" of of the appraisal and extent of processes. While work specifically dedicated to these issues has hardly begun, advances in this area may well provide important insights to our understanding of of the appraisal process as a whole. understanding Conclusions Conclusions
This chapter could do little more than survey a construction site. Deep holes are gaping everywhere but some of of the construction elements can already be found strewn all over the site. We attempted to show how appraisal theory, which we consider as a solid foundation for a future below the conceptual theoretical edifice, needs to add stories or levels (mostly below that has been reasonably consolidated so far). In drawing up the floor that blueprints for such future development we had, by necessity, to be quite of moving towards hierarchical, multi multispeculative in showing the promise of of appraisal, appraisal, buttressing our arguments with some of of the level models of existing approaches approaches in related areas. Similarly, we have been trying to outline
290
Chapter 6
ways in which individual individual differences can be made fruitful for an understanding understanding of of both appraisal biases and appraisal pathology, potentially leading to emotional disorder. Again, individual building blocks found in the literature come in handy to illustrate the way in which one might piece together a more systematic approach to predicting how neurophysiological neurophysiologieal individual towards certain and personality characteristics might predispose an individual modes of of appraisal with foreseeable effects on emotional experience. experience. This review of of the blueprints blueprints and some of the pertinent pertinent building blocks has hopefully demonstrated that future advances on the tower of appraisal will not materialize by a compartmentalized activity of the appraisal specialists involved: neuropsychologists, cognitive other crafts need urgently to be involved: scientists, social psychologists, and clinical researchers, researchers, to name but a few. As on any construction site, speed of building and the quality of site, the speeA of the resulting workmanship will will depend in large part on the cooperation of of the crafts that are immersed in this effort. References References
Abramson, L. Y., Seligman, 1 978). Learned Seligman, M. E. P., & Teasdale, Tcasdale, J. ((1978). LcameA of helplessness in humans: Critique and reformulation. Journal o f Abnormal Psychology, 87, 49-74. Vol.Jl :: Psychological Arnold, M. B. (1960). Emotion and personality. Vol. York: Columbia University Press. aspects. New York: Vol.2: Arnold, M. B. ((1960). 1960). Emotion and personality. Vol. 2: Neurological and physiological aspects. New York: Columbia University Press. Bandura, A. (1982). Self-efficacy mechanisms in human agency. American Psychologist, 337, 7, 122-147. Teasdalr J. D. ((1991). Barnard, P. J., & Teasdale, 1991). Interacting Cognitive Subsystems: A systemic approach to cognitive-affective cognitivc-affcctivc interaction and change. Cognition and -39. and Emotion, 5, 11-39. (1932). Bartlett, F. C. (1 932). Remembering: A study in experimental and social psychology. London and New York: Cambridge University Press. Scheflt, B. K., & Hoffinann, Hoffmann, R R. G. (1 (1994). Basso, M. R., R, Scheffi, 994). Mood-moderating cognition: Sometimes effects of affect intensity on cognition: Sometimes euphoria is not beneficial and dysphoria is not detrimental. f Personality and detrimental. Journal oof Social Psychology, 66, 363-368. Bcchara, A., Damasio, A. R., Damasio, H., & Anderson, S. W. ((1994). Bechara, R, D amasio, H, 1994). Insensitivity to future consequences following damage to human 5. prefrontal cortex. Cognition, Cogmaon, 50, 7-1 7-15.
CM C.M. van Reekum and KR. K.R Scherer
29 2911
something missing: missing: Some Some observations observations prompted by Berkowitz, L. ((1994). 1 994). Is something the cognitive-neoassociationist cognitive-n~associationist view view of anger and emotional emotional aggression. In L. R. Huesmann (Ed.), (Ed.), Aggressive behavior: Current perspectives (pp. 35-57). 35-57). New York: York: Plenum. Berry, D 987). The problem of implicit D.. C. (1 (1987). implicit knowledge. knowledge. Expert Systems, 4, 14415 1 . 144-151. Berry, 1. 1989). Cognitive J. M. ((1989). Cognitive efficacy across the life life span: span: Introduction Introduction to the special special series. Developmental Psychology, 25, 683-686. Berzonsky, 1 993). Identity Berzonsky, M. D. ((1993). Identity style, style, gender, gender, and social-cognitive social-cognitive reasoning. Journal o fAdolescent Research, 8, 289-296. of Bower, G. H. ((1981). 198 1). Mood and memory. 29memory. American Psychologist, 36, 11291148. 48 . Bower, 1 992). How might Bower, G G.. H H.. ((1992). might emotions emotions affect learning? learning? In S.-A S.-A. Christianson Christianson (Ed.), (Ed.), The handbook of of emotion and memory: memory: Research and theory (pp. 3-3 1). Hillsdale, NJ: Lawrence 3-31). Lawrence Erlbaum. Cacioppo, J. 1. T., Petty, 1. A., A, & Jarvis, W. B. G. ((1996). 1 996). Petty, R. E., Feinstein, Feinstein, J. Dispositional differences differences in cognitive cognitive motivation: motivation: The life life and times times of individuals varying in need need for cognition. cognition. Psychological Bulletin, 119, 119, 1197-253 97-253 S.-A. ((1992). Remembering emotional emotional events: events: Potential Potential Christianson, S.-A 1 992). Remembering S.-A. Christianson (Ed.), (Ed.), The handbook oof emotmn and f emotion mechanisms. In S.-A (pp. 307-340). 307-340). Hillsdale, Hillsdale, NJ: Lawrence memory: Research and theory (pp. Erlbaum. Conway, M. A., A, & Bekerian, Bekerian, D. A 1 987). Situational A. ((1987). Situational knowledge knowledge and 145-191. 91. emotions. Cognition and Emotion, 1I,, 145-1 Descartes'' error: error: Emotion, research and the human Damasio, A. A R. ((1994). 1 994). Descartes York: Avon. Avon. brain. New York: emotions. Psychological De Rivera, J. 1. ((1977). 1 977). A structural theory of the emotions. Issues, 11O, 0, No 4, Monograph 40. Feldman, S. I. ((1996). of rejection sensitivity for Downey, G., & Feldman, 1 996). Implications of of intimate relationships. Journal o f Personality and Social Psychology, 770, 0, 11327-1343. 327-1 343 . B.. H., & Teasdale, J. D.. ((1991). Individual differences differences in affect affect1. D 1991). Individual Dritschel, B related cognitive operations elicited experimental stimuli. related cognitive elicited by experimental stimuli. British of Clinical Psychology, 30, 1151-160. Journal of 5 1 - 1 60. Dunn, Dunn, D D.. A, A., Andrews, Andrews, D., & Languis, M. L. ((1992). Dunn, B. R., Dunn, 1 992). Metacontrol: A cognItive cognitive model model of brain functioning functioning for psychophysiological study of complex complex learning. learning. Educational Psychologist, 227, 455-471. Psychologist, 7, 455-47 1.
292
Chapter 6
Ellsworth, P. C., & Smith, C. A. (1988). ( 1 988). Shades Shades of ofjoy: Patterns of of appraisal appraisal and Emotion, 2, 301-331. 30 1-33 1 . differentiating pleasant pleasant emotions. Cognition and difforentiating Folkman, S., & Lazarus, Lazarus, R. S. S . (1985). ( 1 985). If If it changes changes it must must be aa process: process: Folkman, Study of of emotion and coping coping during three stages of of aa college examination. examination. Journal Journal of of Personality and Social Psychology, 48, 1501 50170. 1 70. Folkman, as a mediator of of emotion. Folkman, S., & Lazarus, R. S. (1988). ( 1 988). Coping as Journal o f Personality and Social Psychology, 54, 466-475 . Journal of and 466-475. H (1986). ( 1986). The emotions. Cambridge and Frijda, N. H. and New New York: Cambridge University Press. Frijda, N. H. and action tendency. H (1987). ( 1 987). Emotion, cognitive structure, and and Emotion, 1, 115-143. 1 15- 1 43 . Cognition and H . (1993). ( 1 993). The place of of appraisal iinn emotion. emotion. Cognition Cognition and Frijda, N. H. 357-387. Emotion, 7, 357-387. H, Kuipers, P., & ter Schure, E. ((1989). 1 989). Relations among Frijda, N. H., emotion, appraisal, and emotional action readiness. Journal Journal of of 12-228. Personality and Social Psychology, 57, 2212-228. K R. (1988). ( 1 988). Relating situation evaluation to Gehm, T. L., & Scherer, K. emotion differentiation: Nonmetrie of cross-cultural Nonmetric analysis of questionnaire questionnaire data. In K. K R. Scherer (Ed.), Facets Facets of of emotion: Recent 6 1 -78). Hillsdale, Hillsdale, NJ: Erlbaum. research (pp. 61-78). Graf, P., & Schacter, D. L. ((1985). 1 985). Implicit and explicit memory for new associations in normal subjects and amnesic patients. Journal of of 501Memory, and Cognition, Cognition, 111, 1Experimental Psychology: Learning, Memory, 1 , 50 5518. 18 . processing-time coefficients Hale, S., & Jansen, J. ((1994). 1994). Global processing-time speed. characterize individual and group differences in cognitive speed. Psychological Science, Science, 5, 384-389. Halgren, E. ((1992). neurophysiology of of the amygdala within the 1992). Emotional neurophysiology of human emotion. emotion. In J. P. Aggleton (Ed.), The amygdala: context of ysfunction Neurobiological aspects aspects of of emotion, emotion, memory, memory, and mental mental d dysfunction (pp. 1191-228). 9 1 -228). New York: Wiley-Liss. Hamishfeger, K K. K, K., & Bjorklund, D. F. ((1994). 1994). A developmental Harnishfeger, inhibition. Special Issue: perspective on individual individual differences in inhibition. Developmental perspectives on individual differences in learning and memory. Learning and Individual Diff erences, 6, 3331-355. 3 1 -355 . Differences, Hams, R. M. (1981). Conceptual complexity and preferred coping strategies Harris, in anticipation of temporally predictable and unpredictable threat. Journal of of Personality and Social Psychology, Psychology, 41, 41, 380-390
CM C.M. van Reekum Reelcum and K.R. K.R. Scherer
293
Isen, A. M., Niedenthal, P. M., & Cantor, N. ((1992). 1 992). An influence of of positive 6, 65-78. affect on social categorization. Motivation and Emotion, 116, ( 1 983). A multiple-entry, modular memory system. In O. Johnson, M M.. K K. (1983). G. H. f learning and motivation: Advances in Bower (Ed.), The psychology o of 7, pp. 11-26). -26). New York: Academic Press. research theory (Vol. 117, 1 992). Emotion and MEM. Johnson, M. K, K., & Multhaup, K K. SS.. ((1992). MEM. In S.-A. Christianson (Ed.), The handbook o f emotion of emoaon and memory: Research and Theory (pp. 33-66). Hillsdale, NJ: Erlbaum. Kaiser, S., & Scherer, K. R. (in press). Models of of 'normal' emotions applied to facial and vocal expressions in clinical disorders. In W. F F.. Flack, Jr. psychopathology. New New York: Oxford & J. D. Laird (Eds.), Emotions in psychopathology. University Press. 1 990). Fundamentals of Kolb, B., & Whishaw, I. Q. ((1990). of human neuropsychology (3rd ed.). New York: Freeman. 1 995). Wet mind: Kosslyn, S S.. M., & Koenig, O. ((1995). mind: The new cognitive neuroscience. neuroscience. New York: Free Press. Kunst-Wilson, W. R., R, & Zajonc, R. B. ((1980). 1 980). Affective discrimination of of stimuli that cannot be recognized. Science, Science, 207, 557-558. Lang, P. J., Bradley, M. M., & Cuthbert, B 1990). Emotion, attention, B.. N. ((1990). 377-395.. and the startle reflex. Psychological Review, ReView, 97, 377-395 mindlessness of of 1 978). The mindlessness Langer, E. J., Blank, A., & Chanowitz, B. ((1978). of "placebic" information in ostensibly thoughtful action: The role of interpersonal interaction. Journal o fPersonality and Social Psychology, of 36, 635-642. R. S. ((1968). Lazarus, R 1 968). Emotions and adaptation: Conceptual and empirical Symposium ymposium on Motivation relations. In W. J. Arnold (Ed.), Nebraska S (Vol. 116, 6, pp. 175-270). 1 75-270). Lincoln, NE: University of of Nebraska Press. Lazarus, R. S. ((1984). 1 984). Thoughts on the relation between emotion and R Scherer K. R. Soberer & P P.. Ekman (Eds.), Approaches to emotion cognition. In K (pp. 247-258). Hillsdale, NJ: Erlbaum. Lazarus, R. S 1991). Emotion and adaptation. S.. ((1991). adaptation. New York: Oxford University Press. 1 988). Knowledge and appraisal in the Lazarus, R. SS.,. , & Smith, C. A. ((1988). 281-300. 8 1 -300. cognition-emotion relationship. Cognition and Emotion, 2, 2 of affective LeDoux, J. E. ((1986). 1 986). Sensory systems and emotion: A model of 237-248.. processing. Integrative Psychiatry, 4, 237-248 LeDoux, J. E. ((1989). leDoux, 1 989). Cognitive-emotional interactions in the brain. Cognition and Emotion, 3, 267-289.
294 294
Chapter 6
LcDoux, LeDoux, J. 1. E. (1993). ( 1 993). Emotional Emotional networks networks in the brain. brain. In M. Lewis Lewis & J. 1. Handbook of of emotions emotions (pp. 109-118). 1 09-1 1 8). New New York: (Eds.), Handbook Haviland (Eds.), Guilford Guilford Press. Press. LeDoux, J. 1. E. (1994). ( 1994). Emotion, memory, and and the the brain. Scientific SCientific American, American, LcDoux, June, 32-39. LeDoux, J. 1. E. (1996). ( 1996). The emotional emotional brain: The mysterious mysterious underpinnings underpinnings of ofemotional emotional life. life . New York: Simon and Schuster. Leventhal, Leventhal, H. H (1979). ( 1979). A perceptual-motor processing model of of emotion. In P. Pliner, study of of Pliner, K. Blankstein, Blankstein, & I. M. Spigel (Eds.), (Eds.), Advances in the study communication and and affect: affect: Perception of of emotion in self self and and others communication (Vol. 5). New York: Plenum. Leventhal, motor th~ry Leventhal, H. (1984). ( 1 984). A perceptual motor theory of of emotion. emotion. In K. R. R Scherer & P. Ekman (Eds.), Approaches emotion (pp. Scherer Approaches to emotion (pp. 271-292). 27 1 -292). Hillsdalr Hillsdale, NJ: Erlbaum. Leventhal, relationship of of emotion to Leventhal, H., H, & Scherer, K. R. R (1987). ( 1 987). The relationship cognition: A functional functional approach to a semantic controversy. controversy. Cognition Cognition J, 3-28. and Emotion, 1, and Logan, G. D. (1988). of automatization. automatization. ( 1 988). Toward an instance theory of Psychological Review, 95, 492-527. R G., & Levy, P. E. ((1994). 1 994). Moving from cognition to Lord, R. to action: A Applied Psychology: An International International control theory perspective. perspective. Applied 335-367. ReView, 43, 335-367. Review, J. L., Bonanno, G. A, A., Davis, P., & et r al. ((1994). 1994). Lorig, T. S., Singer, 1. affr162 Repressor personality styles and EEG patterns associated with affective suppression. Imagination, Cognition and memory and thought suppression. Personality, 114, 4, 203-2 1 0. 203-210. A. SS.. R., Tctlock, P. E. ((1989). Manstead, A R, & Tetlock, 1 989). Cognitive appraisals and evidence. Cognition and Emotion, 3, 225emotional experience: Further evidence. 240. Matthews, G., & Amelang, M. ((1993). 1 993). Extraversion, arousal theory and performance: A study of of individual differences in the EEG. Personality and Individual Diff erences, 14, 347-363. Differences, G.A., & Harley, T. A A. (1993). Effects of of extraversion r selfMatthews, G.A, and self report arousal on semantic priming: A connectionist approach. Journal of of Personality and Social Psychology, 65, 6.5, 735-756. R., & Lees, 1. J. L. ((1990). r Matthews, G., Davies, D. R, 1 990). Arousal, extraversion, f and individual differences in resource availability. Journal oof Personality and Social Psychology, 59, .59, 150-168.
C.M. van Reekum and K.R. K.R Scherer CM
295
R.,, Sato, K, K., & Tucker, 1. J. ((1992). Mauro, R 1 992). The role of appraisal in human of f Personality and Social emotions: A crosscultural study. Journal o Psychology, 62, 3301-317. 0 1 -3 1 7. McCrae, R.-R. ((1993). 1 993). Openness to experience as a basic dimension of Cognitmn and Personality, 113, 39-55.. 3, 39-55 personality. Imagination, Cognition K. M. ((1987). of McGraw, K 1 987). Guilt following transgression: An attribution of of Personality and Social Psychology, responsibility approach. Journal of 247-256.. 53, 247-256 Messick, S. ((1994). 1994). The matter of style: Manifestations of personality in PsyChologist, 29, 112121cognition, learning, and teaching. Educational Psychologist, 1136. 36. R. B. ((1993). Murphy, D., & Zajonc, R 1 993). Affect, cognition, and awareness: Affective priming with optimal and suboptimal stimulus exposures. of Personality and and Social Psychology, 64, 723-739. Journal of Murray, N., Sujan, H., Hirt, E. R., of R, & Sujan, M. ((1990). 1 990). The influence of of mood on categorization: A cognitive flexibility interpretation. Journal of Personality and and Social Psychology, 59, 4411-425. 1 1 -425 . Neuberg, SS.. L., & Newson, Newson, J. Neuberg, 1. T. ((1993). 1 993). Personal need for structure: of f Individual differences in the desire for simpler structure. Journal o 3- 1 3 1 . Personality and Social Psychology, 65, 1I 113-131. P.. M M.. ((1990). of affective information. Niedenthal, P 1 990). Implicit perception of of Experimental Social Psychology, 26, 505-527. Journal of heart's's eye: eye: Emotional Niedenthal, P. M., & Kitayama, S. (1994). ( 1 994). The heart influences in perception perception and and attention. San Diego, CA: Academic Press. Oatley, K., of K, & Johnson-Laird, P. P . N. (1987). ( 1 987). Towards a cognitive theory of and Emotion, 1, 29-50. emotions. Cogniaon Cognition and Ohman, A. A (1986). ( 1986). Face the beast and fear the face: Animal and social fears as prototypes for evolutionary evolutionary analyses of of emotion. Psychophysiology, 23, 123-145. 1 23-145. of emotions. In V. Ohman, A. A (1988). ( 1 988). Preattentive processes in the generation of Hamilton, G. H. Bower, & N. N . H. Frijda Frijda (Eds.), Cognitive perspectives perspectives on emotion emotion and and motivation motivation (pp. 127-144). 1 27-144). Dordrecht: Kluwer Academic Publishers. Publishers. Ohman, A., A, & Soares, J. 1. J. 1. F. (1994). ( 1 994). "Unconscious "Unconscious anxiety": Phobic responses to to masked masked stimuli. stimuli. Journal Journal of of Abnormal Abnormal Psychology, Psychology, 103, 1 03, 231-240. 23 1 -240. Ohman, A., A, Dimberg, U., & Esteves, F. (1989). ( 1 989). Preattentive activation of of aversive emotions. emotions. In L. G. Nilsson & T. Archer (Eds.), Aversion,
296
Chapter 6
avoidance, avoidance, and anxiety: anxiety: Perspectives Perspectives on aversively motivated behavior (pp. 1169-193). 69-193). Hillsdale, NJ: Erlbaum. rkinson, B. ((1997). 1 997). Untangling the appraisal-emotion Parkinson, appraisal-emotion connection. connection. Pa Personality and Social Psychology ReView, Review, 11,, 62-79. constructs of of Parkinson, B., & Lea, M. F. ((1991). 1 991). Investigating personal constructs emotion. British fPsychology, 2, 73-86. Briash Journal oof Psychology, 882, Parkinson, B., & Manstead, 1992). Appraisal as a cause of Manstead, A. S. R R. ((1992). fPersonality and Social Psychology, 49. Review oof Psychology, 13, 122-1 122-149. emotion. Review 1 993). Making sense of Parkinson, Parkinson, B., & Manstead, Manstead, A. SS.. R R. ((1993). of emotions emotions in stories and social life. Cognition and Emotion, Emotion, 7, 295-323. life. Cognition Peterson, C., & Seligman, 1984). Causal explanations as a risk factor Seligman, M. E. ((1984). depression: Theory and evidence. evidence. Psychological for depression: Psychological ReView, Review, 91, 347374. Piaget, 1. 1 937/1955). The child child's's construction construction of of reality. reality. London: London: J. ((1937/1955). Routledge and Kegan Paul. Press, A. N., Crockett, W. H., & Delia, O. ((1975). 1 975). Effects of cognitive Delia, 1. J. G. complexity and of of perceiver's set upon the organization organization of impressions. Journal o fPersonality and Social Psychology, of Psychology, 32, 865-872. Pruzinsky, T., & Borkovec, T.-D. ((1990). 1 990). Cognitive and personality characteristics of worriers. Behaviour Research Research and Therapy, Therapy, 228, characteristics 8, 5075512. 12. R.,, & Spielhofer, C. (1994). dimensions of of Reisenzein, R ( 1994). Subjectively salient dimensions Emotion, 18, 331-77. appraisal. Motivation and Emotion, 1-77. emotional appraisal. Hofmann,, T. ((1990). of dimensions dimensions of Reisenzein, R., & Hofinann 1 990). An investigation of cognitive appraisal in emotion using the repertory grid technique. technique. Motivation and Emotion, -26. Emotion, 14, 11-26. Reisenzein, R., & Hofinann, Hofmann, T. ((1993). emotions from Reisenzein, 1 993). Discriminating emotions information: Baseline data for structural appraisal-relevant situational information: Emotion, 7, 7, 27 271-293. models of cognitive appraisals. Cognition and Emotion, 1 -293. SelfRhodewalt, F., Morf, C., Hazlett, S., & Fairfield, M. ((1991). 199 1). Self handicapping: The role of discounting and augmentation in the preservation of self-esteem. self-esteem. Journal of of Personality and Social Psychology, 1 , 1122-131. 22-13 1 . Psychology, 661, J. ((1984). emotion: A structural Roseman, II.. 1. 1984). Cognitive determinants of emotion: Review oof theory. In P. Shaver (Ed.), Review fpersonality and social psychology (Vol. 5, pp. 111-36). 1-36). Beverly Hills, CA: Sage. Roseman, I. J. 1. ((1991). 199 1). Appraisal determinants of discrete emotions. emotions. Emotion, 5, 1161-200. Cognition and Emotion, 6 1 -200.
C.M. van van Reekum Reekum and and K.R. K.R. Scherer Scherer CM
297 297
Roseman, I.I. 1J.,. , Spindel, Spindel, M. M. SS.,., & & Jose, Jose, P. P. E. E. ((1990). Appraisal of of emotion emotion1 990). Appraisal Roseman, eliciting events: events: Testing Testing aa theory theory of of discrete discrete emotions. emotions. Journal Journal of of eliciting 899-915. Personality and Social Psychology, 59, 59, 899-9 1 5 . Personality and Russel, D D.,. , & & McAuley, McAuley, EE.. ((1986). Causal attributions, attributions, causal causal dimensions, dimensions, Russel, 1 986). Causal success and failure. Journal of of Personality and affective reactions to success 74- 1 1 85 . and Social Psychology, 50, 1I 1174-1185. L. ((1987). memory: History and current status. Journal Schacter, D. L. 1 987). Implicit memory:
of Experimental Psychology: Learning, Memory, and Cognition, 113, of 3, 5501-518. 0 1 -5 1 8 . for memory: The brain, brain, the mind, mind, and the L.. ((1996). Schacter, D. L 1 996). Searching f or memory: past. New York: Basic Books. F.,. , & Carver, Carver, C. C. SS.. ((1985) and health: health: Scheier, M. F 1 985) Optimism, coping, and Assessment and implications of generalized outcome expectancies. 4, 2219-247 Health Psychology, 4, 1 9-247 K. R. ((1982). Scherer, K 1 982). Emotion as a process: Function, origin, and Information, 21,555-570. regulation. Social Science Inf ormation, 21 , 555-570. K. R. ((1984a). multicomponent process: A model model and Scherer, K 1984a). Emotion as a multicomponent of personality and some cross-cultural data. In P. Shaver (Ed.), Review of (Vol. 5, pp. 37-63). Beverly Hills, CA: Sage. social psychology (Vol. Scherer, K K. R. ((1984b). emotion: A component Scherer, 1 984b). On the nature and function of emotion: K. R. Scherer & P. Ekman (Eds.), Approaches to process approach. In K 293-318). 1 8). Hillsdale, NJ: Erlbaum. emotion (pp. 293-3 Scherer, K. K R. (1986). ( 1 986). Vocal affect expression: A review and a model model for future research. Psychological Bulletin, 99, 143-165. 143- 1 65 . Scherer, K.R. of affective disorders. In 1.0. J.D. Maser KR. (1987). ( 1 987). Vocal assessment of (Ed.), Depression and and expressive behavior. (pp. 57-82). Hillsdale, NJ: Erlbaum. Scherer, K. Criteria for emotion-antecedent Scherer, K R. (1988). ( 1 988). Criteria emotion-antecedent appraisal: A review. In V. Hamilton, G. H. Bower, Bower, & N. H. Frijda (Eds.), (Eds. ), Cognitive perspectives on emotion and and motivation (pp. 89-126). 89- 1 26). Dordrecht: Nijhoff. Nijhoff. Scherer, Scherer, K. K R. (1992). ( 1 992) . What What does facial expression express? In K. K Strongman (Ed.), (Ed.), International International Review of of Studies on Emotion (Vol. (Vol. 2, pp. 139-165). 1 39- 1 65). Chichester: Chichester: Wiley. Scherer, Scherer, K. K R. (1993a). ( 1 993a). Neuroscience projections projections to to current debates in emotion Cognition and and Emotion, Emotion, 7, 1-41. 1 -4 1 . emotion psychology. psychology. Cognition Scherer, Scherer, K. K R. (1993b). ( 1 993b). Studying Studying the the emotion-antecedent emotion-antecedent appraisal appraisal process process: An expert expert system system approach. approach. Cognition Cognition and and Emotion, Emotion, 7, 325-355. 325-355.
298 298
Chapter 6
Schcrcr, Scherer, K. R. (1997). ( 1997). Profiles Profiles of of cmotion-antccexicnt emotion-antecedent appraisal: testing and Emotion, I 1, I , 1131 13predictions across cultures. cultures. Cognition and theoretical predictions 150. 1 50. Scherer, Scherer, K. R. (in press). Appraisal Appraisal theories. theories. In T. Dalgleish, Dalgleish, & M. Power (Eds.), Handbook of ofcogmtion cognition and emotion. emotion. Chichester: Wiley. (Eds.), Schneider, ( 1 985). Toward Toward aa model of of attention and of Schneider, W. (1985). and the the development of I . Posner & O. O . S. S . Matin Marin (Eds.), Attention automatic processes. In M. I. performance )21 XI (pp. 475-492). 475-492). Hillsdale, Hillsdale, NJ: Eflbaum. Erlbaum. and performance 1., & Streufert, Streufert, S. S . (1967). ( 1 967) . Human information information Schroder, H. M., Driver, M. M . J., Schroder, complex social processing: Individuals and groups functioning in complex New York: Holt, Rinehart and situations. Now and Winston ( 1994) . Relationships between need achievement, Schroth, M. M . L., & Lund, E. (1994). Sehroth, sensation seeking and cognitive performance. performance. Personality and Differences, 16, 1 6, 861-867. Individual Differences, Shiffrin, Controlled and automatic human Shiffrin, R. M., & Schneider, W. (1977). ( 1 977) . Controlled information processing: II. Perceptual learning, automatic attention and ReView, 84, 1127-190. 27-1 90. a general theory. Psychological Review, Smith, C. A. of appraisal and physiological response in A (1989). ( 1 989) . Dimensions of Psychology, 56, 339-353. 339-353. emotion. Journal of ofPersonality and Social Psychology, A, & Ellsworth, Ellsworth, P. C. (1985). ( 1 985). Patterns of Smith, C. A., of cognitive appraisal in ofPersonality and Social Psychology, Psychology, 48, 813-838. 8 1 3-838. emotion. Journal of Smith, C. A., & Ellsworth, of appraisal and emotion Ellsworth, P. C. (1987). ( 1987). Patterns of f Personality and Social related to taking an exam. Journal o of Psychology, Psychology, 52, 475-488. S.. ((1993). 1 993 ) . Appraisal components, core Smith, C. A., & Lazarus, R. S relational relational themes, and the emotions. Cognition Cognition and Emotion, 7, 7, 233269. Smith, C. A., Griner, L. A., A, Kirby, L. D., & Scott, H. S. ((1996). 1996). Toward a process model of of appraisal in emotion. emotion. In N. H. Frijda (Ed.), Proceedings o f the IXth. Con ference of or of Conference of the International Society f for Research on Emotions, '96. Storrs: ISRE Publications. Emotions, ISRE ISRE'96. Publications. 1., & Schutzelaars, A 1986/87). On 'cognitive' Smolenaars, Smole aars, A. J., A. 1. J. H. ((1986/87). semantics of emotion words: Solomon quasi ecologically tested. Journal oof fSemantics, 1. Semantics, 5, 207-23 207-231. Solomon, R. C. ((1976). 1 976) . The passions: The myth f human myth and nature nature oof emotion. Garden City, NY: Doubleday. Sommers, S 1 98 1 ) . Emotionality reconsidered: The role of cognition in S.. ((1981). f Personality and Social emotional responsiveness. Journal o of 553-56 1 . Psychology, Psychology, 41, 41,553-561.
C.M. van van Reekum Reekum and and K.R. K.R. Scherer Scherer CM
299 299
Stelmaek, R. R. M., M., Houlihan, Houlihan, M., M., & & McGarry-Roberts, McGarry-Roberts, P. P. A. A. ((1993). Stelmack, 1 993). Journal of of Personality, reaction reaction time, time, and and event-related event-related potentials. potentials. Journal Personality,
Personality and Social Social Psychology, Psychology, 65, 65, 399-409. 399-409. Stenberg, Stenberg, G., G., Wendt, Wendt, P. E., E., & & Risberg, Risberg, 1.J. ((1993). Regional cerebral cerebral blood blood 1 993). Regional and Individual Diff Differences, erences, 115, 5, 547flow and extraversion. Personality and 554. Testing some attribution-emotion attribution-emotion Stipek, D., Weiner, B., & Li, K. ((1989). 1 989). Testing China. Journal oof relations in the People's Republic of China. f Personality and Psychology, 56, 1109-116. 09- 1 1 6. Social Psychology, of interference in in serial verbal reactions. Stroop, J. R. ((1935). 1 935). Studies of of Experimental Psychology, Psychology, 18, 643-662. Journal of Stuss, D. T., & Benson, D. F. ((1983). 1 983). Emotional concomitants of M.. Heilman & P. Satz (Eds.), Neuropsychology oof f psychosurgery. In K. M human emotion emotion (pp. 1111-140). New York: Guilford Press. 1 1 - 1 40). Teasdale, J. D. ((1988). 1 988). Cognitive vulnerability to persistent depression. Cognition and Emotion, 2, 247-274. Teasdale, J. D., & Barnard, Bamard, P. J. ((1993). Affect, cognition, and change: change: Re Re1 993). Aff ect, cognition, thought. Hove (UK): Erlbaum. modelling depressive thought. of emotion: A Tesser, A. ((1990). 1 990). Smith and Ellsworth's appraisal model of replication, extension, and test. Personality and Social Psychology Bulletin, 116, 6, 2210-223. 10-223. E. . ((1983). of thought. Journal oof Tetlock, P. E 1 983). Accountability and complexity of f Personality 45 74-83 Personality and Social Psychology, 45, Thompson, E. P., Chaiken, S., & Hazlewood, J. D. ((1993). 1 993). Need for cognition and and desire for control as moderators of of extrinsic reward effects: A person •x situation approach to of intrinsic to the study of motivation. Journal of ofPersonality and Social Psychology, 64, 987-999. 987-999. Tobacyk, c . , & Hourcade, J. (1991). ( 1 99 1 ). Self-monitoring Self-monitoring and Tobacyk, J. J., Driggers, E. C., psychological type: A social cognitive information processing model. Journal of ofPsychological Psychological Type, Type, 22, 33-38. 33-38. ( 1 995). Hemispheric specialization and Tous, J. M., Fuste, A., & Vidal, J. (1995). individual differences differences in cognitive processing. Personality and Individual 1 9, 463-470. 463-470. Individual Differences, Differences, 19, Tulving, Tulving, E. (1972). ( 1 972). Episodic Episodic and and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organisation of of memory. memory. New New York: York: Academic (Eds.), Organisation Donaldson Press. pp. 381-403. 38 1 -403 . Waller, Waller, J. J . E. E . (1994). ( 1 994). Dual-processing Dual-processing in value value self-confrontation: self-confrontation: The The role role of of need Social Issues, 50, 175-188. need for for cognition. cognition. Journal Journal of o/Social 1 75- 1 8 8 . ,
300 300
Chapter 66 Chapter
Weiner, B. B. (1982). ( 1982). The The emotional emotional consequences consequences of of causal causal attributions. attributions. In M. Clark & & S. S . T. T. Fiske (Eds.), (Eds.), Affect Affect and and rcognition (pp. 185-209). 1 85-209). S. Clark Hillsdale, NJ: NJ: Erlbaum. Erlbaum. Hillsdale, An attributional attributional theory theory of ofmotivation motivation and and emotion. New New Weiner, B. (1986). ( 1 986). An Weiner, Springer. York: Springer. Amirkhan, J., 1., Folkes, Folkes, V. S., & Verctte, Verette, J. A. A (1987). ( 1 987). An An Weiner, B., Amirkhan, of excuse excuse giving: giving: Studies of of aa naive theory of of attributional analysis of Journal of ofPersonality Personality and and Social Social Psychology, 52, 316-324. 3 1 6-324. emotion. Journal An Weiner, B., Graham, S., & Chandler, Chandler, C. (1982). ( 1 982). Pity, anger, and guilt: An attributional analysis. Personality and and Social Psychology Bulletin, 8, 226-232. 226-232. Weiner, B., Russel, D., & Lcrman, Russel, D., Lerman, D. (1979). ( 1 979). The cognition-emotion process Journal of of Personality and and Social achievement-related contexts. contexts. Journal in achievcment-rdated 3 7, 1211-1220. 1 2 1 1-1 220. Psychology, 37, A, & Matthews, G. (1994). ( 1 994). Attention and and emotion: A clinical Wells, A., (UK): Erlbaum. perspective. Hove (UK): 1. M. G., Mathews, A., A, & M MacLeod, ( 1996). The emotional Williams, J. a c L ~ , C. (1996). Stroop task and psychopathology. 20, 3-24. psychopathology. Psychological Bulletin, 1120, Woodfield, E., Jones, G. V., & Martin, M. ((1995). 1 995). Selective processing of stress-related words in colour naming. Personality and Individual Difforences, 9, 385-387. Differences, 119, 385-387. Zajonc, R. B. ((1980). 1 980). Feeling and thinking: Preferences need no inferences. American Psychologist, 2, 115 5 1l- 176. of affect. In K. R. Scherer & P. Ekman Zajonc, R. B. ((1984). 1984). On the primacy of (pp. 259-270). Hillsdale, NJ: Erlbaum. (Eds.), Approaches to emotion (pp. B.. ((1993). Zucker, G. S., & Weiner, B 1 993). Conservatism and perceptions of poverty: An f Applied Social An attributional analysis. Journal oof Psychology, 23, 925-943 925-943.. Author Notes
The authors would would like like to thank Leonard Berkowitz, Bcrkowitz, Craig Craig Smith, Leslie The Kirby, and in particular particular Tom Johnstone, for their their extremely helpful comments and suggestions on earlier versions of this chapter. Preparation of this 1 1-37504.93) of the Swiss National manuscript was supported by a grant ((11-37504.93) Scientific Research Foundation to Klaus R. Scherer. Schcrcr.
Cognitive Science Perspectives Perspectives on Personality Personality and Emotion Emotion
- G. Matthews Matthews (Editor) (Editor)
-
9 1997 1997 Elsevier Science B.V. rights reserved. reserved. © Science B .V . All rights CHAPTER 7 CHAPTER
Modeling Individual Differences in Negative Information Processing Processing Biases
J Siegle and Rick E. Ingram Greg J Research on depression and research on personality have long been very different fields. Research on personality and individual differences has concentrated on figuring out why individuals respond differently to similar stimuli, process similar information differently, and act differently in similar social situations. It has often focused on understanding differences between individuals who are considered "normal". In contrast, research on depression has focused on commonalties among people who are considered depressed. Questions such as why depressed depressed people respond similarly to negative stimuli and why depressed people appear appear similarly biased in their processing of of negative information have pervaded this literature. The disciplines of of studying differences among otherwise similar people The and similarities between groups of of very different people have recently begun and to converge as theory begins to suggest that depressed people differ that mood may affect qualitatively on a number of of trait-like dimensions, and that the expression of of individual differences other than depression itself. For example, Klein, Wonderlich, and Shea ((1992) 1 992) suggest that enduring personality characteristics such as dependency and perfectionism appear appear to be associated with qualitative differences in the expression of of depression. and Miranda ((1992) Similarly, researchers such as Persons and 1 992) have found that biased memory for negative things, which was once thought to be trait-like, appear to vary with affective state. These paths have begun begun to suggest that appear of personality, and that personality is depression can affect the assessment of intimately associated with the onset, maintenance, and recovery from depression. An area in which which this synthesis has been extremely productive regards the mediating role of of personality variables upon biased information processing in dysphoric individuals. Some dysphoric individuals appear to systematically attend to negative over positive or neutral information of (MacLeod & Mathews, 11991), 99 1 ), remember disproportionate amounts of of negative information (Blaney, 11986), 986), and engage in excessive processing of negative information or "rumination". Such information processing biases
302
Chapter 7
linked by numerous researchers to a vulnerability for depression have been linked (e.g., Ingrain, Ingram, 11984). 984). Other dysphoric and even clinically depressed individuals do not appear to show these biases (e.g., Siegle, Siegle, Ingram, & Matt, 11996a). 996a). To the extent that personality and individual difference variables may mediate these biases, it will be important to understand understand their role in depression. Such variables may be particularly valuable in predicting who will and will not be vulnerable vulnerable to depression after a period of dysphoria. The goal of this chapter is to illustrate illustrate some of the ways in which personality factors might help to explain variation in information processing in dysphoric and depressed individuals. Towards this end, end, we will introduce a number of of proposed roles for personality and individual individual differences in explaining vulnerability to depression. We then describe a framework for evaluating these roles using computational neural networks. To illustrate the use of of this framework, a computational neural network described by Siegle, Siegle, Ingram and Matt ((1995; 1 995; 11996b) 996b) will be employed to simulate the roles of of three personality variables in mediating observed information processing biases on two simulated information processing tasks. Each simulated personality variable will will be related to the goal of integrating integrating depression and personality research through the following steps. First, ways a personality variable may be simulated by an analogous aspect of of the network model model will be explained. explained. The effect of of the parameter on the network's performance on simulated information processing tasks will be assessed before and after atter the network has undergone a simulated of prolonged dysphoria. Ways in which the parameter affects the analog of will be generalized generalizeA to ways in which the network's simulated behavior will analogous personality characteristic might affect performance. Evidence will be provided for this type of a relationship in dysphoric and nondysphoric will provide a few integrative people. After discussing each variable, we will conclusions regarding the applicability of this type of framework for of personality in vulnerability to depression. investigating the role of
Personality Research and Vulnerability to Depression: A History Personality individual difference variables appear to influence influence how people Many individual process information, including what they attend to, what they remember, and life-experiences. Two such variables are cognitive how they interpret their life-experiences. structures and coping processes. A third individual-difference variable which may bear distinct correlates on the onset, maintenance, and recovery from
G.J G.J. Siegle and R.E. Ingram lngram
303
depression depression is intellect. intellect. While the latter variable has not received a great deal of of attention from depression depression theorists, the former variables have. A common theme underlying several contemporary models of of depression that individuals who experience experience this disorder, or who are at at risk for its is that experience, are characterized characterized by cognitive structures structures that that affect information processing in a dysfunctional manner. For example, models by Beck ((1967, 1 967, processing 1984), and Teasdale ((1988), 1 988), although differing on some 11976), 976), Ingram ((1984), dimensions, converge on the notion that that the onset and maintenance maintenance of of depression depression is influenced, influenced, to a significant degree by dysfunctional information that is determined by negative or depressive cognitive structures structures processing that or schemas. A wealth of empirical data wealth of data has, in fact, shown that that depressed individuals process process information in a variety of of domains domains in a negative and self selfdefeating manner manner (e.g., Lester & Schaeffier, Schaeffler, 1993; Schill & Kramer, 1991). From a functional standpoint, such schema driven information processing is thought to lead to life experiences being processed processed in a fashion which which induces selfand maintains depression. Specifically, positive and potentially self enhancing experiences are are processed poorly or not not at at all while neutral experiences experiences take on a negative tone, and genuinely negative life experiences catastrophic. The negative cognitive triad proposed several are viewed as catastrophic. decades ago by Beck ((1967) 1 967) still adequately adequately sums up the interpretation interpretation of of life experiences by depression-prone individuals; one's self, future, and and world world are of viewed in quite negative terms. It is hard to overestimate the implications of way of of structuring structuring life experiences for understanding understanding depression. this way of attention Although cognitive structures have received a great deal of from depression depression theorists, other cognitive processes processes are also considered from important. Of particular relevance relevance for this chapter, chapter, is coping. The literature literature Of particular on coping is extremely large and has been reviewed in depth elsewhere (Zeidner & Endler, 11996). particular style of of 996). Instead, our focus is on a particular cognitive coping that several several depression researchers have have suggested may have cognitive important implications implications for either the alleviation of important of a depressive state or for its that one of of the ways in which the prolongation. Ingram ((1984) 1 984) suggested that activity of dysfunctional cognitive structures could be either exacerbated exacerbated or of dysfunctional diminished was through active efforts at cognitive coping. Ingram proposed that once activated, negative cognitive structures structures are consistently reactivated that by recycling energy through a network of of other interconnected cognitive structures. structures. Phenomenologically, this process may be experienced as a the individual. Cognitive efforts and and behaviors intended to to ruminative state by the interrupt may be considered to constitute constitute coping. Alternately, interrupt this activity may interrupt this cognitive recycling process individuals who make little effort to interrupt
304
Chapter Chapter 7
are are apt apt to to experience experience longer, longer, more more severe severe periods periods of of depression. depression. Coping Coping is is thus jor determinate thus seen seen in in this this model model as as aa ma major determinate of of several several important important
parameters of the the depressive depressive state. state. parameters of More 1 993) have More recently, recently, Nolen-Hoeksema, Nolen-Hoeksema, Morrow, Morrow, and and Fredrickson Fredrickson ((1993) have expanded the cognitive coping coping to give it expanded the notion notion of of cognitive to give it aa central central role role in in what what she she terms the "response "response styles styles theory" theory" of of depression. depression. In In brief brief the the authors authors propose propose terms the that erence variable that an an important important individual individual diff difference variable in in depression depression is is the the extent extent to to which people people are are inclined inclined to to ruminate ruminate on on the the symptoms symptoms of of depression. depression. Those Those which who who ruminate ruminate longer longer and and deeper deeper are are proposed proposed to experience experience more more protracted protracted and while those those who less will and severe severe periods periods of of depression depression while who ruminate ruminate less will have have correspondingly shorter and less intense correspondingly shorter and less intense depressive depressive experiences. experiences. Other Other models models of depression depression have have given given rumination rumination aa similarly similarly central central role role (Pyszczynski (Pyszczynski & of Greenberg, 987). While ashion than Greenberg, 11987). While we we would would view view rumination rumination in in aa broader broader ffashion than 1 993; e.g., rumination on do do Nolen-Hoeksema Nolen-Hoekserna et et al. al. ((1993; e.g., rumination on aa variety variety of of depression depression related related phenomena phenomena in in addition addition to to symptoms symptoms may may be be important), important), we we believe believe that the the central central construct construct of of rumination rumination is is an an extremely extremely important important variable variable in in that the the maintenance, maintenance, if if not not the the onset, onset, of of depression. depression. Simulating Simulating Aspects of of Depression and Personality on a Computer
The research above has has provided provided many many useful useful hypotheses hypotheses The research described described above about about potential potential
relationships depression and relationships between between vulnerability vulnerability to to depression and personality. Most of of these these hypotheses hypotheses refl reflect causal models models (e.g., (e.g., rumination rumination personality. Most ect causal prolongs depression). A common common criticism criticism of of clinical clinical causal models is is that that prolongs depression). causal models they often fundamentally fundamentally unsatisfying unsatisfying to to people people wishing wishing to to understand understand they are are often how aspects aspects of of aa model model interact. interact. For For example, example, diagrams diagrams composed composed of of circles circles how and arrows arrows are are valuable valuable tools tools ffor helping readers readers understand understand an an author's author's and or helping conception of of qualitative qualitative relationships relationships between between measured measured or or hypothetical hypothetical conception constructs. However, However, such such diagrams diagrams do do not not convey convey the the precise precise nature nature of of constructs. mediational interactions interactions between between causal causal agents. agents. Moreover, Moreover, the the entirety entirety of of the the mediational processes which which are are assumed interact, and and potentially potentially even even ffeed back to to re reprocesses assumed to to interact, eed back influence each other other are are rarely rarely directly directly observed observed in in the the generation generation of of causal causal infl uence each explanations; rather rather specific specific steps steps in in the the causal causal path path are are observed observed and and large large explanations; causal diagrams are are constructed constructed from from these these paths. paths. Thus, the technologies technologies causal diagrams Thus, the used to used to specify specify and and evaluate evaluate theoretical theoretical causal causal models models have have rarely rarely caught caught up up with our our complex complex intuitions intuitions of of causal models themselves. themselves. A model model with causal models constructed within within aa framework framework which direct inspection inspection of of the the constructed which allows allows direct processes involved involved in in causal mediation of of clinical clinical phenomena phenomena would would be be much much processes causal mediation more desirable. desirable. Computational Computational models models provide provide this this technology, technology, allowing allowing their their more
and R.E. Ingram Ingrain G.J. Siegle and
305
feeAback between modeled constructs and to creators to interactively observe feedback experiment with how changes in m these constructs might change observed of variables, behaviors. They also allow users to investigate how the addition of which were not empirically measured, might influence the observed behaviors. of computer simulations to understand behavior has thus gained The use of a growing f ollowing in recent years f or addressing causal mediation processes following for manner. in this more satisfying and rigorous mann er. Computational modeling has many advantages ffor psychologists interested in understanding understaadmg causal or mediation processes in abnormal behavior. Some include its ability to aid in of theory to a rigorously specified, the translation of specified, empirically testable causal model, to aid in understanding behaviors, to generate new hypotheses about of novel clinical interventions, abnormal processes, to promote the creation of 996). and to integrate various "granularities" of of research (Siegle, 11996). of model which has sparked a great deal of of interest One particular type of among the computational clinical community is called a neural network ormation throughout a connected information model. Neural network models represent inf of units units.. "Activation" travels between units representing a flow of of network of inf ormation. Units in a connectionist information. conneetionist network have thus been likened to information biological neurons which receive inf ormation on dendrites and send out an aggregate of of that information over axons axons.. Neural network models have become popular for many reasons including their predictive power (Sarle, 11994), 994), biological congruity (Cohen & Servan-Schreiber, 11992), 992), ability to data (Cohen & Servan-Schreiber, Servan-Sehreiber, 11992), 992), and the natural way in handle noisy data information which they can be used to model inf ormation processing tasks (McClelland, Rurnelhart & Hinton, 11986). 986). Additionally, some neural networks can be used Rumelhart to model phenomena which are difficult to represent using more traditional symbolic modeling techniques, such as behaviors for which explicit governing 990). Moreover, Tryon ((1993) 1 993) has rules are not known (Hecht-Neilson, (Heeht-Neilson, 11990). shown that that neural network models help to provide resolutions for traditional mind/body problem and the nature vs. schisms in psychology such as the mindlbody nurture debate.
Why use computer simulations to investigate dysphoria? While computational models, and specifically neural networks, are interesting vehicles for psychologists in general, they are particularly useful
for of cognition in depression f or researchers interested in understanding the role of because neural networks are natural biologically congruent extensions of
Chapter 77
306
of depressive inf information common models of ormation processing. Since the "cognitive revolution" of of the late seventies and early eighties, semantic network models have been proposed to explain inf ormation processing biases in depression. information depression. Adopting Collins Collins and Loftus's Loflus's
((1975) 1 975) notion of of mental events as propositions
spreading between loci representing bits of of knowledge in a "semantic network", network", Bower
(1981) modeled the normal experience of of emotion by
suggesting that emotions emotions might also be nodes in such a semantic semantic network. network. Ingram
that when such emotion nodes become strongly ((1984) 1984) suggested that
of depression might occur. This approach activated, cognitive aspects of for of biased inf information appears to account f or many types of ormation processing observed in depressed people as well as the onset and maintenance of of depression 11984) 984) and 986; Morrow and its eventual treatment (Ingram & Hollon, 11986; & Nolen-Hoeksema, 990) and potential recurrence (Teasdale, 988). Nolen-Hockscma, 11990) (Tcasdale, 11988). The distinction between ssemantic emantic networks composed of ormation of inf information processing nodes, each representing beliefs or logical propositions, and distributed conncctionist connectionist networks, in which each node is individually meaningless, is important. Yet, work relating the two representations has served to provide important bridges between subdisciplines subdisciplines of of experimental 990; Blank et 992; Yates 993). For psychology (Anderson, (Anderson, 11990; ct aI. al.,, 11992; Yatcs & Nasby 11993). Yates and Nasby ((1993) example, Yates 1 993) identify six fundamental assumptions of associative common to semantic network and neural network models of memory including: 1) Both types of l) of models models use nodes representing aspects of of propositional knowledge; are associated at at the time of of encoding; 2) Propositions are 3) Inf Information ormation in the network becomes conscious when it is activated above some threshold; 4) Environmental stimuli activate some nodes in the network; can spread in varying degrees; 5) Activation can 6) Consciousness may may be understood as involving the activation of of nodes. The semantic network approach approach thus provides an important historical The for development of of neural network context f or understanding the subsequent development of depression. depression. The cognitive orientation, and the ffocus of models of ocus on the roles of memory and attention in depression used throughout semantic network models may provide a basis for their use in corresponding neural network reformulations of semantic network representations models. Neural network ref ormulations of of depression preserve the benefits of of semantic network models while of (Ingram,
G.J Siegle G.J. Siegle and R.E. Ingram
307
providing a biologically congruous approach to examining mechanisms responsible for the onset and maintenance of depression. More practically, practically, many types of experiments can be performed on a More computational computational neural network model which it would be unethical or difficult to perform on a depressed human. For example, it would often be desirable to test theories theories of depression by progressively depressing a person, bit by bit, until he or she were on the brink of complete despair. At each step of the induction, the the person could be assessed on a number of variables to ascertain whether the person's person's behavior follows that predicted by a theory. Such an experiment would, of course, be unethical. As As a substitute, if a computational computational model of analogs of processes processes thought to operate in depression could be created, created, the computer's computer' s behavior on some task could be analyzed at various levels of operation of these processes processes with relatively few ethical complications.
Why simulate personality f actors and individual diff erences? factors differences? Our goal in simulating aspects Our aspects of personality within a computational model is to generate a number of possibilities for the role of personality personality in mediating depressive information processing biases. It is our hope that simulations can can help to show how many avenues for differential information simulations processing might exist based on individual differences in just just a few constructs. R It is not the attempt of this paper to provide any concrete demonstrations of the factors may play in depression. empirical demonstrations the role these factors Rather, the following simulations serve as hopeful "might be's" and avenues for empirical confirmation. By simulating personality variables on a computer, theories about them can be rigorously specified. Hidden ambiguities may correspondingly become apparent, and thus thus theories theories of both normal and pathological behavior will stand stand a better chance of validation. Moreover, computational computational models which are biologically motivated often serve to suggest physiological correlates correlates of simulated phenomena, in this case, personality variables. Another reason to consider simulating personality factors on a computer involves the determination of vulnerability markers for depression. If some simulated personality factors are associated associated with greater expression of characteristics characteristics associated with depression after the model is subjected to a computational analog of a depression induction, we can examine the effects computational these factors before factors have on a model bef ore it assumes a state analogous analogous to
308 308
Chapter 7 Chapter
depression. By By examinm" examining the pre-pathologized pre-pathologized model, we may may determine depression. g the model, we markers for for people at at risk for for depression. depression. markers Two final final reasons reasons for for simulating simulating the the role role of of personality personality and and individual individual Two diff erence variables in depression will become more apparent later in the the difference variables in depression will become more apparent later in chapter. investigating the chapter. First, neural neural networks networks are are particularly particularly suited for for investigating role of parameters inherent to network of personality personality variables variables because because parameters simulations correspond correspond neatly neatly to to aspects of of personality. Without Without adding simulations variables variables to to an an existing simulation, simulation, aspects of of personality personality implicit in the models can can be be investigated. Second, by simulating characteristic characteristic individual differences variation in the differences in depressive depressive information information processing biases, biases, variation of aa computational computational model of of depression depression earl can be be attributed attributed to to factors behavior of other other than than random random noise. In this way, the specific specific variation variation in human human information information processing processing experiments experiments can can be approximated approximated more closely than than if if personality factors factors were not simulated. personality
A short introduction to neural short introduction neural networks To appreciate appreciate the types of of models which are following To are discussed discussed in the following sections, aa basic familiarity network models is familiarity with computational neural network assumexi. concepts assumed. The following section discusses some of of the basic concepts necessary for understanding understanding these models. A number of of excellent excellent review articles conncctionist articles are available which summarize summarize specific types of of connectionist networks (Rumelhart, Hinton, Hinton, & McClelland, Mr 1986), the mathematics and networks of computational simulations of of connectionist networks networks (e.g., mechanics of and the relative strengths strengths of of this type of of Arbib, 11987; 987; Hecht-Neilson, 1990), and traditional modeling technologies (Bamden, (Barnden, 1995), ffor modeling over traditional or the interested reader. reader. information Neural network models represent inf ormation throughout a connected network of of individually meaningless units nodes. Information is represented distributed fashion, as a function of of the simultaneous activation of of in a distributed multiple nodes. Each Each unit receives "activation" "activation" from other nodes to which it of these nodes nodes.. The unit then sends is connected in response to the activation of transfer function of of the activations coming into it to other nodes to which it a transfer neuron sends a function of of activations is connected, much as a biological neuron from its dendrites to other neurons through its axons. For example, a collection of of nodes, when activated together, may be said to represent a of these these nodes may spread to another concept such as "sun". Activation of of nodes which, together, represent the concept "moon". This collection of of a mental process might be interpreted as a computational analog of
G.J Siegle and R.E. Ingram G.J.
309
association between sun and and moon. Environmental or internal factors (e.g (e.g.,. , visual stimuli) which cause some nodes tto ob bee activated are arc determined dctcrmmexi and modeled explicitly by the neural network modeler. The The
pattern
of of
connections
between
nodes,
or
the
network' network'ss
"architecture" governs what types o ormation may enter the network, the off inf information ormation may be manner in which nodes may activate each other, and what inf information of the network. Thus, neural network's response to a said to flow out of stimulus generally involves successive activation of of a number of of nodes in the network. This pattern pattern of of activations can represent an association association in memory between two stimuli, a reaction to an external stimulus, or some other construct based on the network's architecture and the designer's conception of of the network. The processes operating within connectionist networks can be be assumed assumed to to correspond to neuronal, cognitive, or behavioral events based based on the intuitions of of the network's designer. By strategically modifying the strengths of of connections between nodes, the network can can be made to produce a a specific set of of activations in response the of activations. This process has been likened to making the to another set of network "learn" an association of of a a stimulus with a response. Numerous procedures for allowing a a network network to learn associations in this fashion have been proposed (e.g. (e.g.,, Fallman & Lebiere, 11991; 99 1 ; Rumelhart, Hinton & Williams, 1986). 1 986). Once a a connectionist network model has been created, its behavior can be evaluated on a a number of of dimensions. The choice of of what dimension is to be evaluated generally reflects the processes which the network is designed to simulate. For example, if performance rmance on if the network is designed to simulate perfo some information processing task, associations made by the network could be compared to associations made made by humans to a stimulus; the frequency of of the network's erroneous associations could be measured as an an analog of of human error rates. Similarly, a network may take a a number of of associative steps to settle on a a learned association. The number of of associative processing cycles or "epochs" the network network needs to to associate a a stimulus with a particular particular response can be examined examined as an an analog analog of of reaction reaction time. may be inf inferred from this this introduction, connectionist models are As may erred from generally associated with a a number of of parameters which can be varied such as the relative strengths of of connections within within the the network, the rate at at which the network network can "learn", and and the number number of of nodes devoted to representing the of inf ormation within the types of information the network. Systematic Systematic manipulations of of these these parameters will be termed "perturbations" throughout this review. Such perturbations can can be used to make make the network respond in different ways to perturbations different ways
3 310 10
Chapter Chapter 7
of network parameters may be seen as similar stimuli. Thus, perturbations of one way of of approximating individual differences. Often, perturbations of of er network parameters are used to make a network function in a mann manner associated with psychopathology. Individual differences will be simulated using three types of of "perturbations" to to a network model, representing three factors which are believed to induce individual differences in humans. As humans are expected to differ in coping styles, or the techniques they use to process negative stimuli, network parameters corresponding to the ways in which information "response to a stimulus will be varied. By feeds through the network in 'response examining the network's perfo rmance on simulated analogs of information performance of some method for processing negative processing tasks given more or less of information, claims may be made about how different humans might process similar negative information. Another way in which individuals are hypothesized hypothesize~ to differ reflects different learning styles. Some individuals learn slowly and thus may take longer to assimilate new information. To gauge the of such a learning style on information processing in depression, effects of parameters affecting the network's learning style can can be manipulated. Finally, uman associations are based on what they have learned, associations just as h human made by the network are based on the stimulus-response pairs on which it has been trained. Thus, to approximate characteristics of of different individuals, the stimulus-response pairs on which the network is trained will be modified. For of depression, more training will be given to example, to represent an analog of the network on stimuli which are deemed to be negative. Similarly, to of experiences which might help an individual recover represent the types of from depression, the network can be trained on stimuli deemed not to be negative, after having been originally overtrained on negative information. of the sequence of of training examples will be important in Manipulations of of sequences of of life events which may lead to information simulating a variety of processing biases. Because the number and types of of parameters which may conneetionist model are too numerous to illustrate in detail here, occur in a connectionist parameters which are manipulated for each simulation will be explained the parameters as they are referenced in the text.
investigating affective affective inf information A computational or investigating ormation computational framework f for processing of personality variables on affective affcetive Before simulating the effects of of information processing, it will be useful to have a computational model of
G.J. G.J Siegle and and R.E. Ingrain Ingram
311 311
affective information information processing processing which which captures basic basic notions about about how how affective most that end, most dysphoric dysphoric and nondysphoric nondysphoric individuals individuals behave. Towards that Siegle, Siegle, Ingram, Ingram, and Matt Matt (1995; ( 1 995; 1996b) 1 996b) have developed developed a computational computational neural network model of of affective affective information information processing processing which simulates the behavior of of dysphoric dysphoric and nondysphoric individuals on two two information information processing tasks. Their model is depicted m in Figure 1. 1. processing
9
.-'.~. ' ~ . - ~ i . , ~
g~
....... ,:~'.,~,'":,i~~"'~,-~:~ ..~.~:~@.~i~
~}~
'
. . . . . .
'~:~.~r
~,~,~, .............
J ..ll,,~t...,,t~,.
! ...................
I.[.
I.
.1
.
.
.
-'..~..,,..-.~t ~?~;..~ia..,.~ii-..,~ ..... ~,,.~.-.:f~,~::.~.~,.,,.~}l
,
i i;. . . . ~iiiii~i~}~i i!i~ ....
~i
i iiiiii:iili!i!i!iii
~,~i~".~!.".'-"~:i::~.'-i~:~,~i~:~.".-'-~:~ i:~','~" .......... !!!,i,,i~i!~iiiiii!iiii!i!iill
....
.....~.',% ~
~'. ~
~.~.
~."
~:.
r~NN
~ ~ ~
~.~:.~ :~.i~.,:."~i
~
" ,
~ i l i
i!!i!iiiiiiiiiii!ii!i! iiiili?i/iiiiii'iii;iiiiiii/!ii~i
~N~I.. ~ii}i!i;i;i!iiiiiii . . . . ~ili! iiiii!ii!iiiiiligi'>'~!i~ii:
/ The affectivesemantic loop
.,iiiii!ii!iiiii!, ii!i!ii!!!!i!!ii!ili!li!iii!ili!i!!ii!i! !iiii!!iiiiiiiiii!iiiii!ii!i!ii!iiiiiill
Orthographic Features
information ormation Figure 11.. A neural network platform for investigating affective inf processmg. processing.
The model model is an attempt to formalize Bower's ((1981) of feedback The 1 98 1 ) notion of between a person's information. representing
internal representations of of aff affective internal ective and semantic
al.'s computational model, 110 In Siegle et al. ' s ((1996b) 1 996b) computational 0 nodes the
most
basic
perceptual
characteristics
of of
stimuli
activate 110 other nodes nodes which represent represent ("orthographic" nodes) are made to activate 0 other
3312 12
Chapter Chapter 7
the ssemantic of the stimuli stimulil1 .. The T h e orthographic aspects of of a stimulus the emantic content of
may be thought of of as the patterns of of light and dark dark which make up an emantic content represents the mental observed visual image whereas its ssemantic
representation of of a learned concept with which the visual representation is
associated. The semantic nodes activate two more nodes which represent the
affective of stimuli. One valence node represents the positivity of of a aff ective valence of stimulus and the other valence node represents its negativity. A neutral represented by neither of valence is representeA of the valence nodes being active. To 1984) notion of eedback between the aff ective and capture Ingram's ((1984) of ffeedback affective semantic aspects of of a stimulus, the nodes representing the affective valence of of the stimulus can reactivate the nodes representing the stimulus's semantic can content a variable number of eeds back to the of times. When activation ffeeds semantic or valence nodes the activation is computed as the sum of of the previous activation plus one half of of the activation from the incoming connections. In this way, the f eedback from other nodes does not immediately feedback exceed the original activation from the stimulus. The f eedback between the feedback affective and semantic representations of of a stimulus is congruous with Tucker and Derryberry's ((1992) 1 992) notion of eedback between the parts of of ffeedback of the brain of a stimulus, responsible for the affective and semantic representation of the hippocampus and and the the amygdala, though we have not tried to namely the aspeOs of of either the hipp hippocampus ocampus or the amygdala in this model other aspects framework. To simulate the knowledge base that a random individual might have, the network was trained on three positive, three negative and three neutral the stimuli. Stimuli were represented in a localist fashion in which only one orthographic, one semantic and one valence node was expected to be active f or a given stimulus. While the restriction to a localist representation is not for essential for the f ollowing simulations, it is useful f or illustrating how following for of personality were network connections changed when various aspects of simulated. Training involved presenting a simulated orthographic of a a stimulus to the network, observing the network's semantic representation of and valence nodes, and adjusting the weights within the network until the ed desired semantic and valence representations were achieved using a modifi modified back-propagation learning algorithm (Rumelhart, Hinton, & Williams, 11986) 986) feedback in which weights were adjusted after f eedback occurred 110 0 times between the 11 For most simulations, simulations, only the first nine nine nodes were used. The tenth node was reserved for simulations simulations involving "novel" stimuli to which the network network was was not exposed during its initial training period.
G.J R.E. Ingram G.J. Siegle and R.E. lngram
313 3 13
uman learning aff ective and affective and semantic nodes. No No claim claim is made here here that that h human
actually takes place via a back-propagation algorithm. Rather, it is assumed umans change their associations that that as in the the back-propagation back-propagation algorithm, h humans
with stimuli based on their experiences, and more association to some stimulus means that that it is learned better. Training Training continued until the sum of of
the the network's mean squared squared error error in in the the semantic semantic and and valence valence nodes nodes was was or a below 0.004 ffor a block of of all nine stimuli.
Simulating Simulating the induction inductmn of of depression in the network In order ormation order to use of of the neural neural network model to represent inf information processing biases in depression, it is necessary to discuss how the network is to to assume a simulated analog of of depression. A popular theory suggests that the induction of of depression involves a single, pervasive negative life event or loss experience (e.g., Beck 11974, 974, Brewin, Andrews, & Gotlib, 11993, 993, Paykel thought about. This process can be 11979) 979) which is continuously thought operationalized by allowing the network to be exposed to one negative of stimulus for a prolonged period after it has been trained on equal numbers of positive, negative negative and neutral stimuli. Thus, to represent the induction of of network was trained trained on a single negative stimulus ffor depression, the network or 70 epochs after at~er the network' network'ss initial training was complete. The network described above was was implemented in the PLANET PLANET neural The SPARC network simulation environment (Miyata, 11991) 99 1 ) on a Sun SP ARC 11 computer. PLANET PLANET is an environment in which neural network simulations constructed using a language language developed specifically for that purpose. may be constructed Users can can interactively examine activations within the the network and its Users "train" associate inputs with outputs, outputs, train" the network to associate contents, interactively " the resulting error error rates rates.. The The PlaNet PlaNet code representing the and observe the author upon network and for presenting stimuli are available from the first author request. To assess the the validity of of this this framework ffor To or representing information processing biases in dysphoric individuals it is useful to examine examine how the performs computational analogs of of inf information network perf orms on computational ormation processing tasks humans are known to be biased. Specifically, it is useful on which dysphoric humans the network exhibits similar biases biases to humans on tasks to examine whether the of the semantic and aff affective content of of stimuli. involving the recognition of ective content
3 14 314
Chapter 7
of information Simulating the recognition o fsemantic inf ormation of the recognition of of the semantic content of of stimuli is Assessment of traditionally done by presenting an individual with a stimulus, and measuring to recognize it. To measure ssemantic the time it takes the individual to emantic of the time it takes an individual to recognize other recognition independent of of the stimulus (e.g. (e.g.,, its size), pronounce the stimulus, move mouth aspects of of a parts, etc., ssemantic emantic recognition tasks are often conducted in the fform orm of two alternative fforc~ task in which an individual is shown a stimulus, two orced choice task a button corresponding to whether the the stimulus does, or and is asked to push a task in which an individual determines the lexicality does not, spell a word. A task of a stimulus is thus termed aa "lexical decision task". To gauge an of of one aff affective individual's recognition of of affective information of ective valence or of diff different another, stimuli of erent affective valences may be included in a lexical decision task. that individuals who have Bower's ((1981) 1 9 8 1 ) network theory suggests that of sadness and some strong connections between their internal representations of personally meaningful negative event should be facilitated at identifying the semantic content relevant to that that event. This is because information relevant to the event is expected to activate the mental representation of of sadness, which would, in tum, turn, more strongly activate the representation of of the event. Some authors use this argument to predict that depressed individuals should therefore affective theref ore be facilitated in responding to all negative stimuli on an aff ective 988; Macleod, Mathews, & lexical decision task (e.g., Challis & Krane, 11988; lexieal Mar Tata, 11986; 986; Matthews & Southall, 11991; 99 1 ; Ruiz Caballero & Bermudez confirmation. Moreno, 11992) 992) with little confi rmation. In contrast, Siegle, Ingram, and 1 996a) suggest that depressed and Matt ((1996a) individuals will only be facilitated in identifying the semantic content of of stimuli which which represent personally meaningful negative events, on the affective lexical decision task. They suggest that negative stimuli which are of negativity not personally relevant will activate the mental representation of which will, in turn, of personally relevant tum, activate the mental representation of connectext. They propose that the negative events to which it is strongly connected. of personally relevant negative events may interfere concurrent activation of of negative stimuli which are not personally relevant. with the identification of They therefore predict that that depressed individuals will, in general, be delayed They of most negative stimuli when the stimuli emantic content of in identifying the ssemantic presented ffor brief period. Siegle Siegle et al. ((1996a) of are presented or a brief 1 996a) demonstrate this type of interf erence meta-analytically ective interference meta-analytieally and empirically for results using an aff affective
G.J G.J. Siegle and R.E. R.E. Ingram
3315 15
lexical decision task. Specifically, Specifically, Siegle et al. ((1996a) show that a sample lexical 1996a) show of 30 dysphoric research participants were, were, in general, delayed in responding of to negative negative words with respect to positive or neutral words presented for a of 1150ms duration of 50ms on an affective lexical decision task. The same delays were not present for a sample of of 46 nondepressed individuals. Their results are illustrated in Figure 2. To test whether Bower's Bower's theory predicts that being overexposed to negativity leads to a general facilitation towards negative information or facilitation only for personally relevant negative information and delays for
9 NonDepressed-Lexical NonDepressed-Lexieal Decisim Decision ---+: Depressed-Lexieal Decision Decision Depressed-Lexical onDepressed-Vak:nce Identification - .... -~- .. - N NonDepressed-Valenee Idenfifr.ation - - Depressed-Vak:nce Depressed-Valence Identification ldentilieation 0.9 0.9
0.85 S"
9
r
9
0.8 S , . . J 'r
%%%~
r162 r "J'
"~ 9 % l ~
0.75 S ~ S
0.7
-%
SS
"N,
0.65 0.65
0.6 0.6
0.55 0.55
0.5 0.5
L-----
0.45 0 . 4 5 1 +--..... I ',---+---j
Positive Positive
Neutral NeWIll
Negative Negative
Word Word Valence Valence
Reaction times for lexical decision (LD) and valence identification Figure 2. Reaction (VID) tasks, 11550ms 0ms condition.
3 16 316
Chapter 7
other other negative information, Sicglc Siegle ct et al. (1996b) ( 1996b) created aa simulation of of the the affective lcxical lexical decision task. task. They They represent represent lcxical lexical decisions in aa manner manner affcctivc analogous to to that that used used by Cohen, Dunbar, Dunbar, and and McClclland McClelland (1990) ( 1 990) to to analogous represent word word and and color naming in aa rconnectionist model of of the the Stroop task. represent Ratcliff's (1978) ( 1978) notion that that semantic semantic identification is aa diffusion Following Ratr process, they they suggest suggest that that aa semantic identification occurs when when the of the mental representation of of aa stimulus reaches aa threshold. activation of They therefore define counters representing the the accumulated evidence for each item their their network might identify. The counters add each possible item add evidence of the the for aa given item proportional to the difference between the activation of ofthat of any other semantic representation representation of that item and the maximum activation activation of to gaussian noise. When any counter cxcce, exceeds semantic representation, subject to ds set to to 1), 1 ), the the network network is said to have made aa semantic aa threshold (arbitrarily sot of a stimulus for a limited duration is identification. The presentation of simulated by turning turning off input from nodes representing the orthographic off input of the brief period (90 epochs). In this way, the content of the stimulus stimulus after after a brief network is forced to identify the stimulus using only the residual activation semantic and affcctivc affective content of of the within the nodes representing the semantic Siegle ct et al. (1996b) ( 1 996b) show that when the network is trained for 70 stimulus. Sicglc epochs on aa negative stimulus after having received its original training, the network assumes relatively similar information processing biases to those observed by Sicglc Siegle ctal. et al. (1996a) ( 1996a) for dysphoric individuals on the affcctivc affective lexical decision task. The results for this case are lcxical arc depicted dcpicte~ in Figure 33.. bc seen from the figure, the network which has been bccn overtrained ovcrtraincd As may be on a negative stimulus is delayed in responding to other negative words on the lcxical lexical decision task, but is facilitated in responding to the particular negative ovcrtraincd. As negative stimuli on most lexical lcxical word on which it was overtrained. arc probably not especially personally relevant for the decision tasks are individuals who are arc being lacing tested, most depressed individuals would thus be bc expected to be bc delayed in responding to negative words on the task, though facilitation is predicted for any words which are arc particularly meaningful for a given depressed person. While Siegle Sicglc et ct al. ((1996b) 1 996b) report results for groups of simulations using of the figures the same parameter configurations, Figure 3, and the rest of depicting simulations in this paper show the simulated performance of a of the single run of the network. A single simulation is a valid indicator of nctwork's perfo performance network's rmance because the parameters governing noise in the arc not high enough to cause more than 11-2 network are -2 epochs of variability in overall simulated reaction times, and thus, there is no need nccd to average over
3317 17
G.J.. Siegle Siegle and and R.E. R.E. Ingram Ingram G.J. _ NonDepressed-LexK:al Dec6Dn
NonDepressed-Lexical Decision
Depmssed-Lexical Dec6Dn Decisbn - Depressed-LexK:al m -- NonDepressed-Valen;:e NonDepressed-Valeme Ideriificafun Idemifr~tion -- -.... 1136 36
-'~ - Depressed-Valeh:e Depressed-Valence Iderti6:afun Identification -
,A,
1131 31 S
126
epochs epochs
#
J
S
J d#
J
N\%\ \
J N
\\\
\
J
1121 21
\ \
1116 16
" -
-
.....
-
1111 11
1 ! I 1106 � "' +------�--r---� Positive
Neutral
Negative-Other Negative-Other
Negative NegativePersomlly Personally Relevant Relevant
Stimulus Valence Valenc.
Figure 3. 3 . Simulated lexical lexical decision and valence valence identification tasks.
multiple simulated subjects. of this paper paper is to show show subjects . As As the primary objective objective of the effects effects of of parameters governing governing individual differences, differences, effectively effectively simulating simulating one one subject subject at aa time sccmeA seemed appropriate. Additionally, Additionally, each each of of the rcportexi simulations was run multiple times to insure that results were not the reported simulations was run product of of a single anomalous anomalous trial.
Simulating Simulating the recognition recognition of ofaffective affective information information Assessment Assessment of of an individual's individual's recognition recognition of of the the affective affective valence valence of of a a stimulus stimulus may may be be accomplished accomplished in much much the the same same way way as as recognition recognition of of semantic semantic information information was was assessed. assessed. Specifically, Specifically, the the individual individual may may be be presented presented with with an an affective affective stimulus stimulus and and asked asked to to respond respond by by pushing pushing aa button button representing representing the the affeetive affective valence valence of ofthe the stimulus. stimulus. The The time time it it takes takes the the individual individual to to respond respond is is assumed assumed to to be be analogous analogous to to the the time time it it takes takes the the
3 18 318
Chapter 7
of the stimulus. This task task will be individual to recognize the affective valence of termed an affective valence-identification task. As with the lexical decision
of task, the valence identification task appears to be a relatively pure measure of an individual's ability to associate a stimulus with a mental representation of of aff ective valence. affective Despite predicting interference on the identification of of the the semantic content of 1 98 1 ) network theory predicts of most negative stimuli, Bower's ((1981) facilitation f or the identification of for of the affective content of of negative stimuli on a valence identification task. That That is, when a depressed person is exposed to a negative stimulus, their their representation of of sadness is expected to become activated. The mental representation of of events associated with sadness, in tum, turn, become activated, and via the feedback process described by Ingram Ingrain ((1984), 1 984), reactivate the mental representation of of sadness. In contrast, when a non-negative stimulus is presented, any incidental activation of of the nodes
affective representing a negative aff ective valence begins the spiraling negative loop which may increase the activation the node representing negativity and cause interference with the identification of of the stimulus as positive. Thus, Bower's ((1981) 1 98 1) theory predicts that that depressed individuals may be be delayed in identifying the affective valence of of positive words. By contrasting results of or a of an affective lexical lexieal decision task task with that ffor valence identification task, it may be possible to determine the relative
magnitude to which the affective valence of of stimuli interferes with an individual's ability to identify their semantic content. The idea that the
of a stimulus can can interf interfere ere with its semantic identification affective content of affective can be referred to as affective interference. The possibility that such aff ective
interf erence is responsible f or the attention biases that may occur in people interference for with features of of depression can
be ref erred to as the aff ective interference referred affective hypothesis. The aff ective interf erence hypothesis, and its implications f or affective interference for understanding depressive information processing are described in detail by 1 996a). Siegle et a1. al. ((1996a). Siegle et al. ((1996a) 1 996a) demonstrate affective interference empirically using Siegle an affective valence identification task, in which participants are asked to affective of briefly identify the aff ective valence of briefly presented stimulus ((150ms) 1 50ms) which of 30 may be positive negative, or neutral. They show that the same sample of as were used for their affective lexical dysphoric research participants as decision task were, in general, delayed in responding to positive words with respect to negative words on the affective valence identification task. The for individuals. Their same delays were not present f or the 46 nondepressed individuals. results are illustrated in Figure 2.
G.J G.J. Siegle and R.E. R.E. Ingram
3319 19
Sicglc et et al. al. ((1996b) simulate these these results results similarly similarly to to the the way way they they Siegle 1 996b) simulate simulated the the lexical lcxical decision decision task. task. Counters Counters are arc kept kept for for the the activation activation of of the the simulated positive and and negative negative valences valences.. positive (arbitrarily set set to to (arbitrarily
When any any counter counter exceeds exceeds aa threshold threshold When
1), the the network network is is said said to to have have made made aa valence valence 1),
identification. The The network network judged judged aa stimulus stimulus to to be be neutral neutral when when little little identification. evidence was was accumulated accumulated for for either either valence valence (both (both accumulators accumulators less less than than evidence after aa temporal temporal threshold threshold of of 1132 epochs plus plus gaussian gaussian noise. noise. Siegle Sicgle et ct 32 epochs 0.8) after al. ((1996b) show that that when when the the network network is is trained trained for for 70 epochs epochs on on aa negative negative al. 1 996b) show stimulus after after having having received its original training, the the network network assumes assumes stimulus received its original training, relatively similar similar inf information processing biases biases to to those those observed observed for for relatively ormation processing dysphoric individuals individuals on on the the affective affectivc valence valence identification identification task. task. Results Results for for dysphoric this case ease are are depicted depicted in in Figure Figure 33.. Siegle Sicgle et ct al al ((1996b) show that that simulated simulated this 1996b) show reaction times times to to negative negative words words in in the the network network which which received received overtraining ovcrtraining reaction more closely closely resembles resembles those those for for the the network network which which did did not not receive receive more ovcrtraining, when when a a distributed representation of of stimuli stimuli is is used2 used 2.. overtraining, distributed representation
A distributed representation representation was was not not used used for for the the current current simulations simulations so so as as to to distributed preserve as as interpretable interpretable aa representation representation as as possible. possible. The The authors authors also also show show preserve that the the magnitude magnitude of of information information processing processing biases biases increases in direct direct that increases in proportion to to the the overtraining ovcrtraining on on negativity negativity received received by by the the network. network. The The proportion network parameters in the the simulations simulations of of the the affective affcctivc lexical lcxical decision decision network parameters used used in task and and affective affectivc valence valence identification identification task task are are shown shown in in Table Table 11.. Siegle Sicglc et et task al. show that that the nctwork's performance performance is is qualitatively qualitatively similar similar when when al. ((1996b) 1996b) show the network's aa number number of of these these parameters parameters arc are manipulated. manipulated.
A weakness in the model: The need for for investigation ofsubpopulations ofsubpopulations The simulations simulations described above provide provide some for the the The described above some validation validation for proposed variation around around mean mean reaction reaction times proposed connectionist connectionist framework. framework. All All variation times in random noise be present in the the model model is is induced induced by by allowing allowing random noise to to be present in in various various aspects at first, that all all variation variation aspects of of the the network. network. This This feature feature might, might, at first, suggest suggest that about data is due to to entirely entirely about mean mean reaction reaction times times in in the the human human data is effectively effectively due random random variation. variation. Yet, Yet, the the human human data data on on which which the the model model is is based based does does not not reflect reflect entirely entirely random random variation variation about about mean mean reaction reaction times. times. Instead, Instead, observed observed reaction reaction times times for for aa given given condition condition are are often often multimodal multimodal and and the the variance variance about means about means does does not not reflect reflect homoscedasticity homoscedasticity (Siegle (Siegle et et al., aI., 1996a). 1 99
