Page i
Developing Theories of Intention Social Understanding and SelfControl Edited by Philip David Zelazo Janet Wil...
340 downloads
871 Views
3MB 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
Page i
Developing Theories of Intention Social Understanding and SelfControl Edited by Philip David Zelazo Janet Wilde Astington David R. Olson University of Toronto
Page ii
Copyright © 1999 by Lawrence Erlbaum Associates, Inc. All rights reserved. No part of this book may be reproduced in any form, by photostat, microfilm, retrieval system, or any other means, without prior written permission of the publisher. Lawrence Erlbaum Associates, Inc., Publishers 10 Industrial Avenue Mahwah, NJ 07430 Cover design by Kathryn Houghtaling Lacey. Cover art: bill bissett, "th sky within us" (1998, mixed media). Private collection, Toronto, Canada. Reproduced with permission.
Library of Congress CataloginginPublication Data Developing theories of intention : social understanding and selfcontrol / edited by Philip David Zelazo, Janet Wilde Astington, and David R. Olson. p. cm. Includes bibliographical references and index. ISBN 080583141X (cloth : alk. paper)—ISBN 0805831428 (pbk. : alk. paper). 1. Selfcontrol in children. 2. Social perception in children. I. Zelazo, Philip David. II. Astington, Janet W. III. Olson, David R. BF723.S25D46 1998 153.8—dc21 9850446 CIP Books published by Lawrence Erlbaum Associates are printed on acidfree paper, and their bindings are chosen for strength and durability. Printed in the United States of America 10 9 8 7 6 5 4 3 2 1
Page iii
For William Kessen (d. 1999)— humanist, scholar, friend
Page v
CONTENTS Preface
ix
1 Introduction: Actions, Intentions, and Attributions
1
David R. Olson, Janet Wilde Astington, And Philip David Zelazo Part I Development of Intention and Intentional Understanding in Infancy and Early Childhood
15
2 Toddlers' Understanding of Intentions, Desires and Emotions: Explorations of the Dark Ages
17
Andrew N. Meltzoff, Alison Gopnik, And Betty M. Repacholi 3 Intentional Relations and Triadic Interactions
43
Chris Moore 4 Having Intentions, Understanding Intentions, and Understanding Communicative Intentions
63
Michael Tomasello 5 Intentions, Consciousness, and Pretend Play
77
Michael Lewis And Douglas Ramsay 6 Language, Levels of Consciousness, and The Development of Intentional Action Philip David Zelazo
95
Page vi
7 Development of Intention: The Relation of Executive Function to Theory of Mind
119
Douglas Frye 8 Executive Functions and Theory of Mind: Cognitive Complexity or Functional Dependence?
133
Josef Perner, Sandra Stummer, And Birgit Lang 9 The Theory of Ascriptions
153
David R. Olson And Deepthi Kamawar Part II Comparative Perspectives on Intentionality
167
10 Primate Representations and Expectations: Mental Tools for Navigating in a Social World
169
Marc D. Hauser 11 Social Understanding in Chimpanzees: New Evidence From a Longitudinal Approach
195
Daniel J. Povinelli Part III The Sociocultural Context of Intentionality
227
12 Making Sense of the Social World: Mindreading, Emotion, and Relationships
229
Judy Dunn 13 Influences on Maternal Attribution of Infant Intentionality
243
J. Steven Reznick 14 Intention and Emotion in Child Psychopathology: Building Cooperative Plans Jenny Jenkins And Rachel Greenbaum
269
Page vii
Part IV Intentionality and Language
293
15 The Language of Intention: Three Ways of Doing It
295
Janet Wilde Astington 16 Intentionality and Interpretation
317
Carol Fleisher Feldman 17 The Intentionality of Referring
329
Jerome Bruner
Author Index
341
Subject Index
351
Contributors
357
Page ix
PREFACE The current volume grew out of a small conference, Developing Intentions in a Social World, that was held at University College, University of Toronto, in April, 1997. The conference was designed to assess the "state of the art" of research on the development of intention visàvis social understanding and selfcontrol, but it also served to commemorate the university's long history as a center for inquiry into this topic. In 1891, J. Mark Baldwin established the first fully equipped psychology laboratory at University College (apparently the first in the British Commonwealth) and began a seminal series of studies (e.g., Baldwin, 1891, 1892a, 1892b, 1894) on the emergence of intentional imitation in infancy and its relation to the developing socius—the child's sense of self and other. These studies culminated in his landmark books, Mental Development in the Child and the Race (1895) and Social and Ethical Interpretations in Mental Development (1897), the influence of which has been pervasive but, until recently, largely filtered through the writings of Jean Piaget. A renewed interest in the issues addressed directly by Baldwin probably derives in part from several sources including: a growing appreciation of the sociocultural work of Vygotsky and Luria; recent work in developmental neuropsychology on the topic of prefrontal cortical function; and widespread fascination with the problem of consciousness. However, one obvious antecedent of this interest is research on children's beliefs about the mind—their "theory of mind." It will be noted that many of the contributors to the current volume were participants in the 1986 Developing Theories of Mind Conference at the University of Toronto. That conference, organized by Janet Astington, Lynd Forguson, Alison Gopnik, and David Olson, helped to establish theory of mind as a major focus on research. As a result of that research, we have learned a great deal about children's egocentrism and their ideas about thought and misleading appearances. Now, however, many theoryofmind researchers are returning to questions posed more than a century ago by Baldwin: questions about the control of action and how this control is related to children's developing selfconsciousness and their increasingly sophisticated appreciation of other people's perspectives. As will be clear from the
Page x
following chapters, intention is at the intersection of current research on imitation, early understanding of mental states, goaldirected behavior in nonhuman animals, executive function, language acquisition, and narrative understanding, to name just a few of the relevant foci. By bringing these lines of research together in a single volume, we hope to shed light on several questions: What are the psychological processes underlying intentional action? To what extent does the use of intentions (i.e., actionoriented plans) by individuals depend upon socialization? Is it necessary to know about intentions (in self and others) in order to act intentionally? Do individuals attribute intentionality to others by analogy to their own experience of acting intentionally? What is the relation between individual and shared (or mutual) intentions? No doubt, these are extremely difficult questions to answer, but they would seem to be difficult in direct proportion to their importance for an adequate understanding of mind. The talks at the conference were organized into sessions corresponding to the sections in this volume. However, several motifs soon became apparent, including: (1) the possibility of a causal relation between the development of intentional action in infancy and the subsequent understanding of intentional states in others; (2) the potential importance of parental attributions of intentionality to the development of intentional action in infancy and early childhood; (3) the close (and possibly complex) relation between executive function and theory of mind; and (4) the instrumental role that language appears to play in the development of social understanding and selfcontrol. These motifs are maintained in the chapters, which were contributed by nearly all of the 18 participants in the conference (regrettably, James Russell declined due to prior commitments). The conference was presented by the Cognitive Science Program at University College (University of Toronto), under the auspices of the Principal of University College, Lynd Forguson. Principal Forguson convened the conference planning committee, which included Keith Oatley, in addition to the editors of this volume. We would like to thank Lynd and Keith for their hard work and perspicacity, and Coleen McColeman, of the Principal's Office, University College, for her secretarial support to the conference organizers. Of course, the conference could not have been held without generous financial support from several sources including the Connaught Fund; University College; the Department of Psychology; the Office of the Vice President and Provost, University of Toronto; Field Services and Research, the Ontario Institute for Studies in Education/University of Toronto (OISE/UT); and the Laidlaw Research Centre, Institute of Child Study, OISE/UT. We are very grateful for that support. We are also grateful for the assistance of those students who graciously volunteered their time: Paul Andrews, Tate Avery, Carla Baetz, Sandra Bosacki, David Collister, Bruce
Page xi
Homer, Sophie Jacques, Deepthi Kamawar, Elizabeth Lee, Stuart Marcovitch, Bruce Morton, Terri Sloss, Tracy Solomon, and Jessica Sommerville. Finally, we are grateful to Judi Amsel, Executive Editor at Lawrence Erlbaum Associates, for her timely interest and helpful advice, and her unflagging efforts to expedite the publication of this volume. —PHILIP DAVID ZELAZO —JANET WILDE ASTINGTON —DAVID R. OLSON
Page 1
Chapter 1— Introduction: Actions, Intentions, and Attributions David R. Olson Janet Wilde Astington Philip David Zelazo University of Toronto Disciplines, like weedpatches, grow most prolifically at their edges, so perhaps it is not surprising that cognitive science, occupying a space between the biological sciences and the social sciences, should show its greatest signs of growth at its borders. At the first border, we find serious consideration of the question, ''How can a biological mechanism, the brain, give rise to intentional states?" At the second border, we see the equally important question, "How can seemingly private intentional states reflect public forms of talk and action?" Together, these questions capture Brentano's (1874/1973) insight that intentionality is the mark of the mental, and they help to frame developmental questions concerning the origins of intentionality. Development of Intentionality In Brentano's (1874/1973) sense, the term intentionality refers to that property of directedness at an internal object that distinguishes not just propositional attitudes (e.g., belief), but even the simplest mental states (e.g., pain; Brentano, 1874/1973, p. 89 ff.) from those states of the world that are merely physical (see Zelazo, 1996, for discussion). In a series of studies started just over a century ago at the University of Toronto, Baldwin (e.g., 1891, 1892a, 1892b) laid the foundation for his comprehensive account of the origin and development of intentionality (Baldwin, 1897). In this
Page 2
work, Baldwin addressed both the biological and the social aspects of intentionality, and considered intentionality not only in Brentano's (1874/1973) sense, but also in the everyday sense of "purposeful." The everyday sense of intentionality is closely related to Brentano's sense, both etymologically and conceptually (pace Meltzoff, Gopnik, & Repacholi, chap. 2, this volume), because intentional actions are directed at goals in very much the same way that mental states are directed at their objects (i.e., what they are of; see Crane, 1998, for a discussion of common misunderstandings of Brentano's notion of intentionality). Moreover, the two senses are definitionally dependent because, for many authors, intentional action is goaldirected behavior that is accompanied by (or caused by) a particular type of intentional state; namely, an intention (e.g., Adams, 1986). Similarly, Baldwin tied intentional action to the emergence of desire, deliberation, and effort: the conscious representation of a goal, the active consideration of alternative means and ends, and the feeling accompanying the selection and execution of a plan. Although early behaviors are mediated by consciousness, and are often complex and welladapted, they are not deliberate. For Baldwin, the maturation of a coordinating center at the end of the first year of life makes possible the conscious represention of goals and deliberation among means (cf. Olson, 1993, on "holding in mind"; Zelazo, chap. 6, this volume), and hence, intentional action. Baldwin (1897) clearly recognized that the development of intentional action—the deliberate use of goals and means—is inseparable from the development of the infant's understanding of intentionality (in its broader sense) in others. According to Baldwin, ego and alter are ideal ends of a continuum that he defined as the socius (which is a synthesis of Self and Other). Ego (one's sense of Self) and alter (one's sense of the Other) develop together through a circular process. First, one observes projective behavior, that is, behavior seen only in terms of its outward aspects. Then, by imitating this projective behavior, one comes to comprehend the subjective side of it, for example, the affect that accompanies it. Finally, one ejects this subjectivity back into the project. Thus, a father may prick himself with a pin, and his child will imitate the behavior, come to comprehend the painful consequence, and then believe that the father felt it, too. Third and firstperson perspectives on behavior are thereby synthesized. Through the dialectic, the child expands the scope of his or her identifications. Baldwin's (1897) slogan, "It is not I, but I am to become it" (p. 36) anticipated Freud's (1933/1940) famous dictum, "Where It was, there I shall be (Wo Es war; soll Ich werden)" (p. 86). By understanding oneself in relation to the Other one acquires means for acting intentionally and comes to conceptualize oneself as separate from a mindindependent world upon which one can act.
Page 3
The essential role of the Other in the development of intentionality and intentional action has been a central concern of social theorists as well as psychologists. Radical social theorists such as Marx and Durkheim (Lukes, 1973) attempted to explain human behavior by appeal to social, not psychological, causes. The real fact, for such theorists, is "the fact of the dependence of the mental (and not only mental) activity of the individual on the system of culture established before him and completely independently of him, a system in which the [mental life] of every individual begins and runs its course" (Ilyenkov, 1977, p. 78). Vygotsky's (1978; see also Luria, 1961) contribution to this approach was to formulate the relation between mind and society in a developmental framework. According to Vygotsky, children's psychological structures come to resemble those of adults in their culture through the internalization of social practices, and behaviors first engaged in interpersonally come to be managed intrapersonally. This is no less true for simple procedures, such as how to play a game of marbles, than for understanding the intentional states of others. Beliefs, desires, and intentions are thus seen to both cause and reflect participation in a social world (see Bruner, chap. 17, this volume; Feldman, chap. 16, this volume). Although the contributions of social life to the development of intentionality likely are essential, it is clear that children must first possess a set of internal, biologically given, cognitive resources in order to benefit from social and cultural practices. These resources, which continue to develop throughout childhood, are much more scarce among our primate cousins, and possibly absent altogether in other animals (for discussion, see Hauser, chap. 10, this volume; Povinelli, chap. 11, this volume; Tomasello, chap. 4, this volume). What children bring by way of resources, and how this interacts with whatever it is that social interactions provide, remains a central, and still unresolved, issue. Indeed, the interaction between these biologically given resources and the social environment is prominent in recent work, including that of Frye (1981, 1991), Moore (1996; also see Moore, chap. 3, this volume), Meltzoff, Gopnik, and Repacholi (chap. 2, this volume), and Tomasello (chap. 4, this volume), and forms the central theme of the current volume: How are we, as scientists, to understand the interplay between biological and social influences on the development of the ability to have intentional states and attribute them to others? More specifically, however, the work reported in this volume can be seen as addressing a common set of three questions about intentionality in general and intentional action in particular. First, on what grounds can we call behavior intentional (i.e., accompanied by the intentional state of intention)? Second, what is the relation between having intentions and having a concept of intention? Third, what is the function or use of the ability to
Page 4
attribute intentional states (including intention) to oneself and others? These questions will now be considered, in turn. When Does Behavior Become Intentional? It is commonly suggested that children below a certain age, the age of accountability, are not intentional, moral agents. Similarly, other animals, including primates, are not moral creatures in that their actions cannot be judged by appeal to social norms. Praise and blame, as opposed to rewards and punishments, are relevant only to creatures who, as we say, know the difference between good and evil. But what are the criteria for deciding that behavior is intentional, and further, for deciding whether it merits reward and punishment, or praise and blame? Searle (1983) distinguished between intentioninaction and prior intention, the latter being involved in deliberate or premeditated actions (see Astington, chap. 15, this volume). Prior intention allows for planning an action and also for deliberately choosing not to act. It seems safe to suggest that actions based on a prior intention can clearly be construed as moral. However, what about intentioninaction? Certain unpremeditated actions (e.g., spontaneous gestures of comfort, or sudden aggressive acts) are often praised or blamed, although this is probably because we assume that they were motivated by a more general prior intention, to do good or to harm, for example. In considering intentionsinaction, there is an important distinction to be made between behavior that is directly "suggested" (Baldwin, 1891) by a situation—grasping for example—and what is properly referred to as "action" (and associated with intentionsinaction). Actions, in this sense, are distinguished by the dissociation between means and ends, as Baldwin (1892b) and Piaget (1936/1952) noted. They imply an end held in view while means are employed repeatedly or alternative means are sought (cf. Bruner, 1973). Such action may or may not be present in lower animals (see Hauser, chap. 10, this volume; Povinelli, chap. 11, this volume; Tomasello, chap. 4, this volume) but it is perhaps first seen in children at the end of the first year in their persistent attempts at imitation (Baldwin, 1892b). If this is the case, it raises the possibility that even intentioninaction is directly related to the recognition of the mental states of others. As Meltzoff shows, 18montholds imitate what others are intending to do, not what they actually do; that is, they appear to recognize the other's intention (see Meltzoff, Gopnik, & Repacholi, chap. 2, this volume). It might be thought that this creates a problem, in that the recognition of belief in others, as indicated by an understanding of false belief (Wimmer & Perner, 1983), does not appear until about 4 years of age. However, it may be that nonepistemic states, such as an intention, are more easily recognized, or it may be that the early recognition of intention does not
Page 5
involve metarepresentation (Astington, 1991). Certainly, other things we associate with the metarepresentation of intention, such as selfcontrol or acting in terms of a social or moral norm, continue to develop well past infancy. There are a number of important developmental achievements here, and the chapters in the first section of this volume attempt to sort them out (e.g., Lewis & Ramsay, chap. 5, this volume; Meltzoff, Gopnik, & Repacholi, chap. 2, this volume; Moore, chap. 3, this volume; Zelazo, chap. 6, this volume). Intentions and Concepts of Intention Historically, intention played a central role in psychological explanation. Conation or volition was one of the trinity of mental faculties—cognition, conation, and affect—that came down to us from classical times. Intention was changed from a faculty into a function in modern psychology by William James, John Dewey, and J. Mark Baldwin, among others. Baldwin (1897), for example, argued that a sense of personal agency arose in the context of solving problems (e.g., imitating others), and lay at the root of social consciousness. However, once one can do as one likes, one can also be held accountable for what one does and this accountability appears to play a role in the further development and elaboration of the concept of intention. Thus, the nature and development of the concept of intention is a central topic in moral development and the development of social cognition, and has been for some time. Nonetheless, it seems likely that the topic of intention has inherited some of its recent popularity from the now widespread interest in children's theory of mind. Like the present volume, Developing Theories of Mind (Astington, Olson, & Harris, 1988) grew out of a conference held at the University of Toronto and helped to crystallize what has become one of the liveliest areas of developmental research. In the introduction to that volume it was argued that in acquiring a theory of mind, "children begin to recognize mental states in themselves and others; they come to recognize beliefs as beliefs, desires as desires, and intentions as intentions" (Olson, Astington, & Harris, 1988, p. 1). Although much has since been learned about children's understanding of intentional states such as belief and, to a lesser extent, desire, we still know relatively little about children's understanding of intention. We know arguably even less about the development of intention itself, and the relation between intention and concepts of intention. In the rush of new work investigating children's theory of mind, it quickly became clear that important developments occur around 4 years of age in children's reasoning about beliefs. Children younger than age 4 tended to attribute to others the beliefs they themselves held, namely, those congruent with the current situation. Those older than age 4 ascribed
Page 6
beliefs on the basis of the causes of those beliefs, namely, what the other had seen, heard, and done. Consequently, these children could ascribe to others beliefs that they themselves knew to be false (hence the significance of the "false belief task" to the understanding of children's theory of mind; Wimmer & Perner, 1983). To some, this development seemed little more than yet another example of Piagetian egocentrism, rather than evidence of failure to understand beliefs, because children did ascribe to others the beliefs they held themselves. However, the picture was soon complicated by the finding that children were no better at understanding their own beliefs than those of others (Gopnik & Astington, 1988; Wimmer & Hartl, 1991). In contrast, children much younger than 4 years of age can understand that others may have desires different from their own, and can remember earlier desires, different from their present ones (Gopnik & Slaughter, 1991; Repacholi & Gopnik, 1997; but see Moore et al., 1995). Such young children also understand that desires motivate actions and they refer to desire states in their speech a year or more before they refer to beliefs (Wellman, 1991). Whereas the concepts of belief and desire quickly found their place in developmental and evolutionary theories of mind, the concept of intention has remained enigmatic. Intentions are more like desires than beliefs. That is, they are neither true nor false but are fulfilled (or not) by outcomes in the world and indeed, serve to motivate action to bring about those outcomes. Certainly, much of the time our desires and our intentions coincide. Such considerations may lead to the assumption that children understand intention when they understand desire. There are, however, important differences between desire and intention (Astington & Gopnik, 1991). Intentions are selfreferential, that is, they are not fulfilled by outcomes in the world unless the intention itself causes the action that brings about that outcome (Searle, 1983). Intentions are also more closely tied to beliefs (Moses, 1993); one can desire, but not intend, what one believes to be impossible. Frequently, theory of mind is used to explain behavior as a product of belief and desire, leaving unexplored just how beliefs and desires turn into intentions or how prior intentions and plans are formed or executed. Perhaps we may get a clearer picture of how intentions are related to other intentional states by thinking of intentional states as generated by a mental calculus. The Mental Calculus Consider the problem of determining when behavior is directly caused by external or internal stimuli and when behavior is, rather, a product of mental representation, including belief, desire, and intention. Can an animal's behavior, say that of a dog chasing a cat, be correctly characterized as intentional? The dog certainly has a goal, uses a variety of means, and even maintains the memory of the goal over time.
Page 7
However, intentional states, including beliefs, desires, and intentions, are arguably constituted by a calculus of agents, modes, and contents: AGENT
MODE
CONTENT
John—
believes—
the dog wants to go out
Mary—
intends—
to let the dog out
Sam—
wants—
the dog to be out
Searle (1983) remarked, "Only someone in the grip of a philosophical theory would deny that small babies can literally be said to want milk and that dogs want to be let out or believe that their master is at the door" (p. 5). However, if a creature can manifest only one of the modes, say, intending, then there are less secure grounds for saying that the behavior is intentional. Hence, contra Searle, the dog who, when shown a bone, can only try to get the bone, rather than think that it is a bone, may not be well described as intentional. Instead, the behavior would seem to be under the control of a schema for bone getting (suggested by the bone). It may be that only linguistic creatures with access to the sort of mental calculus just mentioned, creatures who can both treat expressions as asserting something as true ("This is a bone") and initiate an action can be correctly described as intentional in that they have a set of optional modes for relating themselves to the world. Agreeing on criteria (including linguistic criteria) for recognizing intention remains a critical task, and is explored in more detail in several of the chapters in this volume (Astington, chap. 15, this volume; Feldman, chap. 16, this volume). Structural and Functional Explanations of Intentional Understanding Although the origins of an understanding of intentional states in general and intentions in particular may be traced to structural changes in the organization of the mind and ultimately the brain, the origins of this understanding can also be examined in terms of the functions that these understandings sustain. Piaget's (1936/1952) account of the development of the representational function is a classic example of a structural change. A more recent structural change explanation of the development of an understanding of intentionality was articulated by Leslie (1987), who proposed a specific "decoupling" device that serves to sever the link between situation and content, this content thereby being eligible for metarepresentation. Others in this tradition have pointed to other plausible mechanisms: changes in the executive functions of prefrontal cortex that allow perspective taking (for discussion see Frye, chap. 7, this volume; Perner, chap. 8, this volume; Zelazo & Frye, 1998); changes in memory (e.g., Gordon & Olson,
Page 8
1998; Olson, 1993) that allow children to "hold in mind" a representation across changing situations; or the development of higherorder levels of consciousness that support selfreflection and psychological distancing (Lewis & Ramsay, chap. 5, this volume; Zelazo, chap. 6, this volume). But, it may be argued, structural change occurs to sustain particular functions and it is the function of participating in a social world that is responsible for the growth of an understanding of intentional states. The functions served by such an understanding are both social and cognitive. Socially, the attribution of intentional states allows for increasingly sophisticated social actions and interactions. Not only does it allow the child to recognize the intention behind the action of another, and therefore, to cooperate in the achievement of the action, but it also allows for the transmission of culture through learning and imitation, a point that is explored in the chapters in the third section of this volume (Dunn, chap. 12, this volume; Jenkins & Greenbaum, chap. 14, this volume; Reznick, chap. 13, this volume). The cognitive functions served by the development of intentionality are metacognitive ones, allowing a new kind of consciousness of mental life, not only for introspection of one's own thoughts but also of the sources of one's own beliefs. This is the reflective kind of consciousness that usually warrants the label thinking. Such metacognitive competence is critical to the formation of planned, premeditated actions, the kind of deliberate actions we associate with the "age of accountability." Indeed, the planfulness of one's own actions is identical in form to the participation in cooperative actions with others, a contingency that invites the inference that this aspect of cognitive development is more appropriately thought of as social development (see Jenkins & Greenbaum, chap. 14, this volume). A recent focus on individual differences in the development of intentional understanding has shown this understanding to be related to certain characteristics of children's social interactions (Dunn, chap. 12, this volume). For example, demographic approaches find relations between a child's understanding of false belief and the size of the child's family, particularly the number of siblings the child has (Jenkins & Astington, 1996; Lewis, Freeman, Kyriakidou, MaridakiKassotaki, & Berridge, 1996; Perner, Ruffman, & Leekam, 1994). A relation also exists between children's theory of mind and family background measures, such as parental occupation and education (Cutting & Dunn, in press), or parenting style (Vinden, 1997). Such findings lead to the argument that involvement in social relations is the precise causal mechanism through which social understanding and selfcontrol are acquired. Engagement in a social world, especially the world of complex social interactions mediated by language, is seen to be the immediate cause of the child's acquisition of an understanding of his or her own mental states, as well as those of other people. Dunn (chap. 12, this
Page 9
volume) cautions against making causal inferences from such correlational data; strong evidence, best provided by longitudinal and training studies, is required to support the causal argument. We do, however, know that particular patterns of social interaction and language use in the home, such as sibling cooperation, social pretend play, and family talk about feeling states, are associated with later differences in false belief and emotion understanding (Dunn, Brown, Slomkowski, Tesla, & Youngblade, 1991; Youngblade & Dunn, 1995). In addition, participation in training procedures involving, for example, conversations about surprise, has been shown to promote false belief understanding (Appleton & Reddy, 1996). Consequently, social factors do appear to have an impact on the child's acquisition of a theory of mind. Such social perspectives are sometimes seen to be in conflict with more cognitive ones insofar as these perspectives seem to see little need for specific cognitive mechanisms to acquire adult forms of competence (Raver & Leadbeater, 1993). Sometimes, these perspectives seem to assume that the adult forms of competence are the inevitable product of practice interpreting actions and speech in a social world. Nonetheless, most researchers agree that developing cognitive mechanisms are involved, and it remains to be determined how best to characterize these mechanisms. Any complete account would have to acknowledge both the structural and functional properties of intentionality: There are brain mechanisms that make the formation of intentional states possible; there are functional roles that these states play; these functional roles are, at least in part, and perhaps originally, social ones; and finally, these states can themselves become subject to reflection. Organization of the Volume The chapters in this volume, too rich to be summarized briefly, are organized into four sections, although a number of themes cut across the four sections. The first and longest section, focusing on the development of intention and intentional understanding in infancy and early childhood, is central to all of the issues we have raised in the introductory chapter. For example, Meltzoff, Gopnik, and Repacholi (chap. 2) argue that the origins of social understanding and selfcontrol can be seen in neonatal imitation, and they trace the development of social understanding from this early, intrinsically interpersonal context through what they refer to as the "dark ages" of the toddler period. Tomasello (chap. 4) argues that at the end of the first year, infants can infer others' intentions because they recognize the intentionality of their own actions, and because they recognize others as creatures like themselves. Zelazo (chap. 6) argues that increasingly higher levels of consciousness, mediated by language, allow more complex knowledge structures to govern children's intentional actions.
Page 10
Lewis and Ramsay (chap. 5) also discuss the development of increasingly higher order levels of intention that yield a selfreflective understanding. Frye (chap. 7) and Perner, Stummer, and Lang (chap. 8) consider how best to characterize the wellestablished relation between action control and theory of mind in preschoolers. Some authors' focus is not on the development of intention per se; instead, they are concerned with intentionality in the broader sense. For example, Moore (chap. 3) discusses infants' understanding of attention within an interactive context, arguing that infants understand intentional relations as a property of the interaction, not as a property of either the self or the other exclusively. Olson and Kamawar (chap. 9) discuss intentional attributions, focusing on belief attribution, and propose a strict criterion for the possession of intentional states, that is, the ability to ascribe such states to self and other. The chapters in the second section focus on comparative perspectives on intentionality; they too are concerned with issues of understanding and attributing intentional states, broadly considered. Hauser's (chap. 10) topic is the extent to which humans and other primates differ with respect to the causal relation between representation and expectation. He discusses primates' expectations concerning communicative signals, selfpropelled motions, and the greatest challenge, mental states, suggesting that primates know something about communication and animacy, if not belief. Povinelli (chap. 11) summarizes his extensive body of research on chimpanzees' understanding of seeing, leading to the perhaps startling conclusion that although chimpanzees follow another's gaze, there is no evidence for an intentional understanding of seeing, even though similar gazefollowing behaviors in children are correctly so interpreted. The chapters in the third section lead on from there to discuss the peculiarly human sociocultural context of intentionality. Dunn (chap. 12) makes a compelling argument for the significance of the emotional context in children's acquisition of intentional understanding. She argues that it is emotionally charged social experiences that play a vital role in the development of social understanding, right from the beginning. Reznick (chap. 13) shows how readily mothers attribute intentionality to infant behaviors displayed on videotape or described in a questionnaire, and Dunn (chap. 12) makes the point that in the real world such interactions are rarely emotionally neutral. Jenkins and Greenbaum (chap. 14) are also concerned with the link between cognition and emotion in their investigation of children's ability to engage in cooperative endeavors. They find that children whose emotional lives are dominated by feelings of anger and aggression are less able to negotiate cooperative goals, and less skilled in using intentional terms to discuss joint plans. The fourth and final section, on intentionality and language, picks up a number of issues from the earlier chapters: the themes that cut across the
Page 11
four sections. Astington's (chap. 15) discussion of the mediating role of language in the development of intentional understanding relates to Jenkins and Greenbaum's (chap. 14) findings, and reflects back to Zelazo's (chap. 6) focus on the role of language in the genesis of action. The acquisition of language depends on the infant's understanding of communicative intentions, as Tomasello (chap. 4) argues, and the development of language allows for the ascription of intentionality, as Olson and Kamawar (chap. 9) argue. Feldman (chap. 16) emphasizes that intentions are understood within an interpretive community, which links back to the chapters on the sociocultural context of intentional understanding, and also to Moore's (chap. 3) argument that intentional relations are first construed as part of the interactive context, not as independently belonging to self or other. Bruner's (chap. 14) focus on infant recognition and parental attribution of epistemic and intrumental intentions relates to Astington's (chap. 15) and Feldman's (chap. 16) concerns, and Tomasello's (chap. 4) and Reznick's (chap. 13) chapters also find echoes here. In conclusion, cognitive science, as heir to the discipline that William James (1890/1950, p. 1) defined as ''the Science of Mental Life" centers on intention and intentional understanding—the ability to make judgments of truth and goodness, and the ability to act intentionally in terms of those judgments. These abilities may well turn out to be found only in persons with sophisticated cognitive and linguistic powers—powers that are not only representational, but also metarepresentational and, more to the point, the product of participation in a social world. If so, then a full understanding of these abilities will require understanding of the cognitive and cultural determinants of intentionality and its development. We believe that the contributors to this volume take important steps toward this goal. References Adams, F. (1986). Intention and intentional action: The simple view. Mind and Language, 1, 281301. Appleton, M., & Reddy, V. (1996). Teaching 3yearolds to pass false belief tests: A conversational approach. Social Development, 5, 275291. Astington, J. W. (1991). Intention in the child's theory of mind. In D. Frye & C. Moore (Eds.), Children's theories of mind (pp. 157172). Hillsdale, NJ: Lawrence Erlbaum Associates. Astington, J. W., & Gopnik, A. (1991). Developing understanding of desire and intention. In A. Whiten (Ed.), Natural theories of mind: Evolution, development and simulation of everyday mindreading (pp. 3950). Oxford: Basil Blackwell. Astington, J. W., Olson, D. R., & Harris, P. L. (Eds.). (1988). Developing theories of mind New York: Cambridge University Press. Baldwin, J. M. (1891). Suggestion in infancy. Science, 18, 113117. Baldwin, J. M. (1892a). Infants' movements. Science, 19, 1516. Baldwin, J. M. (1892b). Origin of volition in childhood. Science, 20, 286288.
Page 12
Baldwin, J. M. (1897). Social and ethical interpretations in mental development: A study in social psychology. New York: Macmillan. Brentano, F. (1973). Psychology from an empirical standpoint. (O. Kraus, Ed.; A. C. Ranculrello, D. B. Terell, & L. L. McAlister, Trans.). London: Routledge & Kegan Paul. (Original work published in 1874) Bruner, J. (1973). Organization of early skilled action. Child Development, 44, 111. Crane, T. (1998). Intentionality as the mark of the mental. In A. O'Hear (Ed.), Current issues in philosophy of mind (pp. 229251). Cambridge, England: Cambridge University Press. Cutting, A., & Dunn, J. (in press). Theory of mind, emotion understanding, language and family background: Individual differences and interrelations. Child Development. Dunn, J., Brown, J., Slomkowski, C., Tesla, C., & Youngblade, L. (1991). Young children's understanding of other people's feelings and beliefs: Individual differences and their antecedents. Child Development, 62, 13521366. Freud, S. (1940). Neuefolge der Vorlesungen zur Einfhrung in die Psychoanalyse [New introductory lectures on psychoanalysis]. In A. Freud, E. Bibring, & E. Kris (Eds.), Gesammelte Werke: XV (Whole volume). London: Imago Publishing Co. (Original work published 1933) Frye, D. (1981). Developmental changes in strategies of social interaction. In M. E. Lamb & L. R. Sherrod (Eds.), Infant social cognition (pp. 315331). Hillsdale, NJ: Lawrence Erlbaum Associates. Frye, D. (1991). The origin of intention in infancy. In D. Frye & C. Moore (Eds.), Children's theories of mind (pp. 1538). Hillsdale, NJ: Lawrence Erlbaum Associates. Gopnik, A., & Astington, J. W. (1988). Children's understanding of representational change and its relation to the understanding of false belief and the appearance reality distinction. Child Development, 59, 2637. Gopnik, A., & Slaughter, V. (1991). Young children's understanding of changes in their mental states. Child Development, 62, 98110. Gordon, A. C. L., & Olson, D. (1998). The relation between acquisition of a theory of mind and information processing capacity. Journal of Experimental Child Psychology, 68, 7083. Ilyenkov, E. V. (1977). The concept of the ideal. In R. Daglish (Ed. and Trans.), Philosophy in the USSR (pp. 7199). Moscow: Progress Publishers. James, W. (1950). The principles of psychology. New York: Dover. (Original work published in 1890). Jenkins, J. M., & Astington, J. W. (1996). Cognitive factors and family structure associated with theory of mind development in young children. Developmental Psychology, 32, 7078. Leslie, A. M. (1987). Pretense and representation: The origins of "theory of mind." Psychological Review, 94, 412426. Lewis, C., Freeman, N. H., Kyriakidou, C., MaridakiKassotaki, K., & Berridge, D. M. (1996). Social influences on false belief access: Specific sibling influences or general apprenticeship. Child Development, 67, 29302947. Lukes, S. (1973). Emile Durkheim: His life and work. Markham, ON: Penguin Books. Luria, A. R. (1961). The role of speech in the regulation of normal and abnormal behaviour (J. Tizard, Ed.). New York: Pergamon Press. Moore, C., Jarrold, C., Russell, J., Lumb, A., Sapp, F., & MaCallum, F. (1995). Conflicting desire and the child's theory of mind. Cognitive Development, 10, 467482. Moses, L. J. (1993). Young children's understanding of belief constraints on intention. Cognitive Development, 8, 125. Olson, D. R. (1993). The development of representations: The origins of mental life. Canadian Psychology, 34, 114. Olson, D. R., Astington,J. W., & Harris, P. L. (1988). Introduction. In J. W. Astington, P. L. Harris, & D. R. Olson (Eds.), Developing theories of mind (pp. 1 15). New York: Cambridge University Press.
Page 13
Perner, J., Ruffman, T., & Leekam, S. R. (1994). Theory of mind is contagious: You catch it from your sibs. Child Development, 65, 12281238. Piaget, J. (1952). The origins of intelligence in children (M. Cook, Trans.). New York: Norton. (Original work published in 1936). Raver, C. C., & Leadbeater, B.J. (1993). The problem of the other in research on theory of mind and social development. Human Development, 36, 350362. Repacholi, B. M., & Gopnik, A. (1997). Early reasoning about desires: Evidence from 14 and 18montholds. Developmental Psychology, 33, 1221. Searle, J. (1983). Intentionality. Cambridge: Cambridge University Press. Vinden, P. (1997, April). The effects of parenting style on theory of mind understanding. Paper presented at the Biennial Meeting of the Society for Research in Child Development, Washington, DC. Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Cambridge, MA: Harvard University Press. Wellman, H. M. (1991). From desires to beliefs: Acquisition of a theory of mind. In A. Whiten (Ed.), Natural theories of mind: Evolution, development and simulation of everyday mindreading (pp. 1938). Oxford: Basil Blackwell. Wimmer, H., & Hard, M. (1991). Against the Cartesian view on mind: Young children's difficulty with own false belief. British Journal of Developmental Psychology, 9, 125138. Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception. Cognition, 13, 103128. Youngblade, L. M., & Dunn, J. (1995). Individual differences in young children's pretend play with mother and sibling: Links to relationships and understanding of other people's feelings and beliefs. Child Development, 66, 14721492. Zelazo, P. D. (1996). Towards a characterization of minimal consciousness. New Ideas in Psychology, 14, 6380. Zelazo, P. D. & Frye, D. (1998). II. Cognitive complexity and control: the development of executive function in childhood. Current Directions in Psychological Science, 7.
Page 15
PART I DEVELOPMENT OF INTENTION AND INTENTIONAL UNDERSTANDING IN INFANCY AND EARLY CHILDHOOD
Page 17
Chapter 2— Toddlers' Understanding of Intentions, Desires, and Emotions: Explorations of the Dark Ages Andrew N. Meltzoff University of Washington Alison Gopnik University of California, Berkeley Betty M. Repacholi Macquarie University Our sensory experience of other people tells us about their movements in space but does not tell us directly about their mental states. Although a few radical philosophers and psychologists may deny the existence of mental states, most regular "folk" feel sure that they themselves and others have them. It is adaptive to read another person's mental state because it allows us to explain the actions they have taken in the past and predict their actions in the future. The general aim of "theory ofmind" research is to illuminate the development of this everyday, folk psychological framework for understanding people (e.g., Astington & Gopnik, 1991b; Astington, Harris, & Olson, 1988; Flavell & Miller, 1998; Perner, 1991; Taylor, 1996; Wellman, 1990). This research has taught us that children's understanding of mental life is not all of one piece. There is no single moment at which children develop a theory of mind. Instead, children gradually converge on an adult understanding of mind. The focus on when children understand "false belief" has been misleading in this regard. Beliefs are only one of many mental states that children understand and use in their everyday interactions with people. Children may only develop a firm understanding of false beliefs at about 4 years old, but they have started on their path of developing a folk psychological understanding of people much earlier. Preschoolers understand a great deal about perceiving, wanting, and intending at an age when they still have only a shaky understanding of false beliefs (e.g., Astington & Gopnik, 1991a; Flavell, Flavell, Green, & Moses, 1990; Gopnik & Slaughter, 1991; Gopnik, Slaughter, & Meltzoff, 1994; Wellman, 1990, 1993).
Page 18
However, just as it is a distortion to think that children don't have an understanding of mind until they pass a falsebelief exam, it is likewise a distortion to hold that infants have the adult conception of the mind as soon as they show a special interest in people. An alternative is a genuinely developmental account of children's understanding of the mind. The view we favor is that infants are given a jump start in understanding people because of certain innate structures, but they gradually come to understand the whole range of psychological flora and fauna including pretenses, images, emotions, perceptions, desires, intentions, and beliefs. Newborns do not have anything like this full understanding of the mind, but they do have privileged ways of understanding other people and human acts. In earlier work, we argued that infant imitation may provide the first groundwork for later understanding of the mind (Gopnik & Meltzoff, 1994; Meltzoff & Gopnik, 1993). Imitation is a behavioral measure indicating that newborns, at some level of processing no matter how primitive, can map actions of other people onto actions of their own body. The findings of early imitation have now been replicated and extended in 25 different studies from 13 independent laboratories, both in this country and crossculturally (for a history and literature review, see Meltzoff & Moore, 1977, 1994, 1997). A comprehensive model of early imitation was offered by Meltzoff and Moore (1997) and dubbed the AIM (active intermodal mapping) model. The central notion is that imitation, even early imitation, is a matchingto target process. The goal or behavioral target is specified visually. Infants' selfproduced movements provide proprioceptive feedback that can be compared to the visually specified target. AIM proposes that when babies imitate, they are linking the visual appearance of other people to their own internal kinesthetic and proprioceptive feelings, connecting the visible bodily actions of others and their own internal states. This type of initial state would provide a jump start for infants' understanding of persons and commonsense psychology because it provides the first and most fundamental building block of the folk psychological framework: "Those entities are like me." Thus, when newborns look at the moving adults, they do not simply see "visual complexity," "highcontrast areas," or mere physical motions, but special acts that are like the acts they can and do perform. Newborns are not alone; they perceive that other entity is "like me."1 1
In using the English word "me," we do not suggest that the infant has the fullfledged adult sense of self. Indeed, we have argued that such a sense of self is a developmental product (Meltzoff & Moore, 1995). Our argument could be rephrased by purging the "me" word and instead saying, "That looks like this feels." Elsewhere we have attempted to describe the initial state in a precise technical manner, using a computational model and avoiding the glosses of everyday English (Meltzoff & Moore, 1997). Interested readers are referred to this work for detailed arguments about early selfother relations.
Page 19
Development in social cognition depends on twoway traffic between self and other, on what might be called "projection" from one's own case to the other and "appropriation" from the other to the self. But this depends on a prior assumption that self and other have anything whatever to do with one another. Newborn imitation provides a demonstration that at some primitive level this link has been made in the normal infant. Without this fundamental connectedness, there would be no reasoning bidirectionally from one's own case to another's because the two cases would not be known to be similar (Meltzoff & Moore, 1995). Regardless of our theory about the initial state, there is, admittedly, a substantial gap in the findings between early infancy and early childhood. We know something about the initial state of newborns from studying imitation and other early phenomena like interactional synchrony and face recognition. We know something about the state of 3yearolds who are on the verge of understanding belief. However, "the dark ages," from about 15 to 36 months, remain something of a mystery. Ask graduate students to test 2yearolds and they will often shudder and scurry out of the lab; the "terribletwos" lead to both subject and graduate student attrition. In the dark ages, the established techniques of infancy (e.g., preferencefornovelty procedures) do not work because the children are too old to sit and passively watch; conversely, tests demanding subtle verbal distinctions do not work ("When I first asked you, before we did X, what did you believe . . ."), because toddlers are too young for such verbal gymnastics. We're left guessing. The absence of empirical findings from this period has contributed to our difficulty in laying out a finegrained developmental theory. It is as if biologists had only seen frogs and tadpoles without the transitions in between. It would be hard to tell a developmental story, and no one would believe it if you did.2 A variety of techniques, however, have recently been developed to test children during the dark ages. One set of techniques uses toddlers' language abilities. Such studies suggest, for example, that 18monthold children understand that words refer to objects and can use an adult's attentional cues (e.g., gaze direction, gestures) to identify the referent of a novel label (e.g., Baldwin, 1993a, 1993b; Baldwin et al., 1996). At a similar age, children also take into account the intentions of the other person in their attempts to determine the referent of a novel word (e.g., Tomasello, 1995; Tomasello 2
We do not mean to imply that there has been no work in the 15 to 36monthold age range. There has been a great deal of work (e.g. Damon, 1998; Kagan, 1981), but not much from the "theoryofmind" viewpoint linking what infants know about persons and what 3 to 5yearolds know about the intentional mental states of persons. Researchers have discovered a good deal about social cognition in infancy and a good deal about social cognition in 3to 5yearolds, but not enough about what happens in between. We expect that future research will shed increasing light on "the dark ages."
Page 20
& Barton, 1994; Tomasello, Strosberg, & Akhtar, 1996). Finally, analyses of naturalistic studies of early conversation have illuminated children's understanding of the mind (e.g., Bartsch & Wellman, 1995). A second newly developed technique, called the "behavioral reenactment procedure" (Meltzoff, 1995), also explores what children know about the mind, but does not rely on language. The behavioral reenactment procedure capitalizes on children's natural tendency to reenact or imitate the behaviors they see, but uses it in a more abstract way. A host of research indicates that children, even young infants, do not imitate by rote; they do not immediately imitate the events they see, but rather their interpretation of them (Meltzoff & Moore, 1995, 1997, 1998). This means that if we arrange a situation correctly, we can use their natural tendency to reenact adult behavior as a "read out" of how they understand the world. Such an approach has been extensively used in the psycholinguistic literature to assess children's linguistic structure. This work shows that children who are asked to imitate a sentence of the adult language tend to respond with a similar sentence, often synonymous with the tobeimitated one, but which conforms to the child's own linguistic rules. The behavioral reenactment procedure uses the imitation of goaldirected acts to examine the psychological structures children use in interpreting human behavior. A third set of techniques capitalizes on children's very early tendency to read meaning into human emotional expressions. This underlies early "social referencing" studies but also has been developed in a more sophisticated way by Repacholi (Repacholi, 1998; Repacholi & Gopnik, 1997). There is evidence that basic emotions such as happiness, sadness, and disgust are associated with particular facial expressions from early infancy and universally across cultures (e.g., Darwin, 1872; Ekman, 1980). Emotions are closely and intricately connected to intentions and desires in our adult understanding of mind. In the everyday adult framework, we assume that getting what we want or acting as we intend to will lead to happy emotions, whereas failure will lead to negative emotions. We also assume that we act in a way that will bring about positive but not negative emotions. Some of the new techniques we discuss in this chapter exploit the early nonverbal ability to read emotional expressions as a way of investigating children's understanding of the mind. Taken together, these recent procedural advances in addressing questions to very young children are starting to reveal some of what children understand in the dark ages. There are two important foci of children's developing understanding of the mind in this period. One is their understanding of perception and attention (e.g., Baldwin, 1993a, 1993b; Gopnik, Slaughter, & Meltzoff, 1994; Gopnik & Wellman, 1994; O'Neill, 1996; Slaughter & Gopnik, 1996; Tomasello, 1995). The other is an understanding
Page 21
of what Searle (1983) called "worldtomind states" such as desire and intention. We focus on the latter in the present chapter. Understanding Directedness In fullfledged adult psychology, an important feature of desires and intentions is that they are "directed at" objects and events. In fact, in adult psychology this is one thing that distinguishes desires and intentions from mere moods, feelings, or emotions and makes them similar to perceptions and beliefs. At the same time, desires, intentions, and emotions, unlike perceptions and beliefs, share what we might call valence. Desires carry with them an implication of certain positive or negative attitudes towards objects and events. Getting what we want is good, and being frustrated is bad. Doing what we intend to do is good, failing to do so is bad. A Conceptual Clarification: Intention and Intentionality What we are calling the "directedness" of these states sometimes is referred to in psychology as the "intentional" character of desire and intention, but this derives from a misunderstanding (or loose adaptation) of a technical philosophical term. Intention in the technical philosophical sense refers to the propositional character of a mental state, not solely the fact that it is directed at objects in the world. This is an important distinction inasmuch as some mental states may be directly or causally related to real objects or events in the world, without being intentional. The classical philosophical example is "seeing" (as opposed to "believing" or ''seeing that"). This mental state is related to real objects and events in the world, not to mental representations of events, and this has important consequences. I can substitute different descriptions of the same event and preserve the truth of the sentence when "see" is used in the nonintentional sense. For example, it is true to say that someone sees the author of Waverley when they see Scott, but it is not true to say that someone, who believes this person is Scott, also believes that he is the author of Waverley. The same holds for worldtomind states such as wanting and intending. These states may simply be directed at objects or events, which means that descriptions can be substituted preserving truth, or they may be genuinely intentional (in the technical, philosophical sense), which means that they cannot. The terminological distinction is important for developmental psychologists because it lets us discriminate between two different ways young children might understand the "aboutness" or "directedness" of mental states. It also should prevent us from assuming that if a child has the minimal idea of the directedness of mental states, they must also have an under
Page 22
standing of the fact that mental states have propositional content, involve representations, and therefore are "intentional." In fact, there is likely a developmental change from children first understanding the nonintentional aspects of mind to later understanding its intentional character. To make things even worse, these philosophical concerns about what it means to be "intentional" have literally nothing to do with the everyday use of "intention," as in the intention to act—it is just a homonym, although it is related to the "intension" of sentences (no wonder philosophers prefer to talk about x's and y's). To assume a deep link between the two intentions, as psychologists sometimes do, is like assuming that the Federal bank must be an historical outcome of the economic importance of rivers. Of course, intentions, in the sense of intending to act, are mental states and can be "intentional" (in the sense of having propositional content), although they need not necessarily be, just as desires, perception, and other mental states may or may not be ''intentional." But then the Federal bank may (for all we know) have something to do with the economic importance of rivers. The similarity of the two words is obviously not an indication of any deeper link between the two concepts: Beliefs are "intentional" mental states par excellence, and there's no homonym at play in this case. Finally, it may be true (we think it is) that the earliest "intentional" concepts of children concern intention, desire, and perception, but if so, this will be an empirical discovery of developmentalists, not a logical truth embodied in the homonyms. Exploring Toddlers' Understanding of the Directedness of Emotional Attitudes Infant imitation and other phenomena of early infancy show that infants can link their own feelings and those of others. However, these early behaviors do not involve objects. The feelings that children understand at first are just that, purely internal feelings. When do infants understand both the valenced and directed character of adults' attitudes toward objects? The literature on social referencing suggests to some that this may be understood as early as 9 months of age. In these studies, mothers reacted to objects and events with particular emotions, and babies seemed to adopt these emotional attitudes. However, a closer look at the experiments in that paradigm suggests that this conclusion may be unwarranted. The fact that infants adopted the mother's attitude does not demonstrate that they understood that this attitude was directed at a particular object. First, typically only one object is presented, so it remains unclear whether infants truly understood that the emotional message was directed toward this and not other objects. Second, the fact that infants adopted the mother's attitude does not necessarily mean that they understood that her
Page 23
attitude was directed at any object whatever. Various forms of emotional contagion and/or simple associative processes (e.g., temporal contiguity, stimulus salience) could be at work in these studies, as argued by advocates of a "lean interpretation" of the standard social referencing work (Baldwin & Moses, 1994). To demonstrate that children really understand directedness, you need to show that they understand that a person can have one attitude toward one object and a different attitude toward another object. At least two objects should be presented, and these should be equally attractive and salient. Some researchers have recently adopted a dualobject display (e.g., Baldwin & Moses, 1994; Mumme, Won, & Fernald, 1994), but this by itself has not been sufficient to eliminate the possibility of all simpler nonreferential processes (see Repacholi, 1998, for a discussion). Repacholi (1998) developed a technique to test infants' understanding of directedness of emotional signals that rules out nonreferential mechanisms. In these studies, 14monthold infants saw the same experimenter produce two different emotional expressions, an expression of disgust and an expression of joy, toward two different objects. The objects were in closed boxes so that the children did not see the objects at the time that they saw the emotions. Instead they saw the experimenter peek into each box and make a different emotional response. The children were then given the two boxes. The results showed that children touched and handled both boxes equally quickly and frequently, indicating that there was no simple emotional contagion at work. The important result was that they opened the box with the "happy" object inside significantly more frequently than they opened the box with the "disgust" object. We conclude that the children had inferred that the adult's attitude was specifically directed at the object inside the containers, even though they had not seen the emotions and the objects at the same time. On the basis of these results, we can also firmly address the temporal contiguity and salience issues that have bedeviled socialreferencing research. How do we know that infants were not simply noting the temporal contiguity between the emotional signal and whichever stimulus they were looking at, at the time the signal was issued? This is ruled out because the only visible objects were the two boxes, not the objects they contained. Moreover, infants should have connected an emotion to whichever box was the focus of their own attention when the expression was displayed, and the results showed that this was not the case (because they handled both boxes equally). Similarly, the paradigm rules out the possibility that simple salience was at work. Both boxes were extremely salient and the experimenter's actions (e.g., picking the boxes up, opening their lids) made them all the more so. Yet infants did not link these salient stimuli
Page 24
to the emotions: They were not loathe to touch and examine either box, even the box containing the disgust object. They just did not want to grab the object that was hidden inside that box. Because the object was hidden, it could hardly have been visually salient when the emotion was originally displayed. We conclude that the interpretation of the social referencing demonstrated in 9 to 13monthold infants is still equivocal. Perhaps a "lean" interpretation of those effects are still in order. However, the work discussed here shows that by 14 months infants genuinely understand that emotional attitudes with particular valences may be directed at particular objects. They understand that the objects to which an adult's emotions are directed may be ones that are neither perceptually salient, nor even perceptually present, contiguously with the emotional expression. This is an important step toward the adult understanding of desire. Early Understanding of Simple Intentions Another important aspect of the adult understanding of the mind is a distinction between the actual actions someone performs and their intention in performing those actions. Wittgenstein (1953) asked, "What is left over if I subtract the fact that my arm goes up from the fact that I raise my arm?" (p. 161). Answer: Intention. This pithy example shows that intention is not wholly reducible to bodily movement. Intentions are mental states and bodily movements are physical events in the world. The two have an intimate relation because intentions underlie and cause bodily movements. If we know a person's intentions, we (often) can predict her actions, and conversely if we see her actions, we can often reason backwards to what her intentions must have been. Moreover, within the adult framework, only certain types of movements are ascribed to intention. Chairs and boulders move, but their rocking and rolling are not seen as intentional. Most prototypically, human acts are the types of movement patterns that are seen as caused by intentions. Just as the youngest infants do not show evidence of understanding the directedness of mental states, they also show little evidence of understanding this distinction between underlying intentions and visible movements, although they do link their own intentional movements to the intentional movements of others (as in body imitation). When do children begin to differentiate bodily movements from the underlying psychological states that cause them, and when do they begin to understand that only certain types of movements and not others are intentional? To address these questions, it is not enough to show that young infants act intentionally themselves. We want to know when they begin to under
Page 25
stand the intentions of others, and most important, when they begin to differentiate surface actions from underlying intentions. There has been some excellent research on this question using verbal tests on young children just beyond "the dark ages," between 3 and 4 years of age (e.g., Astington & Gopnik, 1991b; Moses, 1993; Shultz, 1980; Shultz, Wells, & Sarda, 1980). Our goal was to use the behavioral reenactment technique to pose the question to preverbal children. In one study, we tested whether 18monthold children could read below the literal surface behavior demonstrated in an adult act (Meltzoff, 1995). The study involved showing infants an unsuccessful act. For example, the adult accidentally under or overshot his target, or he tried to perform a behavior but his hand slipped several times. Thus the goalstate was not achieved (Fig. 2.1). To an adult, it was easy to read the actor's intentions, even though he was not able to fulfill them. The experimental question was whether children interpreted this behavior in purely physical terms or whether they too read through the literal body movements to the underlying goal or intention of the act. The measure of how they interpreted the event was what they chose to reenact, in particular whether they chose to produce the intended act despite the fact that it was never present to the senses. In a sense, the "correct answer" was to not copy the literal movement, but the intended act that remained unfulfilled and therefore invisible. Using this behavioral reenactment paradigm, Meltzoff (1995) tested four groups of 18monthold infants. The demonstrationtarget group saw the adult successfully fulfill his intentions and perform a series of target acts on five different objects. The demonstrationintention group saw accidental failures for five different events. With each object, the adult strove
Fig. 2.1. The displays used in the study of toddlers' understanding of simple intentions. The top panel shows the human's acts. The adult tried to pull apart the dumbbell three times. Each time, his hand slipped off the end of the cube (first to one side, then the other, and then the first side again). The children did not copy this surface behavior. When given their turn with the dumbbell, they wrapped their hands around the cubes and firmly pulled it apart. The bottom panel shows the inanimate device. Results showed that children did not try to pull the dumbbell apart after seeing these motions. They interpreted the human acts differently than the similar motions of the inanimate device. (From Meltzoff, 1995.)
Page 26
to reach the goal but did not successfully carry out his intention. The adult's motor actions were realisticlooking attempts to reach the goal, but he did not verbalize or show facial expressions of frustration at his failures. Two control groups were used. The control1 group simply omitted any adult demonstrations. The control2 group saw the adult perform control actions on the objects for the same length of time as in the two demonstration groups, but the adult showed neither the target acts nor the intention to achieve them. The control acts were carefully designed so as to control for the possibility that spatial proximity of the adult's hands to the target, or proximity of two objects with each other, might "suggest" the target behavior (see Meltzoff, 1995, for details). The results were very clear cut. Infants in the control groups did not tend to produce the target acts spontaneously or by chance. However, infants in the two demonstration groups produced more than 75% of the target acts. They reproduced the targets after observing the adult do so, which is straightforward imitation. The important, new finding is that they also produced the targets in the intention condition. Indeed infants were equally likely to produce the target after seeing the adult "trying" but failing as they were when the target was actually achieved. They seemed to interpret the adult's effortful behavior as going beyond the literal surface behavior itself, and as being about something else, about the unseen but inferred goal of the act (see also Tomasello & Barton, 1994, for work using a slightly different approach and yielding compatible results). Several followup studies have now been completed that confirm and expand the original findings. If infants are picking up the underlying goal or intention of the human act from seeing the failed attempt they should be able to achieve the act using a variety of means. This was tested in a recent study (Meltzoff, 1996b). As before, the adult showed the failed attempt with his hands sliding off the ends. Then he handed the infant a gigantic dumbbell that was too big for the infant's hands. The infants did not even make an attempt to grasp the ends of the dumbbell. They did not appear to be trying to mimic the surface behavior. Instead, they used novel ways to struggle to get the gigantic toy apart. They put one end of the dumbbell on the table and used both hands to pull the other end upwards; or they put their hands inside the toy and pushed outwards, and so on. In short, they use different means than had been demonstrated by the experimenter, but used them toward the same end. Of course, the interesting thing is that they had never seen the end. They inferred the end and then used previously unseen means to get there. This eliminates the possibility that infants in the original study had merely tried to imitate the surface behavior of the adult (hands slipping off the cubes) and had pulled the toy apart by mistake. It is consistent with the hypothesis that
Page 27
infants had inferred the goal of the act, differentiating it from the literal surface behavior that was observed. Another experiment pressed this point further. In this study, infants were shown the standard "failed attempt" display, but they were handed a trick toy. The toy had been glued shut before the study began. When infants picked it up and attempted to pull it apart, their hands slipped off the ends of the cubes. This of course matched the surface behavior of the adult. The question was whether this match to the adults' behavior satisfied them. The results showed it did not. Infants were not satisfied when they matched the surface behavior of the adult; this did not terminate their behavior. They repeatedly grabbed the toy and yanked on it in different ways, and appealed to their mothers and the adult. Ninety percent of the infants looked up at an adult immediately after the infant failed to pull the trick toy apart. They did so with an average latency of less than 2 seconds and accompanied by vocalizations while they stared into the adults' faces (Meltzoff, 1996b). Why were they appealing for help? They had matched the adult's surface behavior, but evidently were striving toward another goal, not the behavior itself. (Of course it did not make a lot of sense to seek the adult's help because he had already failed. If a large adult failed, the infant's failure was perhaps inevitable. This subtlety escaped them, however.) We have begun to explore the aspects of the adult's behavioral envelope that carry the information that an action was a failedattempt and not a success. When we see an adult relaunch the act and vary the means, we interpret the adult as "effortfully trying" to accomplish something beyond what he is doing. We tested whether infants were sensitive to this tryandtryagain aspect of intentional action (Meltzoff, 1996a). In this study, four independent groups of 18montholds saw the adult perform either the failedattempt or the successful target act either one or three times. (Recall that infants in the original Meltzoff, 1995, study saw three failed attempts.) Infants who saw one failed attempt performed poorly. Their behavior dropped to chance levels, significantly lower than infants who saw three failed attempts. On the other hand, the infants who saw the adult perform the successful target behavior only once did as well as if they saw it three times. This establishes that it is not an across theboard sensory limitation of some kind. These results suggest that seeing a person relaunch his behavior several times is an important cue to the purposiveness of the act for infants, just as it is for adults (Heider, 1958). When an adult relaunches his behavior several times, using different but related actions, infants infer that there is a common cause unifying this surface behavior. In short, 18montholds use the whole pattern of behavior to indicate whether the adult is aiming to do what they are doing or something else.
Page 28
What Kind of Entities Are Interpreted as Purposive? The results we have discussed so far suggest that normal infants can pick up the simple intentions of human actors. One interesting question is whether this intentional reading of behavior is specific to people or at least most readily ascribed to them. To begin to examine this, Meltzoff (1995) tested how 18montholds responded to a mechanical device that mimicked the same movements as the actor in the failedattempt condition. An inanimate device was constructed that had poles for arms and mechanical pincers for hands. It did not look human but it could move very similarly to the human (Fig. 2.1, bottom panel). For the test, the pincers "grasped" the dumbbell at the two ends just as the human hands did. One mechanical arm was then moved outwards, just as in the human case, and its pincer slipped off the end of the dumbbell just as the human hand did. The movement patterns of machine and man were closely matched from a purely spatiotemporal description of movements in space. The results showed that infants did not attribute a goal or intention to the movements of the inanimate device. Although they were not frightened by the device and looked at it as long as at the human display, they simply did not see the sequence of actions as implying a goal. Infants were no more (or less) likely to pull apart the toy after seeing the failed attempt of the inanimate device than they did in baseline levels when they saw nothing. Another study pursued this point. In this study, we had the inanimate device succeed. The inanimate device held the dumbbell from the two ends and successfully pulled it apart. After witnessing this display, infants were given the dumbbell. The results showed that they too pulled it apart (Meltzoff, 1996b). It appears that infants can pick up certain information from the inanimate device (they pull it apart after seeing the device do so), but they cannot pick up other information (concerning failed attempts). We believe 18montholds interpret the person's actions within a psychological framework that differentiates between the surface behavior of people and a deeper level involving goals and intentions. When they see a person's hands slip off the ends of the dumbbell, they infer what the adult was "trying" to do (which is different from what he did do). When they see the inanimate device slip off the end of the dumbbell, they see it as mechanical slippage and sliding with no implications for purposiveness.3 3
It is quite likely that displays can be constructed that fool infants, analogously to those that fool adults (is my computer intentional?). We do not know the necessary and sufficient conditions for attribution of intention, but under certain circumstances infants may see purposiveness in the actions of pretend humans (e.g., stuffed dolls or animals) or in dynamic displays that may seem to be ambiguous as to animacy (2D spots that leap and move spontaneously, as in Gergely, Nádasdy, Csibra, & Bíró, 1995). This does not contradict our thesis, but underscores the need for research on boundary conditions. The 3D, clearly inanimate object used by Meltzoff (1995) gives a lower boundary (infants fail) and real people give an upper boundary (infants succeed). There is a lot of room in between.
Page 29
Human Acts Versus Mechanical Motions On the basis of these findings, it is useful to introduce a distinction that is discussed later in the chapter. We wish to distinguish between construing the behaviors of others in purely physical versus psychological terms. To help keep this distinction clear, we call the former movements or motions and the latter human acts. The behavior of another person can be described using either physics or psychology. We can say, "The president's hand contacted the cup; the cup fell over," or "The president was trying to pick up the cup." Strict behaviorists (and some lawyers) stick to the former description precisely because they eschew appealing to invisible psychological states. Evidently, infants are not behaviorists (not to mention lawyers). They do not construe the behavior of others simply as, "Hold the dumbbell and then remove one hand quickly," but rather construe it as an effort at pulling. Moreover, the new work shows that infants have a differentiation in the kinds of attributions they make to people versus things. By 18 months of age children have already adopted a fundamental aspect of a folk psychology: Persons are understood within a framework involving goals and intentions. Human acts are seen as dripping with purposiveness and are mapped onto one's own like acts. Using a Person's Emotional Reactions to Understand Their Intentions In the adult framework, it makes sense that the same external event may cause one person to become happy and another sad. This is because emotions do not map directly onto outcomes, but are mediated by the person's desires. Using a variety of verbal tests, it has been shown that young children understand quite a bit about the linkage between desires, actions, and emotional reactions. For example, they know that fulfilled desires lead to happiness and a cessation of searching, whereas unfulfilled ones lead to sadness and a continuance of search for the desired object (Hadwin & Perner, 1991; Harris, 1989; Stein & Levine, 1989; Wellman & Banerjee, 1991; Yuill, 1984). There is clear evidence for this sort of understanding in 3 to 4yearolds and some evidence that children as young as 34 months (Wellman & Woolley, 1990) can correctly predict whether a protagonist in a story will be happy or sad, depending on the match between her desires and the outcome. The behavioral reenactment procedure provides a nonverbal way of beginning to explore children's understanding of the links between desire/intention action emotional reaction. In one study, children ranging from 18 to 36 months old were shown an adult performing an action and the adult's emotional reaction to her action was systematically
Page 30
manipulated (Meltzoff, 1996b). After the action was completed, the adult reacted in one of two ways. For half the children, the adult reacted with happy/satisfied facial expressions and exclaimed, "Yeah! There!" and for the other half she reacted with unhappy/dissatisfied facial expressions and exclaimed, "Uhoh! Oh dear!" The important point is that the adult's physical actions were identical in both cases. What differed was the adult's reaction to the event she caused. The question was whether the children's inference about the adult's desire/intention changed as a function of her emotional reactions. Using the behavioral reenactment procedure, three groups of infants aged 18, 24, and 36 months old were randomly assigned to the see the happy/satisfied and unhappy/dissatisfied reaction. The events in themselves were carefully designed on the basis of pilot studies to be ambiguous in themselves. For example, the adult put a toy unstably on top of a 12inchhigh shelf and the toy toppled off making a banging sound as it hit the table. The usual response after the toy fell was to look at his face to clarify the ambiguous event. Immediately after the object fell, the adult gave one of the reactions, happy/satisfied or sad/dissatisfied. The objects were then given to the child. This proved to be a very motivating task, and children leapt at the opportunity to play with the objects. The results revealed an interesting developmental change. The behavior of the 18 and 24montholds did not systematically vary as a function of the adult's emotional reactions. The 36montholds showed a highly significant and orderly response. They carefully put the toy stably on the shelf (which is not what they saw) in the case that the adult seemed dissatisfied by the outcome. Conversely, they exaggeratedly knocked the toy off the shelf if the adult had shown the happy/satisfied reaction. We conclude that in an ambiguous situation children use the adult's emotional reactions to clarify the meaning of the adult's behavior. On this interpretation, children by 36 months but not 18 to 24montholds can reason backwards from an emotional reaction to what the adult was striving to do. It is interesting that this age estimate fits well with that obtained by verbal methods (Wellman & Woolley, 1990). A modest additional piece of information provided by the behavioral reenactment procedure is that children are not simply presented with a multiplechoice verbal response ("will he be happy or sad"). The children are surrounded with the clutter of realworld activity and have to create for themselves the desired end state. The adult never put the toy stably on the shelf and the children who tuck it firmly up there, far from the edge, are imagining and creating the result of what the adult "had in mind" but never achieved. Another interesting point is that the children were forced to reason backwards from emotional reaction to the unseen desire or intention, not forward, as in many of the verbal story scenarios, from the desire and events to the
Page 31
predicted emotion. The results indicate that at least by 36 months old, children know that the adult may desire/intend to do something different from what they do and that the person's emotional reactions after the event are a clue to the underlying desire/intention of the person. It makes some developmental sense that children can first read the goal of the adult's act (by 18 months) and then later can detect regularities in how they themselves and others emotionally react to the successes and failures of goaldirected actions. Presumably, this is the database that allows them (at 36 months) to give meaning to messy, naturally occurring events such as an actor who reacts either positively or negatively to toppling toys. Understanding Differences between One's Own Desires and Those of Others The work discussed so far has focused on one important aspect of our adult understanding of intention, the ability to "read through" surface actions to determine the intentions of the person who performed them. One important aspect of this ability is that the child seems to go beyond the immediate action itself. Children also seem to take into account other aspects of the situation, for example the actor's attendant emotional reactions, to determine intentions. This research assumes that the child identifies their own intentions with the intentions of adults. In fact, one reason both the behavioral reenactment and social referencing paradigms are so effective is that the children so readily "take on" the intentions, desires, and attitudes of adults, even when those intentions are not their own initially. However, an important aspect of our adult theory of mind is the fact that we can differentiate between our own intentions and desires and the intentions and desires of others. We understand that our desires and intentions may differ from and even be in conflict with, the desires of those around us. In some ways these differences among desires parallel the differences in belief that are tested in falsebelief paradigms. This understanding of differences in desires emerges considerably earlier than the understanding of differences in belief (e.g., Flavell et al., 1990; Gopnik & Slaughter, 1991). As in the case of intention, there is a considerable body of work suggesting that children understand this aspect of desire by the time they are about 3 years old (Astington & Gopnik, 1991a; Bartsch and Wellman, 1995), but the origins of this understanding have been lost in "the dark ages." Repacholi and Gopnik (1997) devised a nonverbal method to explore young children's understanding of differences in desires. The method, like that in the previous study of intention and the earlier study on "di
Page 32
rectedness,'' capitalized on young children's ability to detect emotions. Here, however, we asked a rather different question about emotions. In this work, 14 and 18 monthold infants were presented with a plate of raw broccoli and a plate of goldfish crackers. Infants consistently prefer the crackers. The experimenter indicated her preference for one object or another by tasting it and producing a particular emotional expression (disgust or pleasure). Infants were randomly assigned such that half of them saw the adult apparently like the goldfish crackers (the child's own preference) and half saw her apparently like the broccoli (the child's nonpreferred food). In the crucial test condition, the experimenter then reached her hand out to the infant midway between the foods and asked the infant to give her an (unspecified) food. The results showed that 18monthold infants consistently gave the adult the object for which she had expressed a preference, even when the preference differed from their own desire. They gave her broccoli when she had previously expressed a desire for the broccoli, and crackers when she expressed a desire for crackers. This is a developmental achievement inasmuch as 14montholds did not do this. Instead, they always gave the experimenter crackers, their own preference, regardless of the experimenter's expressed desires. This work suggests that even very young children, 18montholds, may have a nonegocentric understanding of the differences between their own mental states and those of others in some cases. This early understanding of desire, like the understanding of intention, goes beyond the simple cues of action or emotion themselves. It confirms that 18monthold children are not limited to the immediate evidence of the experimenter's perceptually present action, or their own present feelings, in determining the adult's desires. Instead, they take into account an earlier, and superficially quite different, piece of evidence about the experimenter's underlying mental state, namely her emotional expression. Like the earlier understanding of social referencing, it also shows that young children understand the directedness of mental states. They understand that disgust and pleasure were directed at different objects. But this understanding also goes beyond understanding the directedness and valence of mental states and the fact that they underlie, but are not identified with, actions. It shows that 18montholds, although not 14montholds, understand differences between their own desires and those of others. By 18 months, a complexity of folk psychology has dawned on children. They have come to understand that people not only have mental states, just as they do, but these mental states may sometimes not be the same as their own. Other people are like me but do not necessarily have my likes. The children no longer live in a mental Garden of Eden without conflict in which everyone is conceived of as sharing the same desires.
Page 33
Developing Theories of Intention for Ourselves and in Our Children We first summarize some of the developmental changes that have been described in children's understanding of mind between 0 and 3 years of age. Next we sketch three mechanisms of change that may induce these developments. Conceptual Change Between Birth and 3 Years of Age Newborns are not as sophisticated as 18montholds, no less 3yearolds. Newborns can imitate actions themselves. They can link the actions they see to internal feelings. However, they do not initially appreciate that those feelings may be directed toward objects. Imitation of objectdirected acts arises in the second half year of life (Meltzoff, 1988; Meltzoff & Moore, 1998). Similarly, young infants can imitate what the adult actually does, but they can't imitate what the adult intends to do but fails to achieve. One of us has tried in vain to get neonates to imitate intended actions (e.g., an adult straining to produce a tongue protrusion or mouth opening). They simply do not seem to read through the actions. Hence a developmental change from the youngest infants imitating what we do to older ones imitating what we meant to do. Younger infants also initially do not appreciate fully the differences between their own states and those of others. In fact, we have seen that they seem to begin by assuming that the two will be similar. By 18 months, they will have learned about all these characteristics of the mind. So there is a rich initial state but also profound developmental change. The fact that newborns have any way at all of interpreting others as equivalent to the self provides a foundation for the development of the notion of persons that will eventually include desire and intention. As we have seen, the 18monthold's abilities are quite different from the newborn's abilities. At the same time, they are also quite different from young 3yearold abilities. Eighteenmontholds differentiate between intentional and unintentional actions, between their own desires and those of others, and understand that desires and intentions are directed at objects. However, there is no evidence that they conceptualize desires or intentions as mental states that exist in the mind prior to and independent from any action at all, although there is evidence that older children do so. Nor, in fact, is there any evidence that they differentiate between desires, intentions, and emotional attitudes toward objects, although we have been using those adult terms differentially in this chapter. Reconstructing the child's world view in terms of adult language is always difficult. One idea we find helpful is Searle's (1983) notion of
Page 34
"intentioninaction." Searle suggests that even adults often understand action as informed and shaped by desires and intentions, even if they do not think a separate mental state preceded that action. When I sit down in the morning to drink my tea and go over my plans for the coming day, I may be said to have formulated the intention and the desire to drive to the office at 9:45 a.m. When in the course of that drive, I swerve to avoid the construction pothole, I do so intentionally and have the desire to do so, but I could not be said to have had those desires and intentions before I started out that morning or even right before I swerved. When, in the course of the drive, my mind is so full of the intention and desire to make a good argument at the 10:00 a.m. meeting that I actually drive into the next construction pothole, the damage to my axle is neither intentional nor desired. We suggest that the 18monthold's conception of desire/intention is akin to my conception of the intention involved in swerving to avoid the pothole. It is not identified with or reducible to any bodily movement in particular, but it is assumed to accompany actions. Mechanisms of Development We are suggesting then that there are important developmental changes in the child's conception of the mind, in particular in their understanding of desire and intention, between birth and 3 years (for a more complete account see Gopnik & Meltzoff, 1997). What mechanisms are responsible for those changes? We suggest that three mechanisms may be particularly important: imitation and interpretation of human acts, spontaneous experimentation, and integration of the evidence. Imitation and Interpretation of Human Acts as "Like Me" We believe that infants start off with a wellstocked toolkit for developing a folk psychology. We can be precise about how the initial state enables later development. In our view, the "human act" may be the most elementary parsing of the world for social cognition. Human acts are especially relevant to infants because they look like the infant feels himself to be and because they are things infants can intend. When a human act is shown to a newborn baby, it may provide a primordial "Aha" experience: "Something interpretable! That (seen) event is like this (felt) event." It is not simply the attractive eyes and lips of the adults that are special for infants but the way the body moves and its relation to the self. The fact that infants can recreate the act allows them to imbue it with special meaning. Thus, we propose that the initial parsing infants impose on the world is not any one of the "usual suspects" found in textbooks and commonly discussed at the biennial meetings of the Society for Research on Child Development. It is not the GelmanSpelke distinction between "animate
Page 35
versus inanimate" (because armadillos are only of passing interest to infants). It is not the PremackLeslieMandler distinction between "selfpropelled versus moved byaseen force" (because swinging clock pendulums and falling leaves are not viewed by infants as special). It is not even the philosopher's distinction between "people (as adults know them) versus things." We believe that the primordial distinction may be something closer to "human acts versus other events'' (see Meltzoff & Moore, 1995, 1997, for more detailed arguments along this line). Infants' construing certain movements in the environment in terms of human acts that can be imitated has cascading developmental effects: (a) The world of material objects is then divisible into those that perform human acts (people) and those that do not (things); and (b) having made the division in the external world, new meanings are possible. I can imitate others, and those entities out there can generatively imitate me. Persons are special entities, the only entities in the world with whom I can share behavioral states. Furthermore, the recognition that others share your states lays the foundation for making further progress toward ascribing psychological properties to these people. This may come about in part because the infant detects regularities in their own behaviors and feelings. When they are acting in a "tryandtryagain" manner they are striving to reach a goal that is not achieved. When they feel happy, they regularly produce a smiling face. There are regularities between the way they act and the way they feel. This would have no interpersonal significance if infants could not recognize that others are acting just like them. But as we have discovered from studies of imitation, infants can readily act like others and recognize when others are acting like them. This recognition of sharing behavioral states is crucial because it allows a foothold for infants attributing like mental states to others. We envision a threestep developmental sequence: (a) When I perform that bodily act I have such and such a phenomenal experience, (b) I recognize that others perform the same type of bodily acts as me, (c) the other is sharing my behavioral state; ergo, perhaps the other is having the same phenomenal experience. (For further analysis of this developmental sequence, see Meltzoff, 1990; Meltzoff & Moore, 1995, 1997.) On this view, imitation and the crossmodal representation of human acts provide a kick start for getting folk psychological thinking off the ground. Without it, people would not be seen as psychological entities, 'just like me." The "likemeness" of others is the essential foundation for all later social cognition—from the attribution of mental states, to empathy, to moral judgments. Experimentation We have suggested elsewhere that young children use psychological devices that bear an interesting similarity to the cognitive devices that are involved in theory change in science (Gopnik, 1996, 1998;
Page 36
Gopnik & Meltzoff, 1997). One such device is active experimentation. We have proposed that normal infants have an early, and perhaps innate, drive to actively experiment with the world in a way that will increase their understanding of it and that this experimentation plays an important role in development. There is evidence for a period of experimentation accompanying the new understanding of desire and intention that emerges in "the dark ages." We can demonstrate this discovery in the laboratory, but it is also dramatically apparent in ordinary life. Parents know it as the "terrible twos." (The dark ages, in development as in history, are dark in both senses, in mystery and in retchedness.) What makes the terrible twos so terrible is not that the babies do things you do not want them to do, but that they do things because you do not want them to. Twoyearolds are deliberately perverse, what the British call bloodyminded. The 2yearold does not even look at the forbidden computer keyboard as you type your grant proposal. Instead his hand goes out as he looks, steadily, gravely, and with great deliberation, at you. Why do they torture us by seeking to play with the very things in the world we desire that they do not touch—the computer, the lamp cord, the lipstick, the power tools? This perverse behavior may turn out to be quite rational. Consider that 2yearolds are only just in the course of discovering that people may have different desires. The broccoli experiment shows that children first start to realize that there are differences between their own desires and those of others when they are about 18 months old. The terrible twos seem to involve a systematic exploration of that idea, almost a kind of experimental research program (see Gopnik & Meltzoff, 1997; Repacholi &.Gopnik, 1997, for further arguments). Toddlers are systematically testing the dimensions on which their desires and the desires of others may be in conflict. The grave look is directed at you because you and your reactions, rather than the forbidden power tools, are the really interesting things. If the child is a budding psychologist, parents are the laboratory rats. Moreover, the experimentation is striking because it actually conflicts with the child's apparent interest in domestic peace. The Fall has come: The young child now understands that their own desires and those of others are not only not the same but that they often conflict. They are forced from the mental Garden of Eden. Integration of Evidence A further common factor in both conceptual changes in childhood and theory change in science is the importance of relevant evidence. Children have extensive evidence about the nature of human action, intention, and desire. There are two sets of experimental findings suggesting that evidence about desires and intentions may induce
Page 37
developmental change and propel toddlers toward a fuller understanding of mind. First, there is a consistent finding in the literature that younger siblings do better on theoryofmind tasks than only or older children (Jenkins & Astington, 1996; Perner, Ruffman, & Leekam, 1994). The most likely explanation for this effect is that siblings provide children with rich evidence about the mind and particularly about differences in minds. Remember that much of what children learn involves the differences between their own minds and the minds of others. They largely take the similarities for granted; in fact, the assumption that we are like other people seems to be part of a basic foundation for understanding mind found in infancy. Parents, and perhaps especially some mothers, tend to minimize the distance between their own mental states and those of the babies. They look for commonality and understanding instead of difference, and their lessons are largely lessons about congruence. Siblings may provide a necessary counterweight. They are much more likely to emphasize differences between what they want and the baby wants, or to witheringly contrast their highly superior 4yearold knowledgeand the baby's pitiful 2yearold ignorance. Other data showing a positive correlation between early parentchild talk about feelings and later performance on theoryofmind tasks (Dunn, Brown, & Beardsall, 1991; Dunn, Brown, Slomkowski, Tesla, & Youngblade, 1991) might also be interpreted from this viewpoint, assuming that the frequencyofconversations measures capture increased talk about conflicts in desires, not solely increased congruence talk, a reasonable assumption given the functions of everyday conversations (e.g., Bruner, 1990). The second set of data comes from work in which we explicitly tried to induce changes in children's understanding of the mind by giving them evidence (Slaughter & Gopnik, 1996). The focus in this study was on developments in falsebelief, appearancereality, and source understanding between 3 and 4 years of age. The theory guiding the research was similar to that of this chapter—that an understanding of mental states such as beliefs emerge from a prior understanding of the mind that includes concepts such as desire, intention, perception, and so on. One prediction from this viewpoint is that giving children experience with understanding these earlier states should induce them to develop more quickly and to acquire an understanding of belieflike mental states even though the latter were not part of the training. Such an acceleration study was conducted and the results confirmed that providing children with evidence relevant to desire and perception significantly increased their understanding of belieflike mental states, including passing the falsebelief exam. This acceleration study strongly suggests that experience with reasoning about desires and perceptions are developmental precursors to understanding of belief.
Page 38
Conclusions The challenge now is to articulate a theory about early development that takes seriously the richness of the initial state of infants' understanding of persons, as revealed in modern infancy research, and still embraces developmental change (Gopnik & Meltzoff, 1994, 1997; Meltzoff & Moore, 1997, 1998). We have argued for what we call a "startingstate nativism" that doesn't reduce to Fodor's (1987) "modularity or final state nativism" in which the outcomes are fixed to begin with and simply mature with age. On the contrary, we propose that development proceeds by a constant process of revision, like the process of theory change in science. Infants never face the empiricist dilemma of inducing the existence of the mind from the raw data of behavior. On the other hand, they also are not trapped by the constraints of a single, biologically fixed construal of other people. The analogy to science enables us to suggest that the sophisticated mental life of the 4yearold could emerge without being preprogrammed in the mind of the newborn. The folk psychological framework of Western adults is neither innate nor maturationally determined; it is fashioned by the child largely to account for his experiences with other persons. Children, like the adults who study them, start off with certain powerful assumptions, they experiment, and the theory they construct is deeply influenced by the evidence they receive. Our understanding of children and their understanding of us is not fixed by nature but cobbled together as we interact with each other. References Astington, J. W., & Gopnik, A. (1991a). Developing understanding of desire and intention. In A. Whiten (Ed.), Natural theories of mind: Evolution, development and simulation of everyday mindreading (pp. 3950). Oxford: Basil Blackwell. Astington, J. W., & Gopnik, A. (1991b). Theoretical explanations of children's understanding of the mind. British Journal of Developmental Psychology, 9, 731. Astington, J. W., Harris, P. L., & Olson, D. R. (1988). Developing theories of mind. New York: Cambridge University Press. Baldwin, D. A. (1993a). Early referential understanding: Infants' ability to recognize referential acts for what they are. Developmental Psychology, 29, 832843. Baldwin, D. A. (1993b). Infants' ability to consult the speaker for clues to word reference. Journal of Child Language, 20, 395418. Baldwin, D. A., Markman, E. M., Bill, B., Desjardins, R. N., Irwin, J. M., & Tidball, G. (1996). Infants' reliance on a social criterion for establishing wordobject relations. Child Development, 67, 31353153. Baldwin, D. A., & Moses, L. J. (1994). Early understanding of referential intent and attentional focus: Evidence from language and emotion. In C. Lewis & P. Mitchell (Eds.), Children \ early understanding of mind: Origins and development (pp. 133156). Hillsdale, NJ: Lawrence Erlbaum Associates.
Page 39
Bartsch, K., & Wellman, H. M. (1995). Children talk about the mind New York: Oxford University Press. Bruner, J. S. (1990). Acts of meaning. Cambridge, MA: Harvard University Press. Damon, W. (1998). Handbook of child psychology (5th ed.). New York: Wiley. Darwin, C. (1872). The expression of the emotions in man and animals. Chicago: University of Chicago Press. Dunn, J., Brown, J., & Beardsall, L. (1991). Family talk about feeling states and children's later understanding of others' emotions. Developmental Psychology, 27, 448455. Dunn, J., Brown, J., Slomkowski, C., Tesla, C., & Youngblade, L. (1991). Young children's understanding of other people's feelings and beliefs: Individual differences and their antecedents. Child Development, 62, 13521366. Ekman, P. (1980). The face of man: Expressions of universal emotions in a New Guinea village. New York: Garland STMP Press. Flavell, J. H., Flavell, E. R., Green, F. L., & Moses, L.J. (1990).Young children's understanding of fact beliefs versus value beliefs. Child Development, 61, 915 928. Flavell, J. H., & Miller, P. H. (1998). Social cognition. In W. Damon (Series Ed.) & D. Kuhn & R. Siegler (Eds.), Handbook of child psychology: Vol. 2. Cognition, perception, and language (pp. 851898). New York: Wiley. Fodor, J. A. (1987). Psychosemantics: The problem of meaning in the philosophy of mind. Cambridge, MA: MIT Press. Gergely, G., Nádasdy, Z., Csibra, G., & Biró, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165193. Gopnik, A. (1996). The scientist as child. Philosophy of Science, 63, 485514. Gopnik, A. (1998). Explanation as orgasm. Minds & Machines, 8, 101118. Gopnik, A., & Meltzoff, A. N. (1994). Minds, bodies, and persons: Young children's understanding of the self and others as reflected in imitation and theory of mind research. In S. T. Parker, R. W. Mitchell, & M. L. Boccia (Eds.), Selfawareness in animals and human: Developmental perspectives (pp. 166186). New York: Cambridge University Press. Gopnik, A., & Meltzoff, A. N. (1997). Words, thoughts, and theories. Cambridge, MA: MIT Press. Gopnik, A., & Slaughter, V. (1991). Young children's understanding of changes in their mental states. Child Development, 62, 98110. Gopnik, A., Slaughter, V., & Meltzoff, A. N. (1994). Changing your views: How understanding visual perception can lead to a new theory of the mind. In C. Lewis & P. Mitchell (Eds.), Children's early understanding of mind: Origins and development (pp. 157181). Hillsdale, NJ: Lawrence Erlbaum Associates. Gopnik, A., & Wellman, H. M. (1994). The theory theory. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture (pp. 257293). New York: Cambridge University Press. Hadwin, J., & Perner, J. (1991). Pleased and surprised: Children's cognitive theory of emotion. British Journal of Developmental Psychology, 9, 215234. Harris, P. L. (1989). Children and emotion: The development of psychological understanding. Oxford: Basil Blackwell. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley. Jenkins, J. M., & Astington, J. W. (1996). Cognitive factors and family structure associated with theory of mind development in young children. Developmental Psychology, 32, 7078. Kagan, J. (1981). The second year. Cambridge, MA: Harvard University Press. Meltzoff, A. N. (1988). Infant imitation and memory: Ninemontholds in immediate and deferred tests. Child Development, 59, 217225. Meltzoff, A. N. (1990). Foundations for developing a concept of self: The role of imitation in relating self to other and the value of social mirroring, social modeling, and self practice in infancy. In D. Cicchetti & M. Beeghly (Eds.), The self in transition: Infancy to childhood (pp. 139164). Chicago: University of Chicago Press.
Page 40
Meltzoff, A. N. (1995). Understanding the intentions of others: Reenactment of intended acts by 18monthold children. Developmental Psychology, 31, 838850. Meltzoff, A. N. (1996a, April). The role of imitation in developing a theory of mind. In P. D. Zelazo (Chair), Intentionality: An interdisciplinary approach. Symposium conducted at the meeting of the International Conference on Infant Studies, Providence, RI. Meltzoff, A. N. (1996b, August). Understanding intentions in infancy. In A. Leslie (Chair), Children's theory of mind. Symposium conducted at the XXVI International Congress of Psychology, Montreal, Canada. Meltzoff, A. N., & Gopnik, A. (1993). The role of imitation in understanding persons and developing a theory of mind. In S. BaronCohen, H. TagerFlusberg, & D.J. Cohen (Eds.), Understanding other minds: Perspectives from autism (pp. 335366). New York: Oxford University Press. Meltzoff, A. N., & Moore, M. K. (1977). Imitation of facial and manual gestures by human neonates. Science, 198, 7578. Meltzoff, A. N., & Moore, M. K. (1994). Imitation, memory, and the representation of persons. Infant Behavior and Development, 17, 8399. Meltzoff, A. N., & Moore, M. K. (1995). Infants' understanding of people and things: From body imitation to folk psychology. In J. Bermúdez, A. J. Marcel, & N. Eilan (Eds.), The body and the self (pp. 4369). Cambridge, MA: MIT Press. Meltzoff, A. N., & Moore, M. K. (1997). Explaining facial imitation: A theoretical model. Early Development and Parenting, 6, 179192. Meltzoff, A. N., & Moore, M. K. (1998). Object representation, identity, and the paradox of early permanence: Steps toward a new framework. Infant Behavior and Development, 21, 201235. Moses, L. J. (1993). Young children's understanding of belief constraints on intention. Cognitive Development, 8, 125. Mumme, D. L., Won, D., & Fernald, A. (1994, June). Do one year old infants show referent specific responding to emotional signals? Poster presented at the meeting of the International Conference on Infant Studies, Paris, France. O'Neill, D. K. (1996). Twoyearold children's sensitivity to a parent's knowledge state when making requests. Child Development, 67, 659677. Perner, J. (1991). Understanding the representational mind. Cambridge, MA: MIT Press. Perner, J., Ruffman, T., & Leekam, S. R. (1994). Theory of mind is contagious: You catch it from your sibs. Child Development, 65, 12281238. Repacholi, B. M. (1998). Infants' use of attentional cues to identify the referent of another person's emotional expression. Developmental Psychology, 34, 1017 1025. Repacholi, B. M., & Gopnik, A. (1997). Early reasoning about desires: Evidence from 14 and 18month olds. Developmental Psychology, 33, 1221. Searle, J. R. (1983). Intentionality: An essay in the philosophy of mind New York: Cambridge University Press. Shultz, T. R. (1980). Development of the concept of intention. In W. A. Collins (Ed.), The Minnesota Symposium on Child Psychology (Vol. 13, pp. 131164). Hillsdale, NJ: Lawrence Erlbaum Associates. Shultz, T. R., Wells, D., & Sarda, M. (1980). Development of the ability to distinguish intended actions from mistakes, reflexes, and passive movements. British Journal of Social and Clinical Psychology, 19, 301310. Slaughter, V., & Gopnik, A. (1996). Conceptual coherence in the child's theory of mind: Training children to understand belief. Child Development, 67, 29672988. Stein, N. L., & Levine, L. J. (1989). The causal organization of emotional knowledge: A developmental study. Cognition and Emotion, 3, 343378. Taylor, M. (1996). A theory of mind perspective on social cognitive development. In R. Gelman & T. Au (Eds.), Handbook of perception and cognition: Vol. 13. Perceptual and cognitive development (pp. 283329). New York: Academic Press.
Page 41
Tomasello, M. (1995).Joint attention as social cognition. In C. Moore & P. . Dunham (Eds.), Joint attention: Its origins and role in development (pp. 103130). Hillsdale, NJ: Lawrence Erlbaum Associates. Tomasello, M., & Barton, M. E. (1994). Learning words in nonostensive contexts. Developmental Psychology, 30, 639650. Tomasello, M., Strosberg, R., & Akhtar, N. (1996). Eighteenmonthold children learn words in nonostensive contexts. Journal of Child Language, 23, 157176. Wellman, H. M. (1990). The child's theory of mind. Cambridge, MA: MIT Press. Wellman, H. M. (1993). Early understanding of mind: The normal case. In S. BaronCohen, H. TagerFlusberg, & D.J. Cohen (Eds.), Understanding other minds: Perspectives from autism (pp. 1039). New York: Oxford University Press. Wellman, H. M., & Banerjee, M. (1991). Mind and emotion: Children's understanding of the emotional consequences of beliefs and desires. British Journal of Developmental Psychology, 9, 191214. Wellman, H. M., & Woolley, J. D. (1990). From simple desires to ordinary beliefs: The early development of everyday psychology. Cognition, 35, 245275. Wittgenstein, L. (1953). Philosophical investigations (G. E. M. Anscombe, Trans.). New York: Macmillan. Yuill, N. (1984). Young children's coordination of motive and outcome in judgments of satisfaction and morality. British Journal of Developmental Psychology, 2, 7381.
Page 43
Chapter 3— Intentional Relations and Triadic Interactions Chris Moore Dalhousie University The second half of infancy, from about 9 months to about 18 months of age, sees the advent of a particularly important set of social skills. During this period, infants start to engage with other persons in various interactions that involve a third object or event. An essential component of these "triadic" (Bakeman & Adamson, 1984) interactions is that, for the first time, infants become able to make use of another person's intentional action to find out about the world. In an essential way, these interactions are developmentally homologous with all later cultural activities involving a shared focus of attention, including, for example, the writing and reading of this volume. It is this ability to engage in triadic interactions that defines us as human because human knowledge is knowledge that is shared with others (e.g., Tomasello, Kruger, & Ratner, 1993). Even the earliest triadic interactions of late infancy clearly involve a sensitivity to the intentions of others. However, the nature of this sensitivity is still at issue. This chapter is about what the infant of about 12 months knows of the intentions of others. I do not say much more about the development of intentional action itself. As a number of authors have noted (e.g., Frye, 1981, 1991; Russell, 1996; Tomasello, 1995), it is probably no coincidence that infants show coordinated meansends sequences of action at about the time they start to respond appropriately to others' objectoriented action. In fact, part of my story is that the infant's own intentional orientations to objects and states of affairs are a necessary component of the first understanding of intention. However, my main
Page 44
goal is to elucidate how the infant at this age understands others' intentional action. To be clear from the outset, I should say something about what I mean by intentional. I am concerned with intentions in the broader sense of intentionality, and not limited to purposeful behavior. In this sense, intentional covers all psychological orientations to an object or state of affairs. Indeed, of special interest are those psychological orientations for which the object is real but displaced in space, of which the paradigm case in my opinion is looking at something. We have argued in recent years (e.g., Moore, in press; Moore & Corkum, 1994) that infants of about 12 months do not understand that others can look at (or attend to) things. Our claim has been that 12montholds interpret others primarily in terms of their actions and not in terms of their intentions (see also Povinelli, chap. 11, this volume). After giving my talk at the Toronto conference on which this book is based, I was accused of asserting that infants are behaviorists. That was not the first time the charge had been laid, nor has it been the last. I think this accusation occurs frequently because in many people's minds, either one allows that infants (or anyone) interpret action as intentional and mediated by mental states or one disallows the involvement of such mental properties in infants' interpretations and thereby consigns infants to being limited to behavioral interpretation. Because I have taken a skeptical stance in the evaluation of infants' social understanding, my account tends to get assimilated to the latter view. In fact, my position is that both viewing infants as having an understanding of others' intentions and viewing infants as little behaviorists are problematic and equally so in that they impute an individualistic form of interpretation to the infant. In contrast, my view is that at 12 months the infant's understanding of intention is grounded in interactive contexts, not in individuals. In the second half of the chapter, I consider some recent work from my laboratory using the novelty preference technique to examine infants' understanding of others. Although by no means conclusive, this work is, I believe, consistent with the theoretical account developed in the first part of the chapter. Intentional Relations In earlier work (Barresi & Moore, 1996), Barresi and I argued that the development of the understanding of intentionality is best seen as a progressive coordination of what we call intentional relations. The latter term is meant to denote the idea that all intentional activity is activity by an agent in relation to something. The "something" can be a real object, directly perceived, or it can be a mental object, which includes propositions.
Page 45
Thus, the idea of intentional relations is supposed to capture not only the most complex propositional attitudes that are usually the domain of philosophers but also the simplest intentional acts. Significantly, our goal was to unite the philosophical notion of the mental as intentional with the commonsense notion of purposeful action. Although there is clearly a difference between mental activity and behavioral activity, casting both as forms of intentional relation recognizes their common roots. As adults, we have an individualistic conception of intentional relations. Our commonsense or folk psychology is individualistic in the sense that agents, commonly persons, are taken to be individual centers of intentional activity capable of taking up a range of intentional relations with real and represented objects. Importantly, one of those agents is the self. The fact that the self is recognized to be an individual agent like others around us is selfevident and yet arguably one of the most significant features of commonsense psychology. When it comes to agenthood, self and other are understood to be essentially equivalent. We understand that other people have conscious experience of the real and imaginative worlds in the same way that we do and at the same time, we understand that we are objective agentive entities of the same kind as the others we see around us. In our earlier work, Barresi and I (Barresi & Moore, 1996; Moore & Barresi, 1993) used the understanding of selfother equivalence as a starting point for our theoretical investigation into the origins of intentional understanding. Understanding this equivalence, we argued, must present a problem for a naive observer of intentional activity because the information available to an observer about their own intentional activity is qualitatively very different from the information available to that observer about anyone else's intentional activity. To elaborate, we argued that information that is available to an observer about another agent's intentional activity emphasizes the agent's action. Correspondingly, the object is less salient, in particular when the intentional activity is in relation either to a mental object or to a real object that is displaced in space. The observer cannot know directly the object of another's intentional relation (the intentional object) and consequently it will have to be inferred in some way. In contrast, the information that is immediately available to an observer about his or her own intentional activity is primarily focused on the intentional object. The agent (self) and any action involved are less salient. In this way, thirdperson information and firstperson information tend to emphasize different components of the intentional relation. Thirdperson information emphasizes the agent and his or her activity. Firstperson information tends to emphasize the phenomenal characteristics of the object and one's attitude towards it. The epistemic problem for a naive observer of intentional activity is how to recognize that both first and thirdperson information about that
Page 46
activity correspond to the same kind of thing, namely an intentional relation between an agent and an object. We (Barresi & Moore, 1996) argued that there must be some way for the young infant to construct the equivalence of intentional relations of self and other from the available input and suggested that interactive contexts in which intentional relations between self and other are shared would provide the right kind of experience from which to construct such an understanding. Gopnik and Meltzoff (1994) also pointed to selfother equivalence as an epistemic problem for understanding intentionality, although their proposed solution to its development is somewhat different from ours (see Moore, 1996). Our account indicates that there is a period in development during which infants can participate in shared intentional relations without being able to attribute an intentional relation to an individual agent, either self or other. The suggestion is that this period comes in the second half of infancy from about 9 months of age. As noted earlier, it is at about this age that infants start to show a variety of behaviors that allow them to interact with another person in relation to some third object. I adopt the term triadic (cf. Bakeman & Adamson, 1984) to refer to these interactions because it captures in a relatively theoretically free manner the important issue— that there are now three things involved: infant, adult, and third object. Commonly cited examples of triadic interactions are joint visual attention or gaze following, social referencing, and prelinguistic communicative acts such as pointing (see Moore & Corkum, 1994). Although some have interpreted these phenomena as revealing an early form of theory of mind (e.g., Bretherton, 1991; Wellman, 1993), I think it is fair to say that such an interpretation is not warranted on the basis of the phenomena alone. Elsewhere (Moore & Corkum, 1994), we have provided a skeptical account of infants' joint visual attention, whereby infants will turn to look in the same direction as an interactive partner. We argued that such gaze following does not require the attribution to young infants of an understanding of attention. Instead, infants may understand that others' gaze behavior is predictive of where interesting sights will occur. The empirical literature on the topic is fully consistent with such a view (see Moore, in press, for a review). Baldwin and Moses (1996) similarly considered the phenomenon of social referencing, or checking of a caregiver's emotional expression in times of uncertainty, and concluded that its origins may lie in the attachment system and not in intentional understanding (see also Povinelli, chap. 11, this volume). If infants at the end of the first year do not understand others in terms of intentional relations, then how do they represent others? We have argued that, at this age, infants process others' activity in terms of thirdperson information, or information about the spatiotemporal properties of behavior. They are able to engage in triadic interactions because the third
Page 47
person information provided by the other's intentional activity directed at objects has become linked in familiar and arousing interactive structures with the infant's own intentional objectoriented activity. The thirdperson information provided by the observation of the other's activity is linked to the corresponding firstperson information provided by the infant's coordinated response. To take again the example of gaze following, infants turn in the same direction as someone else, not because they understand that the other is looking at something, but because they are able to participate in an interactive structure whereby the observation of another's head turn leads to an expectation that an interesting sight will appear to the side, and to a consequent head turn to look to the side. This account of gaze following fits the data on gaze following. For example, Butterworth and others (e.g., Butterworth & Cochran, 1980; Morissette, Ricard, & Gouin Decarie, 1995) showed that when infants first start to follow gaze to objects outside of the immediate visual field, they do not seek out the target of the other's gaze but only turn as far as necessary to see something of interest. At about the same time as they start to follow gaze to objects outside the visual field, infants also start to direct others' attention towards novel or interesting objects using gestures such as pointing (see Carpenter, Nagell, & Tomasello, 1998). Again, however, this behavior can be understood in terms of the thirdperson information provided by the activity of the other being linked within familiar and arousing interactive structures with the firstperson information provided by the infant's own intentional orientation to the novel or interesting event. Here, the infant's own observation of something of interest leads him or her to perform a gesture with the expectation that the adult will then provide some interesting feedback (Moore, 1998; Moore & Corkum, 1994). In such interactive episodes, the infant is presented with corresponding first and thirdperson information about the intentional relation, and it is the combination of these two forms of information that provides the first form of intentional understanding in that an agent is thereby connected to an intentional object. The critical point for present purposes, however, is that the intentional relation is not understood to be a property of either the self or the interactive partner. Why This Is Not a Behavioral Account of Infant Social Understanding It is important to point out here that the claim that infants process others' activity in terms of thirdperson information does not commit one to a behaviorist view of infant social understanding. As I noted earlier, it is sometimes thought that either infants understand the intentional nature
Page 48
of others' activity or they are little behaviorists. If infants only process information about others in thirdperson terms, then is that not a behavioral view? To see why not, consider what it means to have a behavioral understanding of human activity. To construe human action behaviorally means that the observer understands there to be a relation between the action of the other person and the objective state of affairs. This relation is deemed to be a causal one, with the objective state of affairs causing the person's behavior, and one that is unmediated by that person's mental states. Important, the objective state of affairs is understood to be affecting the other person independently of its effects on the observer. Thus, behavioral understanding resembles adult intentional understanding in that it is individualistic. It differs from intentional understanding in that it eschews mentalistic variables. Our claim is that infants do not understand that the other is in a relation, intentional or otherwise, to the objective state of affairs. The other's activity is not seen to be related to an object in either a behavioral or an intentional manner. The activity of another agent is linked to an object or state of affairs only through the connection with the infant's own firstperson intentional orientation to that object or state of affairs. Thus, the other is not understood to have intentional orientations to objects independently of the infant's own orientations to those objects. Equally, whereas the infant has intentional orientations to objects, the self is not understood to be an intentional agent. It is the thirdperson information provided by the observation of the other's activity combined with the firstperson information provided from the infant's own intentional orientation to the object or state of affairs that constitutes the first form of understanding of intentional relations. One might say that the 12 monthold recognizes intentional relations but only to the extent that she can share such a relation with an interactive partner. For the 1yearold, intentional relations exist in the interaction and are not a property of, or descriptive of, individuals. This, then, is a constructivist developmental story in that the interactive structures that contain the essential components for intentional understanding form the basis on which more mature, individualistic forms of understanding develop. Our claim has been that the individualistic form of understanding depends on the capacity for imagination or secondorder representation, which develops around the middle of the second year. Only when the child can imagine the component of the intentional relation that is not given in immediate experience (i.e., firstperson component for others and thirdperson component for self) will the notion of intentional relations that are descriptive of persons be possible. Imagination means that when the child observes another person engaged in intentional activity, she can imagine the corresponding firstperson information com
Page 49
ponent and attribute it to the other. At the same time, when the child herself acts intentionally, she can imagine herself from a thirdperson perspective. In this way a uniform understanding of self and other becomes possible because the child can represent both as agents acting in relation to intentional objects. Notice that although this account has some similarities to the simulation view of social understanding (e.g., Harris, 1992), it differs in the sense that not only is self used as a model for others, but also others are used as a model for self. Empirical Approaches So, one part of our claim is that 1yearolds represent others' activity in terms of thirdperson information, not in terms of intentional relations, but how can one examine infants' understanding of others' intentional activity? Some years ago, Corkum and I (Moore & Corkum, 1994) proposed that the novelty preference paradigm could be used to shed some light on these issues. We were particularly intrigued by the use of the novelty preference method by Cohen and Oakes (1993; Oakes & Cohen, 1990) to study infants' representation of causal relations. Like intentional relations, causal relations involve a relation between an agent and an object, albeit one that is based on the physical rather than psychological properties of the agent. In an ingenious series of experiments, Cohen and Oakes showed that 10 monthold infants are sensitive to the characteristics of the object serving as the agent and the type of action, but not to the characteristics of the object serving as the recipient of the causal event. Our idea (see also Woodward, 1995) was that if 12montholds understand others mainly in terms of thirdperson information, then, when observing an agent engaged in an intentional relation with an object, they should pay more attention to the agent and his or her activity and less attention to the overall relation including the object, especially for intentional activity operating at a distance. Take the example of someone pointing at an object. Figure 3.1a represents the case of a person pointing at one of two objects, indicating attention or interest in that object. From a purely thirdperson perspective, the relevant information comes from the spatial properties of the agent's action, namely the direction of the pointing gesture. Imagine showing an infant this scene repeatedly until habituation occurs and then substituting either of Fig. 3.1b or Fig. 3.1c in which the position of the two objects has been switched. Both of these figures show the same agent pointing at one of the two objects. In Fig. 3.1b, the agent continues to point to the same location but now it is towards the other object. In Fig. 3.1c, the agent continues to point to the same object as before, thereby changing the spatial properties of the pointing gesture. According to the logic of novelty
Page 50
Fig. 3.1. Schematic representation of habituation and dishabituation pointing events.
preference, if the infant processed the original scene in terms of the intentional relation, ''Agent is attending to Object A," then there should be greater recovery of looking to Fig. 3.1b. In contrast, if the infant processed the original scene in terms of the agent's activity alone, then there should be greater recovery of looking to Fig. 3.1c. We have now conducted a series of three experiments using this logic (see Table 3.1) and although the results did not come out exactly as anticipated, I believe they do shed some light on the issues discussed. TABLE 3.1 Overview of the Three Experiments Using Novelty Preference Paradigm to Examine Infants' Representation of Intentional Action
Experiment
Format
Habituation Stimulus
Dishabituation Stimuli
1
Live
Adult 1 points at Toy A
Adult 1 points at Toy A Adult 1 points at Toy B
2
Live
Adult 1 points at Toy A
Adult 2 points at Toy A Adult 2 points at Toy B
3
Video
Adult 1 points at Toy A
Adult 1 points at Toy A Adult 1 points at Toy B
Page 51
General Method In all three experiments, participants ranged in age from 12 to 14 months. Forty infants participated in each of the first two experiments with average ages of 13;1 (i.e., 13 months, 1 day) and 13;6, respectively. Thirtysix infants of average age 13;7 participated in Experiment 3. At each age, infants were divided into two equal presentation order groups. The experiments were conducted in a laboratory room enclosed in plain brown curtains. The infant sat in a high chair pulled up to a table, 60 cm high, 90 cm long, and 60 cm wide. In the middle of the table was a slot into which a cloth screen (60 cm wide × 40 cm high) was folded. This screen was attached via fishing line and a pulley system to a handle that the experimenter, who was present in the room with the infant, manipulated. The screen thereby could be raised and lowered by the experimenter at the appropriate points in the session. The stimuli were presented on the other side of the screen from the infant so that when raised, the screen obscured the infant's view of the stimulus materials. The stimulus included an adult who produced the pointing gesture and two toys placed at just over arm's length in front of the adult. The toys, a small stuffed dog and a similarly sized plastic car, were placed on small platforms so that they were about 4 inches off the table. A novelty preference paradigm was employed. In the first phase, infants were presented repeatedly with the habituation stimulus. Infant looking was coded by an observer in an adjacent room from a monitor. A different observer was used for each experiment. The observer could communicate with the experimenter through the use of wireless intercom. Looking time was measured using MacXHab 1.4 software, ©1995 (J. Pinto, Stanford University, 1995) running on a Macintosh Powermac 7100 computer. Whenever the infant was judged to be looking at the display, the observer depressed a computer key and released it when the infant looked away. For the purposes of scoring infant looking, the display included the adult and the two toys. If the key was released for 2 seconds, the computer signaled that the trial was over and the observer relayed this information to the experimenter. The experimenter then raised the screen for 2 seconds and lowered it again to begin the next trial. The computer calculated looking time automatically and signaled when the habituation criterion (a 50% decrement in mean looking time over three trials compared to the first three trials) was reached. At this point, the observer signaled to the experimenter that the criterion had been reached. The experimenter raised the screen and switched to the dishabituation stimuli. There were 6 dishabituation trials comprising three different toy trials and three same toy trials in alternating sequence. Order of the trials was counterbalanced. For different toy trials the experimenter pointed at the second toy, which
Page 52
was now in the same location as the first toy had been during habituation trials. In this way, the pointing gesture was exactly the same as before, but it was now towards a different toy. For the same toy trials the experimenter pointed at the same toy as before, but now it was in a different location requiring a change in the spatial organization of the pointing gesture. Dishabituation trials proceeded and were coded in the same manner as habituation trials. Reliability In order to check reliability, all subjects' looking behavior was coded from videotape by a new coder who was naive to the experimental hypotheses. Reliability coding was done in the same way as live coding. Rather than generate an arbitrary reliability statistic for coding acceptability, we wanted to determine if the patterns of results were affected by who did the coding. To this end, for each experiment, we analyzed all of the data for the live coder and for the reliability coder in overall repeated measures analyses of variance (ANOVAs), treating coder as a withinsubjects variable. These analyses all showed a main effect of coder with the looking times being longer for the reliability coder. In addition, for Experiment 2 only, the coder variable interacted significantly with the other variables. Further analyses showed that the effects of the other variables were stronger for the reliability coder than for the live coder. For Experiments 1 and 3, the coder variable did not show interaction effects with any of the other variables. Thus, although the reliability coder tended to record longer looking times in general, the other effects were robust independently of who did the coding or whether it was done live or from replay. In what follows, only the results from the live coder are reported. Experiment 1 In the first experiment, we essentially conducted the experiment sketched in the introduction to this section. Infants sat opposite an adult experimenter who stood in full view and facing the infant. On each trial the experimenter said, "Oh look," and produced a point with arm extended at one of two toys. Throughout the habituation phase the experimenter pointed at the same object on the same side. The particular object used in the habituation phase as well as the side on which it was located was counterbalanced across subjects. Once the habituation criterion had been reached, the experimenter switched the locations of the toys and the experiment moved into the dishabituation phase. Over six trials, the experimenter alternated the location of her point, thereby presenting different toy and same toy trials in alternating sequence. For half of the infants, the
Page 53
experimenter started the dishabituation phase with a different toy trial and for the other half, this phase was started with a same toy trial. To determine whether infants looked longer at the dishabituation stimuli, a twoway repeated measures ANOVA was carried out on the results. For each presentation order, the mean looking time of the last three habituation trials was compared with the looking times for the first different toy trial and for the first same toy trial. This analysis yielded only a significant effect of trial, F(2, 76) = 28.62, p < .001. Analysis of least square means showed that all three means were significantly different from each other, with the greatest recovery of looking for the different toy trials (see Fig. 3.2). A threeway repeated measures ANOVA was also conducted on the dishabituation trials alone with presentation order as a betweensubjects variable and trial type (different toy, same toy) and trial number (first, second, third) as withinsubjects variables. This analysis showed that although looking times decreased significantly through the dishabituation phase, F(2, 76) = 34.32, p < .001, the difference in looking time between the trial types was maintained with a significant main effect for trial type, F(1, 76) = 6.51, p < .05, and no interaction effects. The important result from this experiment was that although infants showed a significant recovery of looking to both of the two dishabituation
Fig. 3.2. Mean looking time for the last three habituation trials and for the two types of dishabituation trials for Experiment 1.
Page 54
stimuli, this recovery was greater for the stimulus event in which the experimenter pointed at the different toy. Because the location of the toys had been switched, the pointing gesture in this event had the same spatial properties as in the habituation event. Therefore, it appears that the infants saw more novelty in the experimenter pointing to the same location at the different object than in the experimenter pointing to the different location at the same object. According to the logic outlined above, this finding is not consistent with our theoretical account but rather implies that infants do represent agent's activity in terms of the intentional relation between the experimenter and the toy. However, we believed that another interpretation was also possible. Because the infants were presented with the same adult pointing gesture in the habituation phase, they may have followed this point repeatedly and thereby ended up attending more during habituation to the first toy. If so, then habituation may have occured to the toy rather than to the whole event of experimenter pointing at the toy. Having habituated to the first toy, it would be natural for the infant to show more recovery of looking to the event in which the experimenter pointed at the second toy. In order to examine this possibility, we conducted a second experiment in which the only modification from the first experiment was the use of two different experimenters to perform the habituation and dishabituation events. If the infants were representing the pointing events in terms of an intentional relation between the adult agent and the toy, then when a new adult was introduced to perform the dishabituation events, both events should be equally novel because the agent of the intentional relation has changed. There should, therefore, be no difference in recovery of looking to the two types of dishabituation trials. However, if the effect was achieved merely through the relative familiarity of the toys, then the results should be unchanged from the first experiment by the introduction of a second experimenter. Experiment 2 In the next experiment, we closely followed the design and method of the first with one important change. Two experimenters were used to deliver the stimuli, one for habituation and one for dishabituation. The first experimenter delivered the habituation stimuli as in the first experiment while the second experimenter waited behind the curtain to the rear of the first experimenter. After habituation had been achieved, the first experimenter raised the screen, switched the locations of the two toys, and then changed places with the second experimenter, who completed the experiment, delivering the dishabituation stimuli. In all other respects, this experiment was an exact replication of the first.
Page 55
The same analyses as for Experiment 1 were carried out and essentially the same pattern of results was observed. The twoway repeated measures ANOVA comparing the mean looking times for the last three habituation trials with the looking times for the first different toy trial and for the first same toy trial showed a significant effect of trial, F(2, 76) = 17.42, p < .001, but no effect of presentation order, or interaction. Analysis of least square means (p < .05) showed that all three means were again significantly different from each other, with the greatest recovery of looking for the different toy trials (see Fig. 3.3). The threeway repeated measures ANOVA (Order × Trial Type × Trial Number) conducted on the dishabituation trials showed that the difference between different toy trials and same toy trials maintained itself throughout the dishabituation phase, F(1,76) = 6.98, p < .05, even as looking times decreased through the phase, F(2, 76) = 10.84, p < .001. The results from this experiment closely matched those from the first experiment. Infants showed significant recovery of looking to both dishabituation events, thereby showing evidence of discriminating both from the habituation event, but significantly more recovery to the event in which the experimenter pointed to the same location at the second toy. These results cannot easily be interpreted to mean that infants represent others in terms of their intentional relations to objects. The infants had become habituated
Fig. 3.3. Mean looking time for the last three habituation trials and for the two types of dishabituation trials for Experiment 2.
Page 56
to the event involving the first experimenter pointing at the first toy. Therefore, the two events in the dishabituation phase involved equally novel intentional relations because a second experimenter acted as the intentional agent. However, the two events were not treated as equally novel by the infants. The experimenter pointing at the second toy was seen as more novel. As a result, a more reasonable interpretation of the results is that the infants had habituated largely to the first toy and thus were more interested in the dishabituation event involving the second toy. These two experiments have left us in a position where we might simply doubt the utility of this approach to investigating infants' understanding of intentional relations. In a sense, the activity of the agent did not enter into the infants' representations; the infant just attended to the designated toys. However, it is important to recognize that the reason the infants attended to the toys in the way that they did was because of the activity of the experimenters. Although we may think of the events as experimental stimuli, for the infants these events included real interactive people, even if they were interacting in a rather stilted manner. The infants attended to the objects that they did because they were directed to do so by the experimenters. In short, the experimental events were themselves triadic interactions between the infants and the experimenters. In order to try to remove this interactive component from the experimental situation, we decided to repeat the first experiment with videotaped stimulus events. Our hope was that by presenting the events on a TV monitor, the infants would be less likely to respond to them as interactive events. Clearly adults, for the most part, respond differently when observing social behavior live and on videotape. Although there is evidence that younger infants will respond interactively to social stimulation presented on TV screens (e.g., Murray & Trevarthen, 1986), this evidence involves dyadic not triadic interaction. Because triadic interaction includes a third object in the shared world of the interactive partners, it is possible that it is less likely to occur when the interactive partners are not simultaneously present with the third object. In any case, our hope was that 12monthold infants would be less likely to respond interactively to videotaped images of people pointing to objects and consequently more likely simply to attend to the overall events. Experiment 3 In the third experiment, we repeated the first experiment except that the stimulus events were presented on a 20inch TV monitor from videotape. The monitor sat facing the infant on the table behind the slot for the screen. The VCR was on the floor behind the table and the experimenter sat on a small stool outofview behind the monitor and controlled the
Page 57
presentation of the stimuli. A total of six videotapes were constructed, two habituation and four dishabituation. For the habituation tapes, an adult female experimenter was taped as she said "Oh look!" and pointed at one of the toys, a different toy for each tape. These tapes contained 14 trials, each 30 seconds long. Two dishabituation tapes were made for each habituation tape, with the position of the two toys switched. These tapes corresponded to the two orders of presentation of the dishabituation trials and all six dishabituation trials were recorded on each tape. During the habituation phase the chosen habituation tape was played until the infant looked away for 2 seconds. The screen was then raised and the tape fast forwarded to the start of the next trial, which was electronically marked on the tape. The screen was lowered and this next trial played. After the habituation criterion had been reached, the screen was raised and the habituation tape was ejected from the cassette recorder. The assigned dishabituation tape was inserted and the screen lowered. The first dishabituation trial was then played. Once the infant had looked away for 2 seconds, the screen was raised and the tape fastforwarded to the beginning of the next trial, which was electronically marked. Again, the same analyses as for the previous experiments were carried out. Only a significant effect of trial, F(2, 78) = 17.24, p < .001, was found for the twoway repeated measures ANOVA comparing the mean looking time of the last three habituation trials with the looking times for the first different toy trial and for the first same toy trial showed for the two presentation orders. Analysis of least square means (p < .05) showed that all three means were significantly different from each other. This result shows that dishabituation occurred to both stimulus events, which is important because it confirms that the infants did notice the switching of the toys in the videotaped displays. In this experiment, unlike the previous two, the greatest recovery of looking was for the same toy trials in which the experimenter pointed to the different location (see Fig. 3.4). The threeway repeated measures ANOVA conducted on the dishabituation trials showed only a small main effect of trial number, F(2, 68) = 3.17, p < .05, with trial one having greater looking times than trial three (least square means, p < .05). The other trial number comparisons were not significantly different from each other. There was not a significant difference between same toy trials and different toy trials when the dishabituation phase was considered as a whole. Whereas the effects in this experiment were not as strong as in the previous two experiments, it was nevertheless the case that infants showed more recovery of looking to the event in which the adult's action changed. It is worth pointing out that this effect occurred even though the infants had noticed the switch in the toys, given that they showed significant recovery of looking to both dishabituation events.
Page 58
Fig. 3.4. Mean looking time for the last three habituation trials and for the two types of dishabituation trials for Experiment 3.
In contrast to those from the first experiment, the results from this experiment are consistent with our original idea that infants represent intentional activity largely in terms of the agent's action and not in terms of the relation between the agent and the object. When the spatiotemporal properties of the agent's action changed, the infant showed more recovery of looking even though the action was still directed at the same object. Less recovery during the dishabituation phase was seen when the action had the same form as in the habituation phase but was directed at the new object. Summary and Conclusion The results from Experiment 1 showed that infants saw more novelty when the adult pointed to a new target, even though the pointing gesture had exactly the same spatiotemporal properties, than when she pointed, using a gesture with different spatiotemporal properties, to the same target as before. The results from Experiment 2 showed that this effect was not likely due to the infants representing the events in terms of the intentional relation between the particular adult and the toy because the same results were achieved even when a different adult performed the pointing gesture
Page 59
in the dishabituation phase. In Experiment 3, the infants saw more novelty when the adult used a different point in relation to the same toy than when she used the same point in relation to a different toy. It is the contrast between the results from Experiments 1 and 3 that is particularly striking. The structure of these two experiments was very similar, with the main difference being the method of event presentation. In Experiment 1, a live adult pointed to one of two toys in front of the infants; in Experiment 3, the same events were shown on TV. Inevitably, this change in method of presentation entailed a number of differences in the characteristics of the stimuli; for example, the overall size of the stimuli was reduced in the TV presentation. However, perhaps most important was the change from live to recorded presentation. The adult in Experiment 1 was a real adult standing in front of the infant with two real toys interposed between them. In Experiment 3, the adult and toys were on TV. Correspondingly, the infants in Experiment 1 appeared to respond appropriately for a live interaction, attending more to the toy to which the adult pointed. However, the infants in Experiment 3 did not respond as though it was a live interaction and appeared to represent the videotaped displays in terms of the adult's gesture. In a sense, then, the change between Experiments 1 and 3 was that the pointing adult had become an object of contemplation rather than an interactive partner and as such the infants paid more attention to the form of the action. Final Words Although the results from our habituation experiments did not come out exactly as we had initially predicted, I think that the results are relevant to the issues considered in the first half of the chapter. The contrast between the results of Experiments 1 and 2 on the one hand and Experiment 3 on the other characterizes the differences between an interactive understanding of intentional relations and a representation of other's agentive activity from a thirdperson perspective respectively. Infants of this age are quite adept at participating with others in shared intentional relations (e.g., Carpenter et al., 1998). When someone points, they will attend to whatever is indicated. Although in effect the infants are thereby attending to the partner's intentional object, such responsivity does not entail an inference that the other is attending to that object. The infant's own response may be linked to the observation of the other's behavior because the thirdperson information provided by the other's activity leads in interactive contexts to an expectation of an interesting experience. However, when the interactive context is removed and the infants do not respond to the adult's pointing, it is possible that a purer assessment of the infants'
Page 60
intepretation of the intentional event is achieved. In that case, it appears that the other's activity is represented in terms of its spatiotemporal properties (or thirdperson information), not in terms of the relation to the intentional object. Nevertheless, the experience that the infants received in Experiment 3 was particularly artificial. In real life, infants do not sit as passive observers of other people's activity. Rather, they participate with others in interactive exchanges where the combined experience of the available first and thirdperson information enables a genuine apprehension of an intentional relation. Where infants differ from older children is that, for infants, the intentional relation is not held to be a property of either of the interactive partners individually but rather is a property of the interaction itself. Acknowledgments Many of the theoretical ideas in this chapter were worked out in collaboration with John Barresi. I am grateful to Amanda Woodward for a number of stimulating discussions and suggestions on the empirical work and to the following who helped with the data collection: Karen Anthony, Paula Bennett, Shannon Macgillivray, and Krista Walters. Finally, thanks to Philip Zelazo for organizing the Toronto conference on intention in action and to David Olson for his valuable feedback on this chapter. References Bakeman, R., & Adamson, L. (1984). Coordinating attention to people and objects in motherinfant and peerinfant interactions. Child Development, 55, 1278 1289. Baldwin, D. A., & Moses, L. (1996). The ontogeny of social information gathering. Child Development, 67, 19151939. Barresi, J., & Moore, C. (1996). Intentional relations and social understanding. Behavioral and Brain Sciences, 19, 107122. Bretherton, I. (1991). Intentional communication and the development of mind. In D. Frye & C. Moore (Eds.), Children' theories of mind: Mental states and social understanding (pp. 4975). Hillsdale, NJ: Lawrence Erlbaum Associates. Butterworth, G., & Cochran, E. (1980). Towards a mechanism of joint visual attention in human infancy. International Journal of Behavioral Development, 3, 253272. Carpenter, M., Nagell, K., & Tomasello, M. (1998). Social cognition, joint attention, and communicative competence from 9 to 15 months of age. Monographs of the Society for Research in Child Development, 63(Serial No. 255, 1142). Cohen, L. B., & Oakes, L. M. (1993). How infants perceive a simple causal event. Developmental Psychology, 29, 421433. Frye, D. (1981). Developmental changes in strategies of social interaction. In M. Lamb & L. Sherrod (Eds.), Infant social cognition (pp. 315331). Hillsdale, NJ: Lawrence Erlbaum Associates.
Page 61
Frye, D. (1991). The origins of intention in infancy. In D. Frye & C. Moore (Eds.), Children's theories of mind. Mental states and social understanding (pp. 15 38). Hillsdale, NJ: Lawrence Erlbaum Associates. Gopnik, A., & Meltzoff, A. N. (1994). Minds, bodies, and persons: Young children's understanding of the self and others as reflected in imitation and ''theory of mind" research. In S. Parker, M. Boccia, & R. Mitchell (Eds.), Selfawareness in animals and humans: Developmental perspectives (pp. 166186). New York: Cambridge University Press. Harris, P. L. (1992). From simulation to folk psychology. Mind and Language, 7, 120144. Moore, C. (1996). Theories of mind in infancy. British Journal of Developmental Psychology, 14, 1940. Moore, C. (in press). Gaze following and the control of attention. In P. Rochat (Ed.), Early social cognition. Mahwah, NJ: Lawrence Erlbaum Associates. Moore, C. (1998). Social cognition in infancy. Monographs of the Society for Research in Child Development, 63(Serial No. 255, 167174). Moore, C., & Barresi, J. (1993). Knowledge of the psychological states of self and others is not only theoryladen but also datadriven. Behavioral and Brain Sciences, 16, 6162. Moore, C., & Corkum, V. (1994). Social understanding at the end of the first year of life. Developmental Review, 14, 349372. Morissette, P., Ricard, M., & Gouin Decarie, T. (1995). Joint visual attention and pointing in infancy: A longitudinal study of comprehension. British Journal of Developmental Psychology, 13, 163175. Murray, L., & Trevarthen, C. (1986). The infant's role in motherinfant communications. Journal of Child Language, 13, 1529. Oakes, L. M., & Cohen, L. B. (1990). Infant perception of a causal event. Cognitive Development 5, 193207. Russell, J. (1996). Agency. Its role in mental development. Mahwah, NJ: Lawrence Erlbaum Associates. Tomasello, M. (1995). Joint attention as social cognition. In C. Moore & P. Dunham (Eds.), Joint attention: Its origins and role in development (pp. 103130). Hillsdale, NJ: Lawrence Erlbaum Associates. Tomasello, M., Kruger, A. C., & Ratner, H. H. (1993). Cultural learning. Behavioral and Brain Sciences, 16, 495511. Wellman, H. M. (1993). Early understanding of mind: The normal case. In S. BaronCohen, H. TagerFlusberg, & D. Cohen (Eds.), Understanding other minds: Perspectives from autism (pp. 1039). Oxford, England: Oxford University Press. Woodward, A. (1995, March). Infants' reasoning about the goals of a human actor. Paper presented at the biennial meeting of the Society for Research in Child Development, Indianapolis.
Page 63
Chapter 4— Having Intentions, Understanding Intentions, and Understanding Communicative Intentions Michael Tomasello Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany In much of my previous work I have focused on human infants' earliest understanding of other persons as intentional agents beginning at around 9 months of age (e.g., Tomasello, 1995a; Carpenter, Nagell, & Tomasello, 1998). In this chapter, I would like to look at a major cause and a major consequence of this 9month social cognitive revolution; and both of these also concern infant intentionality. First, I argue that young children's understanding of other persons as intentional agents results in large part from newly emerging forms of intentionality in their own sensorymotor actions. This account also specifies why nonhuman primates under normal circumstances do not come to understand conspecifics as intentional agents. Second, I explore young children's understanding of a special type of intention that emerges directly on the heels of the 9month revolution, namely, communicative intentions. This understanding is of a piece with infants' other forms of intentional understanding at this age, but it has some special characteristics as well. Once again, comparison with nonhuman primates is instructive in that it helps to identify some of the particular socialcognitive skills involved. The 9Month Revolution Sixmonthold infants interact dyadically with objects, grasping and manipulating them, and they also interact dyadically with other people, expressing emotions back andforth in a turntaking sequence. But at around
Page 64
9 to 12 months of age infants begin to engage in interactions that are triadic in the sense that they involve the referential triangle of child, adult, and some outside entity to which they share attention. Thus, infants at this age begin to flexibly and reliably look where adults are looking (gaze following), use adults as social reference points (social referencing), and act on objects in the way adults are acting on them (imitative learning)—in short, to "tune in" to the attention and behavior of adults toward outside entities. At this same age, infants also begin to use communicative gestures to direct adult attention and behavior to outside entities in which they are interested (imperatives and declaratives)—in short, to get the adult to "tune in" to them. In many cases several of these behaviors come together as the infant interacts with an adult in a relatively extended bout of joint engagement with an object (Bakeman & Adamson, 1984). Most often the term joint attention has been used to characterize this whole complex of triadic social skills and interactions, and it represents something of a revolution in the way infants relate to their worlds (see chapters in Moore & Dunham, 1995). My own view is that infants engage in joint attentional interactions when they begin to understand other persons as intentional agents (Tomasello, Kruger, & Ratner, 1993; Tomasello, 1995a). Intentional agents are animate beings with the power to control their spontaneous behavior, but they are more than that. Intentional agents have goals and make active choices among behavioral means for attaining those goals. Important, intentional agents also make active choices about what they pay attention to in pursuing those goals (see Gibson & Rader, 1979, for the argument that attention is intentional perception). The central theoretical point is that all of the different joint attentional behaviors in which infants follow, direct, or share adult attention and behavior are not separate activities or cognitive domains; they are simply different behavioral manifestations of this same underlying understanding of other persons as intentional agents. Support for this view is provided by a recent study in which a group of 24 infants were followed longitudinally from 9 to 15 months of age (Carpenter et al., 1998). At monthly intervals these infants were assessed on nine different measures of joint attention: joint engagement, gaze following, point following, imitation of instrumental acts, imitation of arbitrary acts, reaction to social obstacles, use of imperative gestures, and use of declarative gestures (including proximal such as "show" and distal such as "point"). In each case, very stringent criteria were used to ensure that infants were attempting either to follow into or to direct the adult's attention or behavior (e.g., alternating attention between goal and adult)—not just reacting to a discriminative stimulus. The findings of most importance in the current context were: (a) all nine of these triadic joint attentional skills emerged for most children by 12 months of age; (b) all of these skills
Page 65
emerged in close developmental synchrony for individual children, with nearly 80% of the infants mastering all nine tasks within a 4month window; and (c) age of emergence was intercorrelated for all the skills. Further support for the hypothesis that joint attentional skills are underlain by an understanding of other persons as intentional agents is provided by two recent sets of findings, using a very different methodology (looking time), which find that 9 months of age is the age at which infants first understand intentional action. First, Csibra, Gergely, Bíró, and Koos (in press) showed 9monthold infants one dot on a screen moving in a goaldirected manner toward another dot on that same screen, detouring around an obstacle to do so (see also Gergely, Nádasdy, Csibra, & Bíró, 1995). Infants clearly showed that they viewed the movements of the first dot as goal directed because they dishabituated if that dot made identical movements only with the obstacle removed (thus making the phantom detour inexplicable from an intentional point of view), but they remained habituated to the dot's different trajectories so long as they were directed at the same goal. Important, 6monthold infants did not behave in this same way. Second, Rochat, Morgan, and Carpenter (1995; see also Rochat & Morgan, 1998) found similar evidence for 9montholds', but not 6montholds', understanding of intentional action in a situation in which infants viewed one moving ball "chasing" another in a goaldirected manner. These two sets of looking studies thus demonstrate, once again, that 9 months of age is a special age in the understanding of intentionality, and they do this using a methodology that does not require the infant to coordinate different pieces of information in the same way as naturally occurring joint attentional skills—thus undermining the view that these skills simply reflect the expansion of some general information processing capacity or the like. Together, all of these different studies serve to undermine several wellknown theories of the emergence of joint attentional skills. Thus, Trevarthen (1979) expresses the view that human infants have adultlike social cognition from birth, and that the emergence of joint attentional behaviors at 9 to 12 months of age simply reflects the development of behavioral performance skills for manifesting this cognition in overt behavior. But the looking time studies of Gergeley and Rochat would seem to undermine this view because they require minimal skills of behavioral performance. BaronCohen (1995) believes that infants are preprogrammed with several independent socialcognitive modules, including an Eye Direction Detector, an Intention Detector, and a Shared Attention Mechanism, each of which has its own predetermined developmental timetable that is affected neither by the ontogeny of the other modules nor by the organism's interactions with the social environment. Similarly, although stressing processes of learning, Moore (1996; Barresi & Moore, 1996) believes that in
Page 66
fants' joint attentional interactions at 9 to 12 months of age represent independent accomplishments as well—in this case independently learned behavioral sequences, each of which having its own critical stimuli, environmental contingencies, and learning history. The intercorrelated and near simultaneous developmental emergence of the different joint attentional behaviors, however, as evidenced most clearly by the study of Carpenter et al. (in press), seriously undermines any view that posits that the different skills are, for one reason or another, independent. It is of course possible that all of the different joint attentional behaviors emerge together at 9 months of age for some other reason, for example, the information processing characteristic that they all require the coordination of two cognitive elements (Moore, 1996; Zelazo, in press). However, to my knowledge, there are no independent assessments of this purported expansion in information processing capacity at 9 months of age—the way there are independent assessments of the understanding of intentionality at 9 months in the studies of Gergeley et al. and Rochat and Morgan—and so I remain skeptical of these somewhat underspecified proposals. Having Intentions and Simulating Others' Intentions In addition to the fact that infants' many skills of joint attention emerge together ontogenetically and in a correlated fashion, any serious theory of early social cognition must also answer the question: Why does this happen at precisely 9 months of age and no other? None of the Nativistic or Learning Theory accounts has a coherent answer to this question, other than some vague and unsubstantiated appeal to a maturational timetable for either innate modules or information processing skills. My own answer is quite specific. Infants' understanding of others as intentional agents depends on the emergence of their own intentionality (cf. Frye, 1981)—in combination with another key component that I elaborate shortly. The story is as follows. In the first months of life infants understand that their behavioral actions achieve results in the external environment, but they do not seem to understand how they do this. Piaget (1952, 1954) devised a number of clever experiments in which infants produced interesting effects on mobiles, toys, and household objects, and then were given the opportunity to reproduce those effects—sometimes in slightly modified circumstances that called for an accommodation on the infant's part. For the first 6 to 8 months of life, the infants repeated behaviors that reproduced interesting results, but they made very few accommodations for the exigencies of particular situations. For example, if the infant managed to shake a rattle and produce an interesting sight and sound because her hand was tethered via a string to the suspended rattle, removal of the
Page 67
string did not lead to any changes of behavior; the infant made the same arm movements as before. Piaget observed many other instances of this "magical" thinking about how actions produce results in the external world. But at around 8 months of age, Piaget's infants seemed to display a new understanding of actionoutcome relations. The new behaviors that evidenced this new understanding were as follows: (a) the use of multiple behavioral means to the same goal; and (b) the recognition of behavioral intermediaries in the pursuit of goals. For example, when the infants wanted to reach a toy, and Piaget placed a pillow as an obstacle in the way, prior to 8 months of age the infants would either start interacting with the pillow, forgetting the original toy, or else they would stay focused on the toy and simply become frustrated. But at 8 months of age the infants reacted to the intervention of the pillow by pausing, then removing the pillow or smashing it down, then proceeding deliberately to grasp the toy. The converse of the removal of obstacles was the use of intermediaries, mostly human intermediaries, to achieve goals. For example, when the infants wanted to operate some toy and could not, they would push the adult's hand toward it and wait for a result (in a few cases they attempted to use inanimate intermediaries, but these were mostly used only a few months later). In all cases the infants showed much more flexibility in using different means to the same goal than in the immediately preceding months, so that Piaget characterized this stage of infant development as "the dissociation of means and ends." Although it is fair to say that prior to 8 months of age infants are acting intentionally in the general sense that they are acting toward a goal, the use of multiple means to the same end and the use of intermediaries indicates a new level of intentional functioning. A means that was useful toward a goal in one circumstance is replaced by another in another circumstance. A behavior that on one occasion was an end in itself, for example, smashing down a pillow, is now only a means to a greater end (grasping the toy). The implication is thus that infants now have a new understanding of the different roles of ends and means in the behavioral act. They have come to differentiate the goal they are pursuing and the behavioral means they use to pursue that goal much more clearly than in their previous sensorymotor actions. The fact that actions can in different circumstances be either end or means, and the fact that some actions are in some cases subordinated as means to others actions, suggests that at this age we have the emergence of truly intentional behavior—in the sense that infants have a goal in mind ahead of time (presumably in the form of an imagined state of affairs in the world) that they clearly differentiate from the various behavioral means from which they must choose. Acting intentionally in this way, of course, does not directly imply anything about infants' understanding of other persons as intentional agents.
Page 68
Many nonhuman animals engage in flexible, intentional behavior without understanding that their conspecifics are intentional as well. The missing component, obviously, is that infants must come to the realization that other persons operate in the same way they do. My own best guess is that this is a cognitive competency that is actually developmentally prior to the emergence of intentionally organized action. Based on a number of phenomena, but especially neonatal imitation, Meltzoff and Gopnik (1993) believe that human infants understand other persons as "like me" very early in ontogeny. Why else would they imitate the facial movements of adults? Because he is dubious about neonatal imitation, Moore (1996) believes that this understanding only emerges somewhat later, based on the infants' experience in social interactions in which they may observe both their own and the adult's simultaneous relations to some third object (i.e., they see the "same" relation from a firstperson and a thirdperson perspective simultaneously). In either case, the main point is that because infants view other persons as "like me" from a very early age, any new understanding of their own functioning leads immediately to a new understanding of the functioning of others: They more or less simulate other persons' behavioral and psychological functioning on analogy to their own more directly and intimately known functioning. Thus, following social theorists such as Vico, Dilthey, Cooley, and Mead, the hypothesis is that our understanding of other persons rests on a special source of knowledge that is not available in other cases of knowing: the analogy to the self. The simple fact is that we have sources of information about the self and its workings that are not available for external entities. As I act I have available various forms of proprioception (correlated with my exteroception) and the internal experience of goals, and the striving for goals, and how these relate to behavioral expression. To the extent that I understand some other being in the world as "like me," and can therefore understand it in terms of the same kinds of internal workings as my own, to that extent can I gain extra knowledge of a special type about how it works. Presumably, the analogy is closest and most natural when it is applied to other human beings, but it can also be stretched to other cases. This is of course one version of a simulation theory of human social cognitive development. Interesting to note, this simulation account may also be applied to findings with even younger infants, before they understand others as intentional agents. Thus, in the studies of Leslie (1984) and Woodward (1998), infants 5 to 6 months of age show surprise (look longer) when they observe other people's hands doing things they normally do not do, for example, changing the object they are grasping or grasping for no apparent reason. These findings clearly suggest that 5 to 6monthold infants have an understanding that other persons are animate beings, with
Page 69
powers of selfmovement, that behave in certain ways. These findings do not imply, nor do any other behaviors at this age imply, that these infants understand others as intentional agents with clearly differentiated means and goals. Consistent with the current simulation theory, this understanding of others as animate (but not yet intentional) corresponds precisely to the way infants understand their own actions at this age, that is, as "magical" procedures that simply make things happen, as studied by Piaget (1954). Thus, 5 to 6monthold infants are surprised when people do things they do not normally do—because they have learned to expect certain things out of these animate beings like themselves—but they do not show surprise in either the Gergeley or the Rochat paradigms when entities change from intentionally guided behavior to random behavior—because they do not even understand their own behavior as intentional. When infants understand themselves as animate, they understand others as animate; when they understand themselves as intentional, they understand others as intentional. The ontogenetic account is thus that human infants identify with other human beings from very early in ontogeny. As long as infants understand themselves only as animate agents with the ability to make things happen in some generalized way, for the first 7 to 8 months or so, that is how they will also understand and simulate other persons. When they begin understanding themselves as intentional agents in the sense that they recognize that they have goals that are clearly separated from behavioral means, at 8 to 9 months of age or so, that is how they understand and simulate other persons as well. This understanding also paves the way for understanding the perceptual choices that others make—their voluntary attention as distinct from their perception. Although at this point we should not push the argument too far, it is also possible that infants make some of these same kinds of simulations, perhaps somewhat inappropriately, to inanimate objects, and this is the source of their understandings of how some physical events "force" others to happen: the first billiard ball is pushing the second with the same kind of force that I feel when I push it (Piaget, 1954). Perhaps this kind of simulation is weaker for infants than the simulation of other persons because the analogy between themselves and inanimate objects is weaker. This account also provides a plausible explanation for why nonhuman primates do not seem to understand others as intentional agents in humanlike ways (for evidence on this point see Tomasello, 1996, 1998; Tomasello & Call, 1997). Chimpanzees and some other nonhuman primates clearly understand something of the efficacy of their own actions on the environment, and indeed they even engage in many kinds of intentional sensorymotor actions in which they use different means toward the same end, remove obstacles, and use intermediaries such as tools. If
Page 70
they do not understand others as intentional agents, then it cannot be because of this factor. Instead, the reason they do not understand others as intentional agents is because they do not identify with conspecifics in the same way as do human beings. And so, even though they themselves act intentionally, they do not understand their conspecifics as intentional beings like themselves (and this might also be the source of their difficulty with physical problems in which they must attempt to understand the causal relations among the actions of inanimate objects). Although we are far from understanding these issues fully, a similar approach might be viable in attempting to account for autistic children's problems with joint attention and related socialcognitive skills (Tomasello et al., 1993). Understanding Communicative Intentions There is a special form of intentional understanding that human children employ soon after the emergence of their other joint attentional skills, and this is the understanding of communicative intentions. The conceptualization and explanation of communicative intentions has a rich philosophical history, beginning in its modern form with Grice (1975; see Levinson, 1983, for an especially useful analysis). I follow along the lines of Clark (1996), who gives a more psychologically based account, and I ignore many subtleties and complexities to focus on one key characteristic: The understanding of communicative intentions involves a partially recursive understanding in which I understand your intentions toward my intentional states. An example helps to clarify. If you come and push me down into a chair, I recognize your intention that I sit down. But if you tell me, "Sit down," I recognize your intention that I attend to your verbal proposal—and as a result that I intentionally sit myself down. If you tell me, "That chair is mine," I recognize your intention that I attend to the chair and its ownership status (from which I may also make inferences about my own sitting behavior). I thus understand: You intend for [me to share attention to [X]]. Understanding of this type is clearly more complex than simply understanding another person's intention simpliciter. To understand that another person wants to kick a ball, I must simply determine their goal with respect to the ball. But to understand what another person intends when they make the sound "Ball!" in my direction, I must determine his or her goal with respect to my intentional/attentional state. Figure 4.1 illustrates a situation in which the infant comprehends a referential communicative intention in which the adult wants the child to follow into her attention
Page 71
Fig. 4.1. Schematic depiction of infant's understanding of a (referential) communicative intention.
to an object—an especially interesting and important case because it is the basis for much of language acquisition (Tomasello, in press). It should be noted that this formulation makes explicit reference to the self ("me"). I believe this is important, and indeed in other places I have attempted to show how the infants' understanding of self emerges from their joint attentional skills in general (Tomasello, 1993, 1995b). The way it goes is this. As just described, 9 to 12 monthold infants begin to follow into and direct the attention of others to outside entities, indicating a new understanding that other people perceive the world and have intentions toward it. Now that the infant can monitor the attention of others to outside entities, it happens that on occasion others are focused on the infant him or herself. The infant then monitors their attention to him or her in a way that was not possible before this new understanding of others as intentional agents. At this point the infant's facetoface interactions with others—which appear on the surface to be continuous with her facetoface interactions from early infancy—are radically transformed. He or she is now interacting with an intentional agent who perceives and intends things toward him or her in a way that was simply not perceived before. When the infant did not understand that others perceive and relate to an outside world, there could be no question of how they perceived and related to me. After coming to this understanding, the infant now can
Page 72
monitor the adult's intentional relation to the world including him or herself (the ''me" of Mead, 1934). This is a necessary prerequisite for understanding that someone else wants me to attend to X, that is, for understanding a communicative intention. I am emphasizing the understanding of communicative intentions because I want to illustrate how the understanding of others as intentional agents leads the child immediately into a new reality—cultural reality. As the child begins to understand others as intentional agents, he or she begins to understand their intentions towards him or her and his or her intentional states. This is the prerequisite form of social understanding for children's comprehension of: (a) the communicative intentions embodied in natural languages, (b) the adult's intentions toward their intentional states in instructional situations, and (c) the deontic forms of culture in which certain forms of behavior and cognition are morally "expected" and "forced" (Bruner, 1993). It is for this reason—that it is at the foundation of language, teaching, and morality—that the understanding of communicative intentions, as a special form of the understanding of intentions, is of such crucial importance in human social cognitive development and in the transformation of human infants from social beings into cultural beings. There is currently some controversy over whether nonhuman primates understand others as intentional and cultural agents. The issues are very difficult, however, as there are wellknown problems in specifying exactly what behavioral differences we should expect to see if an organism does or does not understand the intentions of others (Tomasello & Call, 1997). Although I myself have expressed skepticism on this point (e.g., Tomasello, 1996, in press)—and could argue it at length if needed—in the current context I will simply maintain that nonhuman primates, whatever they may or may not understand about the simple intentions of others, clearly do not understand the communicative intentions of others. First, consider the behavior of chimpanzees as they attempt to comprehend communicative behaviors that are totally novel to them. Tomasello, Call, and Gluckman (1997) indicated for both chimpanzees and 2 to 3yearold human children which of three distinct containers contained a reward by pointing to the correct container, placing a small wooden marker on top of the correct container, or holding up an exact duplicate of the correct container. Children already knew about pointing, but they did not know about using markers and replicas as communicative signs. They nevertheless used these novel signs very effectively to find the reward. In contrast, no ape was above chance for any of the communicative signs that it did not know before the experiment. My explanation of these results is that the children understood that the adult was attempting to direct their attention to the correct container so they could find the prize (i.e., they
Page 73
understood her communicative intentions), whereas the apes were not able to understand that the human being had intentions toward their own attentional states. The apes thus treated the communicative attempts of the human as discriminative cues on a par with all other types of discriminative cues that have to be laboriously learned over repeated experiences. Second, because this experiment was done with chimpanzees attempting to understand the communicative intentions of humans, we should also look at the way chimpanzees communicate with one another in their natural social interactions. The key point here is that in their natural habitats chimpanzees do not ever actively direct the attention of others to outside entities by pointing, showing, offering, teaching, or any other indicative behaviors (Tomasello & Camaioni, 1997). And the signals chimpanzees and other nonhuman primates use are not intersubjectively understood symbols, but simply ritualized intention movements that signal what they are likely to do next. Chimpanzees have intentions towards the interactant's behavior, not towards her intentional/attentional states. And so even if individuals comprehend these behavioral intentions (a dubious proposition, in my opinion), they are not comprehending intentions toward their intentional states (i.e., communicative intentions). Whether chimpanzees raised by humans in humanlike cultural and communicative settings come to understand the intentions and communicative intentions of humans is a fascinating question in need of systematic research (Call & Tomasello, 1996). Conclusion What I have attempted to do here is to focus attention on the human understanding of conspecifics as intentional agents. It is a coherent and powerful understanding that emerges all in a piece at around 9 months of age, in the later phases of human infancy. Although many theories have been propounded to account for the new behaviors that emerge at this age, only a theory that focuses on infant understanding of others as intentional agents can account for the fact that they all emerge together in a correlated fashion at precisely 9 months of age. In the current account, the reason it emerges at this age is because this is the age at which infants' own sensorymotor action begin to display a clear intentional organization in terms of clearly differentiated means and goals. Coupled with infants' previously developed understanding of other persons as "like me," the infant assumes that these persons must operate in the same way as the self, that is, as an intentional agent—which includes not only the intentional control of behavior but also of the intentional control cf perception in the form of voluntary attention. They are thus in a position to understand others at least partially on analogy to the self,
Page 74
that is, by simulation. Nonhuman primates, and some autistic children, do not understand others as intentional agents because they cannot make the link between self and other in the same way as typically developing children. Important, as the 9monthold infant monitors the attention of others to outside entities, including their attention to him or her, this leads to an understanding of "me" as an object in the world. This then opens up the possibility that the infant can understand adults when they express intentions toward his or her intentional states, that is, when they express communicative intentions. Understanding the communicative intentions of others forms the socialcognitive foundation for the acquisition of language and many other cultural skills, as well as for children's internalization of teaching and social norms (Tomasello et al., 1993). Human infants' understanding of the intentionality of others, especially as it applies recursively to the self, is thus in large measure responsible for human infants' entry into the world of culture, including language, which leads them down cognitive paths unique in the animal kingdom. References Bakeman, R., & Adamson, L. (1984). Coordinating attention to people and objects in motherinfant and peerinfant interactions. Child Development, 55, 1278 1289. BaronCohen, S. (1995). Mindblindness.: An essay on autism and theory of mind. Cambridge, MA: MIT Press. Barresi, J., & Moore, C. (1996). Intentional relations and social understanding. Behavioral and Brain Sciences, 19, 107154. Bruner, J. (1993). Commentary on Tomasello et al. "Cultural Learning." Behavioral and Brain Sciences, 16, 515516. Call, J., & Tomasello, M. (1996). The role of humans in the cognitive development of apes. In A. Russon (Ed.), Reaching into thought: The minds of the great apes (pp. 401431). Cambridge, England: Cambridge University Press. Carpenter, M., Nagell, K, & Tomasello, M. (1998). Social cognition, joint attention, and communicative competence from 9 to 15 months of age. Monographs of the Society for Research in Child Development, 63. Clark, H. (1996). Uses of language. Cambridge, England: Cambridge University Press. Csibra, G., Gergely, G., Bíró, S., & Koos, O. (in press). The perception of 'pure reason' in infancy. Cognition. Frye, D. (1981). Developmental changes in strategies of social interaction. In M. E. Lamb & L. R. Sherrod (Eds.), Infant social cognition (pp. 315331). Hillsdale, NJ: Lawrence Erlbaum Associates. Gergely, G., Nádasdy, Z., Csibra, G., & Bíró, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165193. Gibson, E., & Rader, N. (1979). Attention: The perceiver as performer. In G. Hale & M. Lewis (Eds.), Attention and cognitive development (pp. 103141). New York: Plenum Press. Grice, H. (1975). Logic and conversation. In P. Cole and J. Morgan (Eds.), Syntax and semantics, Vol. 3 (pp. 119). New York: Academic Press. Leslie, A. (1984). Infant perception of a manual pick up event. British Journal of Developmental Psychology, 2, 1932.
Page 75
Levinson, S. (1983). Pragmatics. Cambridge, England: Cambridge University Press. Mead, G. (1934). Mind, self, and society. Chicago: University of Chicago Press. Meltzoff, A. N., & Gopnik, A. (1993). The role of imitation in understanding persons and developing a theory of mind. In S. BaronCohen, H. TagerFlusberg, & D.J. Cohen (Eds.), Understanding other minds: Perspectives from autism (pp. 335366). New York: Oxford University Press. Moore, C. (1996). Theories of mind in infancy. British Journal of Developmental Psychology, 14, 1940. Moore, C., & Dunham, P. (1995). Joint attention: Its origin and role in development. Hillsdale, NJ: Lawrence Erlbaum Associates. Piaget, J. (1952). Origins of intelligence in children. New York: Norton. Piaget, J. (1954). The construction of reality in the child. New York: Norton. Rochat, P., & Morgan, R. (1998). Detection of intentional action in infancy. Manuscript submitted for publication. Rochat, P., Morgan, R., & Carpenter, M. (1997). The perception of social causality in infancy. Cognitive Development, 12, 537561. Tomasello, M. (1993). The interpersonal origins of self concept. In U. Neisser (Ed.), The perceived self: Ecological and interpersonal sources of self knowledge (pp. 134140). Cambridge, England: Cambridge University Press. Tomasello, M. (1995a).Joint attention as social cognition. In C. Moore & P. Dunham (Eds.), Joint attention: Its origins and role in development (pp. 103130). Hillsdale, NJ: Lawrence Erlbaum Associates. Tomasello, M. (1995b). Understanding the self as social agent. In P. Rochat (Ed.), The self in early infancy: Theory and research (pp. 350360). Amsterdam: North HollandElsevier. Tomasello, M. (1996). Chimpanzee social cognition. Commentary for Society for Research in Child Development Monographs, 61(3, pp. 161173). Tomasello, M. (1998). Uniquely primate, uniquely human. Developmental Science, 1, 124. Tomasello, M. (in press). Perceiving intentions and learning words in the second year of life. In M. Bowerman & S. Levinson (Eds.), Language acquisition and conceptual development. Cambridge, England: Cambridge University Press. Tomasello, M., & Call, J. (1997). Primate cognition. Oxford, England: Oxford University Press. Tomasello, M., Call, J., & Gluckman, A. (1997). The comprehension of novel communicative signs by apes and human children. Child Development, 68, 1067 1081. Tomasello, M., & Camaioni, L. (1997). A comparison of the gestural communication of apes and human infants. Human Development, 40, 724. Tomasello, M., Kruger, A. C., & Ratner, H. H. (1993). Cultural learning. Behavioral and Brain Sciences, 16, 495552. Trevarthen, C. (1979). Instincts for human understanding and for cultural cooperation: Their development in infancy. In M. von Cranach, K. Foppa, W. Lepenies, & D. Ploog (Eds.), Human ethology: Claims and limits of a new discipline (pp. 530571). Cambridge, England: Cambridge University Press. Woodward, A. (1998). Infants selectively encode the goal object of an actors reach. Manuscript submitted for publication. Zelazo, P. D. (in press). Selfreflection and the development of conciously controlled processing. In P. Mitchell & K. Riggs (Eds.), Children's reasoning and the mind. London: Psychology Press.
Page 77
Chapter 5— Intentions, Consciousness, and Pretend Play Michael Lewis Douglas Ramsay Robert Wood Johnson Medical School, New Brunswick, NJ In this chapter we argue that intentions vary in nature and that this variation has a developmental sequence. Second, we argue that this variation is supported by, and is a function of, the development of consciousness. We also argue that intentions are seen in children's early pretend play and that the emergence of consciousness is related to the emergence of pretend play. On the Nature and Development of Intentions Recently, an 8yearold was talking about how the leaves on the plant in her room moved toward the sun when the plant was placed on the table instead of the window. She said, "The leaves changed their minds." Hers was an intentional stance in regard to plants (Dennett, 1987). That is, she believed in intention and attributed it to plants. She was told by her parents that "it was automatic." Theirs is not an intentional stance visàvis the plant. This example highlights the problem before us. It is clear that some forms of life have intentions, whereas others do not. The developmental question is this: When can we say that intentions emerge? As Searle (1983) pointed out, intentionality is a property of certain mental states. If the organism has no mental states, it can have no intentionality and no inten
Page 78
tions. Most adults believe in a more limited intentional stance than children do, in part because of our belief in what kinds of things have a mental life. On almost everyone's list of those things lacking mental life, intentionality, and intention are plants. When it comes to animal life, the choice becomes more difficult. In general, we say that worms and slugs do not have intentions, whereas domestic animals (such as cats and dogs) as well as higher living forms do. Approaching this same question from a developmental point of view, we say that the adult human has intention, whereas it seems that most would agree that newborns do not. When we ask on what basis such decisions are made, we find that the decision is made around some a priori belief about mental life. It is this problem—the connection between mental and emotional life as it relates to intentionality/intention, consciousness, and pretend play—that we raise. In addressing the issues and problems raised, we identify several specific intentions. These are conscious intentions that are associated with mental states and affect. What Do Infants Do? Lewis and associates (Alessandri, Sullivan, & Lewis, 1990; Lewis, Alessandri, & Sullivan, 1990) found that infants given the opportunity have no difficulty learning how to control a particular set of events in their environment. This finding has been known for some time. Using a variety of procedures, it was shown that infants as young as 6 weeks are able to learn the relation between their response and a consequence (e.g., Papousek, 1967). What is unique about this finding is the observation of the affective component associated with this learning. Infants showed interest and enjoyment as they learned the relation between their simple motor behavior and an outcome. Six to 8weekold infants were observed with smiling and interested faces as they got something to occur when pulling on a string. Infants did not show the same interest and enjoyment when they experienced the same outcome without causing it to happen. How might we explain such behavior? If we observed this same behavior in an adult, we would likely conclude that the adult has mental activity and is working to obtain an intended outcome and that having obtained it, he or she is happy doing this activity. Most of us would have little trouble assuming that the adult "knows" of the relation of string to picture and intends to pull the string to see the picture. On the other hand, we are uncomfortable in ascribing to infants at this age such mental activity, and, therefore, intentional behavior, unless we wish to consider different types of intention (Dennett, 1987). Nonetheless, the idea that the behavior of these young infants is complex and in need of explanation seems reasonable in light of their response
Page 79
when their learned control is removed. Now the infant shows anger and increased arm pulling (cf. Amsel, 1958, 1992). This specific increase in activity and change of affective expression from joy to anger speaks to the directed response of the infant to the loss of control. Again imagine that, instead of observing the infant, we are observing an adult. Imagine also that this adult lives in New Jersey and is working a slot machine in Atlantic City. The machine suddenly stops working and the adult becomes angry and tries even harder to pull the lever to get it to work. We would have no trouble in ascribing intention to this action: The adult intends to get the machine to work. We do not think that conditioning results of the kind we observe in the first part of the study (learning to pull the string) can be readily explained without some notion of mental life and intention (see Rescorla, 1987). However, it is certainly the case that the infant's changing behavior as a consequence of the changing actionoutcome patterns seems to require that we consider the infant's behavior intentional. Again, however, we are confronted with the question of ascribing to so young a child a mental state of the same kind of intention that we are willing to acknowledge in an adult. Our choice, then, is either to argue that the infant's behavior is not intentional or to suggest that the infant has intention, but of a sort somehow different from adult intention. Does Intention Develop and If So, When? The observations made from the study of these infants raises the General question of intentionality and the more focused question of the development of intention. Although the question of intention is relevant for humans at any age, we are particularly concerned with the question of its early development. The problem of development carries with it certain unique features not found when we study intention from the point of view of the adult human; it raises the same questions addressed when intention is considered from a phylogenetic point of view. For us, the topic of the development of intention is particularly important because of this. Moreover, because we are talking about the development of intention, we are confronted with the issue of whether there is change, when it occurs, and how it comes about. Moreover, we need to consider whether it is possible to have intentionality without mental states. As we conclude later, emotional states (specifically, what we call desires) are a type of intention without mental states. In trying to solve this problem, Lewis (1990a, 1990b) argued for the generous distribution of intentionality. Like Searle (1983), Lewis argued that intentionality, in part, is a property of mental states in the same way
Page 80
as is emotionality. The general model presented, framed as "goaldirected behavior is intention," marks the idea that intentionality should be more broadly defined than a property of mental states. Emotional action is intentional, but may not involve the same type of intention as found for mental states. Clearly, such broad definitions run the risk of making everything intentional. We can take advantage of the overinclusiveness of the term and focus instead on the more important issue of the changing nature or different types of intention. The more generous definition should not be construed as a nativistic effort to give the newborn the same capacities as those of the adult. To suggest that it does this constitutes a misreading of any effort at a levels analysis. Indeed, the entire effort of a levels analysis is to argue that very young infants' capacities are not the same as adults'. Levels of Intention Levels or types of intention can be applied either to an ontogenetic process or to differences in species. They may even be helpful in our concern for unconscious intentions (Bargh, 1990). Because Lewis (1990a) articulated these levels elsewhere, we state them here in their broadest form. These levels represent different types of intentionality as well as different levels of intention. Roitblat (1990) also found it useful to distinguish between two types of intentionality, the type without any "mental state about its goals" and the type with. In fact, the type with mental states about its goals looks much like our final level of objective awareness of goals. First, adaptive intentions are goaldirected behaviors that are built into the organism's affectaction patterns. These are desires which are not representational. Second, knowledge intentions have to do with an organism's learned action patterns which have some adaptive features, but which come about through the organism's interaction with its environment. These intentions can be representational; what they are not is objectively selfreflected (see Duval & Wicklund, 1972). It is this lack of selfreflection that distinguishes this level from the third or highest level, conscious intention, which we explain in more detail later. These levels or types of intentions, and the mental and emotional states supporting them, bear some similarity to Piaget's (1936/1952) levels (especially after the first level) and, in fact, also bear some similarity to Dennett's (1987) different orders of intention. This is particularly true for our third level. Indeed, we need to recognize intentional states beyond the second level in order to discuss consciousness, or awareness of the self. Once we recognize that the model proposed is not a nativistic one—newborns are different from 1yearolds, who are different from adults—it is clear that any focus on development must account for changes in level.
Page 81
The details of the movement from level to level have been raised, but they present a continuing problem. Although Piaget for one has tried to articulate the stages of change, the problem has not been solved. Briefly articulated, the developmental problem is this: How do we go from a set of primitives to some higher form not at all like that beginning state? In fact, in transformational and constructivist models of development, these earlier primitives are transformed and disappear in the process of development itself. Understanding development is no easy task. Fodor (1975, 1981) and Sugarman (1987) pointed out the logical problems in explaining the developmental process. Mounoud (1990), however, offered a solution to this problem: He allowed for different types of primitives that occur at each new level. Thus, they are not constructed, but emerge, much like the earliest primitives, from the biology of the organism. We agree, then, that each level has different and more elaborate primitives. Consciousness and Intention as Mental and Emotional States Our three types of intentional processes, adaptive intentions, knowledge intentions, and conscious intentions, are all related to goals. Adaptive intentions are affectively bound and are connected to goals as actions of desire. The emotional property of goals gives them action because emotions, in part, are themselves action patterns (see Darwin, 1872; Searle, 1984). Knowledge intentions are, as the name implies, connected to goals through knowledge. Consider Newell's (1982) construct of a knowledge level. In every system, there is a hierarchy of knowledge. Even at the lowest level of knowledge, the system has a body of knowledge that includes knowledge of the system's goals. Lewis (1995) argued for two aspects of the self, the "machinery of the self" and the cognitive self where there is the "idea of me." The machinery of the self may be thought of as an automatic nonconscious activity, for example, the functioning of the amygdala or hippocampus (see LeDoux, 1990, 1993). The machinery of self aspect controls much of our behavior, has emotional states and affects, learns from experience, and affects our thought processes. For the most part, these processes are unavailable to us. What is available is the mental state or the idea of me. The machinery of self is the equivalent of knowledge, whereas the idea of me is the equivalent of my knowledge of knowledge. Both the machinery and cognitive aspects of the self have intentions. Newell (1982) would not need our third process of selfconsciousness (or knowledge of knowledge of goals) because he would include this as another level of the hierarchy system. To the extent that we claim this third process has no different status—it is the same substance as the second—we have little disagreement. Nevertheless, we see this process as uniquely different from the other two. We base such a claim on no more than the general claim that
Page 82
metacognition is not in the same knowledge level as cognition. Consider the case of memory. A memory of a memory is not of the same class as another memory because the systems (or levels) that support them and the material from which they are made may not be the same. We can think of this difference in knowledge and in knowledge of knowledge (what we wish to call consciousness) as emanating from different parts of the brain (LeDoux, 1990, 1993). Duval and Wicklund (1972) made use of objective selfawareness (knowledge of knowledge of goals) by differentiating it from subjective selfawareness (knowledge of goals). Consider the metaphor, often mentioned, that the property of wetness cannot be derived from the properties of oxygen and hydrogen. In the same fashion, "consciousness emerged at some point in evolution and in ontogeny, in a way underivable from its constituent parts" (Jaynes, 1977, p. 12). Searle (1984) made a similar claim in regard to properties of the mind. In discussing artificial intelligence, he asserted that different types of knowledge come from different types of "machines." A human brain is not a computer and because "understanding is a property that comes from a certain kind of machine only, a machine like the human brain" (Gardner, 1985, p. 174), Searle applied such an analysis to intentionality as well. He saw intentions as caused by the specific properties of the human brain. For this analysis, we claim that knowledge processes or levels can be divided into two levels. In the first, a nonconscious knowledge level exists, that part which we have called the machinery of self, or what Duval and Wicklund (1972) called subjective selfawareness. Most human thinking occurs at this level. Although we are usually unaware (what we mean here by nonconscious) of the processes of our thought, we may become aware of them through special effort. For example, there are times when we desire to trace a set of associations. We all have experienced this ability to retrace our mental steps and arrive at a satisfactory recognition of what occurred by using our objective awareness to tap the subjective self. What we find is that this subjective selfawareness has knowledge and goals and therefore intentions. For example, incomplete actions toward goals set up intentions (perhaps we might call these tensions if we used another language—see Freud, 1915/1959), which are then satisfied. The other knowledge level, objective selfawareness, is when we are aware of our thoughts, feelings, and actions. Phenomenologically, we appear to be watching ourselves. This knowledge level often involves our emotional life, as when we become embarrassed by some action, thought, or feeling of the self (Lewis, 1990b; Lewis, Sullivan, Stanger, & Weiss, 1989). Blushing, as Darwin (1872) first noted, is the most human of all emotional expressions. He stated that "it is not the simple act of reflecting on our own appearance, but the thinking what others think of us, which excites a blush" (p. 325). As an aside, Darwin did not believe that these emotions
Page 83
could emerge much before 3 years of age. It seems clear that this knowledge level (objective selfawareness) is different from the other form of knowing (subjective selfawareness) and that it is reasonable to assume that a metaknowledge (knowledge of knowledge) has differential effects visàvis the levels of intention we seek to explain. Intentions associated with this level of knowledge phenomenonologically appear quite different from intentions at other levels. For example, we intend to study a particular problem. This is a direct operation on a plan, and is quite different from the intentions following this, which proceed from the plan of which we have no awareness (although we might), and which act, so it seems, independently of our generation of other plans or intentions. It is a familiar experience to most of us to be thinking about a problem—a direct intention—when another intention arises, for example, a desire for a glass of water. Intentional action on this desire does not prevent intentions at other levels from proceeding. This possibility appears to occur. In fact, it is one way of conceiving of intrapsychic conflict that was given so much attention in Freud's tripartite notion of the psyche (Freud, 1923/1961). In summary, there are three levels of intention: (a) adaptive intention, or desires, (b) knowledge about intention, or subjective selfawareness, and (c) conscious intention, or objective selfawareness. These levels have an ontogenetic course and may have a phylogenetic one as well. Most important from our point of view, these levels, once developed, do not become transformed as new levels are reached. For each adult human, therefore, these three levels coexist and at times may even conflict. Moreover, a particular level may be involved for one set of goals and another for another set. The regulation of our immunocompetence system is usually left to the first level or perhaps even to the second. Yet even here the third or conscious level may be employed. There are those who subscribe to the view that deliberate attempts (conscious) to make yourself happy, or to reduce stress, will affect the immune system (Dantzer, 1991). Let us state four general propositions that we have so far identified: 1. Intentionality is a property of all goaldirected systems. 2. Development is not transformational. Such a view allows the development of emerging primitives. It also allows for the operation of all achieved levels throughout development. This avoids the inherent problems of regression and how transformations come about. 3. Emotional properties (the desire/goal feature to all representations) underlie all intentions, although emotion is only part of the last two intentions. 4. The levels view allows for the consideration of consciousness as the highest level of intentionality.
Page 84
The Use of Consciousness: Conscious Intentions As we indicated here and elsewhere (Lewis, 1990a, 1990b), the final level of intentionality involves consciousness or objective awareness of intention. Here the organism not only has the flexibility of abstract representation of actions and goals, but is now aware (the self turned toward the self) of these goals and actions. This consciousness allows the child to consider that it has divergent intents. Consciousness is a metaprocess (e.g., the mature human, unlike other organisms, has memory of its memory, as captured by the recursive statement, ''I remember that I remembered to want to do something"). At this point in development, from about 18 months on, intentions are now viewed by the child itself. By this act of consciousness, intentionality itself becomes available for consideration. As such, intentionality itself is changed. The child can now say, "I am aware that I intended to do X, but that is really not what I wish to do." The manipulation of intentions themselves is one property of objective awareness or consciousness. Conscious intent, like all levels of intentions, is supported by emotion. Nevertheless, the new emergent structure of objective selfawareness now becomes the material of desire. Here we mean that to be aware of desires, as opposed to having them but not being aware, becomes a new kind of desire. This supports a new kind of intention. Now, and for the first time, intentions can be disassociated from the direct effects of emotion. Thinking about the self, independent of direct emotion, can sustain intentions. In a sense, one's awareness that "I wish for . . ."—a cognitive act—creates its own emotion. The degree to which the self is not selfaware (not conscious) that it wants something is the degree to which emotions, rather than the other levels of knowledge, control its action. Such a view has been addressed in considering the topic of primary versus secondary processes (Freud, 1915/1959) and is consistent with the assumption that unconscious thought processes are not the equivalent of conscious thought processes. Intentional systems are organized into a series of increasingly more complex structures. Elsewhere we showed that there is a significant change in emotional life with the introduction of objective awareness or consciousness. When consciousness emerges, two classes of emotion are possible: (a) those we have called selfconscious emotions, which include embarrassment, empathy, and envy; and (b) those we called selfconscious evaluative emotions, which include pride, shame, guilt, and satisfaction/competency (Lewis, 1990a, 1992, 1995, 1997; Lewis et al., 1989). These emotions become the basis of a new set of intentions because humans act so as to avoid shame and guilt and to achieve preferable emotions, for example, pride.
Page 85
Although we showed that 2monthold infants act to produce certain learned outcomes and thus possess some type of intention (Alessandri et al., 1990; Lewis et al., 1990), our interest is on the last stage of the development of intentions, conscious intention. We showed that around 15 to 18 months, infants show clear indication of selfrecognition (Lewis & BrooksGunn, 1979; Lewis et al., 1989). All infants show this ability by 24 months. However, this ability does not emerge without a mental age of 15 to 18 months (e.g., BrooksGunn & Lewis, 1980; Loveland, 1987a, 1987b; Mans, Cicchetti, & Sroufe, 1978; Spiker & Ricks, 1984). Moreover, there is little evidence that experience affects this window of emergence. In addition, we showed that temperamentlike characteristics may explain individual differences in the emergence of selfrecognition within this window (DiBiase & Lewis, 1997; Lewis & Ramsay, 1997). Finally, and perhaps most important, we demonstrated that the emergence of selfrecognition is highly associated with the emergence of specific selfconscious emotions including embarrassment and empathy (Lewis, 1992; Lewis et al., 1989). Although these findings point to the emergence of an internal representation of the self, little work has been done relating intentional behavior to the emergence of this capacity. Pretend Play As Conscious Intention It is our belief that pretend or symbolic play reflects the capacity for conscious intention and the objective selfawareness associated with this level of intention. Pretend play requires conscious intention and selfawareness in that children know that their actions on objects are not real; otherwise, the play would not be pretense but a hallucination. The child feeding a doll imaginary food, or the child drinking an imaginary liquid from a cup must know that what he or she is doing is not real. This early pretend play involves, in effect, a negation; namely, this is not what "I pretend it to be but something else." Piaget's (1951/1962) view of symbolic functioning indicated that pretend play requires an awareness of self in order for children to distinguish between what is reality and what is a fantasy. Measuring Pretend Play One difficulty in studying pretend play is the relative infrequency of spontaneous pretend play, particularly in young toddlers. To overcome this problem, investigators used modeling procedures to increase the incidence of target play acts (Fenson & Ramsay, 1981; Largo & Howard, 1979; Pipp, Fischer, & Jennings, 1987; Watson & Fischer, 1977). In large part, the validity of using modeling to index pretense is based on the belief that
Page 86
children only imitate what they understand, although some researchers (e.g., Harris, Kavanaugh, & Meredith, 1994) believe that the reproduction of a given action following modeling may more likely involve motoric imitation than pretend play. Meltzoff (1995) and Asendorpf and colleagues (Asendorpf & Baudonnière, 1993; Asendorpf, Warkentin, & Baudonnière, 1996) successfully used 18montholds' imitative play of the actions of others to indicate their selfawareness. This result does suggest that modeling may well indicate differences in toddlers' emerging symbolic capacities. Pretend play can take place with the self as the object of the pretense (i.e., children pretend to feed themselves) and/or others as the object of the pretense (i.e., when a child pretends to feed a doll; Fenson & Ramsay, 1980, 1981; Lowe, 1975; McCune, 1995; Nicolich, 1977). This past work found a developmental transition from predominately self to dolldirected pretend play. Dolldirected play becomes more frequent than selfdirected play from the latter part of the second year of life. Thus, at its onset pretend play is likely to be much more self than dolldirected. Whereas there is general agreement that dolldirected play involves symbolic capacity, there is some disagreement about whether or not selfdirected play without language requires this capacity (e.g., McCune, 1995). Pretend Play and SelfRecognition If pretend play reflects conscious intention and objective selfawareness, there should be a relation of pretend play to visual selfrecognition, an index of objective self awareness (e.g., Lewis & BrooksGunn, 1979; Lewis et al., 1989). Surprisingly, although there is considerable evidence that pretend play and selfrecognition emerge at approximately the same point in development (e.g., Pipp et al., 1987), previous work did not examine whether this relation is present in individual cases. The likelihood of such a relation is suggested by Asendorpf and colleagues' (Asendorpf & Baudonnière, 1993; Asendorpf et al., 1996) finding that selfrecognition was related to imitative play. We examined the relation between selfrecognition and pretend play. We assessed this relation at 15 months of age when selfrecognition was found to first emerge (Lewis & BrooksGunn, 1979). Evidence for the relation suggests that pretend play does reflect objective selfawareness and conscious intention. Our work involved observation of selfrecognition differences in spontaneous and imitative self and dolldirected pretend play. Selfrecognition and pretend play were observed in a sample of sixtyfive 15montholds. As in previous work (Lewis & BrooksGunn, 1979; Lewis et al., 1989), selfrecognition was assessed by surreptiously applying a spot of rouge to the children's faces and then observing whether they touched
Page 87
the mark on their faces when in front of a mirror. Of the 65 participants, only 11% (7 children) demonstrated selfrecognition. Thus, as would be expected based on our past research, only a minority of children had begun to show selfrecognition at this point in development. Both spontaneous and imitative pretend play were assessed. For spontaneous pretend play, children were observed during a 3minute free play period with a standardized set of toys. We scored the occurrence of pretend play acts involving the following three scenarios: feeding the self or a doll with a spoon; having the self or doll talk on a telephone; and giving the self or doll a drink from a glass. For imitative pretend play, an examiner modeled for the children dolldirected pretend play for the three scenarios in succession after each of which the children's behavior was observed for a 1minute period. The number of self and dolldirected acts involving the three scenarios was summed for spontaneous and for imitative pretend play. Spontaneous Versus Imitative Pretend Play For the whole sample, as might be expected, imitative pretend play was more frequent than spontaneous pretend play (M = 4.52 vs. 1.30).1 Moreover, there was considerable consistency in the amount of spontaneous and imitative pretend play (r = .55), suggesting that both reflect the same underlying symbolic capacities. It was of interest to determine whether selfrecognition differences in pretend play were present for both spontaneous and imitative pretend play. Figure 5.1 shows the frequency of total pretend play as well as the frequency of spontaneous and imitative pretend play for the selfrecognizers and nonselfrecognizers. Selfrecognizers showed significantly more spontaneous and imitative pretend play than nonselfrecognizers. This difference was the same for spontaneous and imitative pretend play.2 The relation between selfrecognition and pretend play supports the view that pretend play reflects both conscious intention and objective selfawareness. That this relation was present for both imitative and spontaneous pretend play indicates that modeling can be used to assess symbolic functioning in toddlers (Fenson & Ramsay, 1981; Largo & Howard, 1979; Pipp et al., 1987; Watson & Fischer, 1977). 1
For the whole sample, the greater frequency of imitative than spontaneous pretend play was significant, F(1, 64) = 99.71, p < .001.
2
Selfrecognition differences in spontaneous versus imitative pretend play were analyzed by a twoway MANOVA with one withinsubjects factor (condition) and one between subjects factor (selfrecognition). Across spontaneous and imitative play, selfrecognizers showed more pretend play than nonselfrecognizers, F(1, 63) = 13.53, p < .001. The interaction between condition and selfrecognition was not significant.
Page 88
Fig. 5.1. Spontaneous and imitative pretend play by selfrecognition.
Self Versus DollDirected Pretend Play For the whole sample, spontaneous pretend play was more self than dolldirected (M = .94 vs. .36), whereas imitative pretend play tended to be more doll than selfdirected (M = 2.49 vs. 2.03).3 Thus, there was no overall difference in amount of self and dolldirected play. That spontaneous pretend play was predominately selfdirected is not surprising given the age of the present sample. Previous findings support the earlier emergence of spontaneous self as opposed to dolldirected pretend play (Fenson & Ramsay, 1980, 1981; Lowe, 1975; McCune, 1995; Nicolich, 1977). The modeling procedure, which involved demonstration of only doll directed acts, led to more imitative doll than selfdirected acts. Nonetheless, the modeling procedure did lead to a considerable amount of selfdirected acts. This likely reflects the tendency for imitative play to indicate children's general level of symbolic understanding at this point in development. Selfrecognition differences in pretend play were examined for both self and dolldirected pretend play. Figure 5.2 shows the frequency of self and dolldirected pretend play for the selfrecognizers and nonselfrecognizers. Selfrecognizers showed significantly more self and dolldirected 3
For the whole sample, differences in self versus dolldirected spontaneous and imitative pretend play were examined by a twoway MANOVA with two withinsubjects factors (condition, play type). This MANOVA yielded a significant interaction, F(1, 64) = 16.78, p < .001. There was a greater frequency of spontaneous self than dolldirected pretend play (p < .001 by post hoc t test) and a trend for a greater frequency of imitative doll than selfdirected pretend play (p < .06 by post hoc t test). Across imitative and spontaneous pretend play, there was no significant difference in the frequency of self and dolldirected pretend play.
Page 89
Fig. 5.2. Self and dolldirected pretend play by selfrecognition.
pretend play than nonselfrecognizers. This difference was the same for self and dolldirected pretend play.4 Thus, in contrast to the view by some (e.g., McCune, 1995) that self as opposed to dolldirected play may not indicate symbolic functioning, the present findings suggest that both self and dolldirected play do reflect emerging pretense in toddlers. Personal Pronoun Use As Another Index of Objective SelfAwareness The use of personal pronouns including "me" and "mine" provides a linguistic demonstration of objective selfawareness. It was of interest to us to obtain confirming evidence that selfrecognition does indicate objective selfawareness by examining the relation between selfrecognition and personal pronoun use in individual children. Previous studies (e.g., Bertenthal & Fischer, 1978; Pipp et al., 1987) examined the relation between the markdirected behavior and selfreferential verbal measures of selfrecognition while children are in front of a mirror. These verbal measures included children's use of the personal pronoun "me" and their name. These studies generally found that verbal measures appear after the markdirected behavior (cf. Harter, 1983). It is 4
Selfrecognition differences in self versus dolldirected pretend play were analyzed by a twoway MANOVA with one withinsubjects factor (play type) and one between subjects factor (selfrecognition). Across self and dolldirected play, selfrecognizers showed more pretend play than nonselfrecognizers, F(1, 63) = 13.52, p < .001. The interaction between play type and selfrecognition was not significant.
Page 90
possible that language behavior during mirror selfrecognition testing may not provide an adequate indication of children's selfreferential language ability, especially in other situations. For example, Lewis and BrooksGunn (1979) found a general inhibition in activity when young children are placed in front of mirrors. Although it is clear that personal pronoun use continues to develop well after the onset of selfrecognition, it may well be that the earliest manifestations of personal pronouns begin at the same point that the selfdirected behavior appears, particularly if measured independently from the mirror procedure. We examined the relation between selfrecognition and personal pronoun use in the same sample of 15montholds. Personal pronoun use was assessed by a self concept questionnaire (cf. Stipek, Gralinski, & Kopp, 1990) completed by the children's mothers. We relied on maternal report to provide a sensitive index of their children's emerging selfreferential verbal capacities, which might not be readily observed in a laboratory setting, particularly when the children were in front of a mirror. Six items from this questionnaire reflected children's use or not of the personal pronouns "me," "my," or "mine," and the phrases, "I'm . . . ,'' "Do it myself," or "I can't." There was evidence for a relation between selfrecognition and personal pronoun use. Selfrecognizers used significantly more personal pronouns that nonself recognizers (M = 1.14 vs. .22). In addition, a significantly greater proportion of selfrecognizers than nonselfrecognizers had begun to use at least one personal pronoun (.71 vs. .16).5 This relation between selfrecognition and personal pronoun use provides additional evidence that selfrecognition, as given by markdirected behavior in front of a mirror, does reflect emerging selfawareness. Consciousness, Intentions, and Theory of Mind In our view, there are three different levels or types of intention that emerge at different points: adaptive intention; knowledge intention, or subjective selfawareness; and conscious intention, or objective selfawareness. At the third level, conscious intention, the individual not only has the flexibility of abstract representation of actions and goals, but is now aware (the self turned toward the self) of these goals and actions. This 5
The greater frequency of personal pronoun use by selfrecognizers than nonselfrecognizers was significant, F(1, 64) = 11.48, p < .001. The greater proportion of selfrecognizers than nonselfrecognizers using at least one personal pronoun was significant (p < .004) by a Fisher's Exact test.
Page 91
awareness is a metacognition as reflected in the statement, "I know that I know." Emotions themselves are changed (added to) as intentions develop from the first to the third level. With conscious intention, one's awareness that "I wish for" creates its own emotion. Elsewhere we showed that there is a significant change in emotion when objective selfawareness emerges (Lewis, 1992, 1995, 1997; Lewis et al., 1989). It is in large part this metaawareness that distinguishes the third level of intention from other levels present at earlier points in development (Butterworth, 1992; Lewis, 1990a, 1990b, 1995; Lewis et al., 1989; Rochat, 1995). When observing play especially in young children, one needs to infer pretense because their language is relatively underdeveloped. Moreover, because spontaneous pretend play is relatively infrequent particularly at its inception, modeling procedures have been used to assess the emerging capacity for pretense. Our results support the presence of pretend play by the middle of the second year of life and indicate the utility of modeling to assess this capacity. Moreover, there is a relation between selfrecognition and pretend play at the point of emergence of selfawareness. This suggests that toddlers at this age are developing a theory of mind (cf. Gallup, 1991; Leslie, 1987; Povinelli & Eddy, 1996), which involves a self pretending, that is, the appearance of the self that knows that it knows or knows that its play is not real. In our view, pretend play reflects the metacognitive ability, "I know that I know," of conscious intention and selfawareness. It marks not only the beginning of a theory of mind, but also likely starts the process that leads to the 3 to 4year switch when children know that what they know is not necessarily what another knows (e.g., Perner, 1991; Rosen, Schwebel, & Singer, 1997; Taylor & Carlson, 1997; Wimmer & Perner, 1983). Although Piaget's (1936/1952, 1937/1954, 1951/1962) early theory of sensorimotor intelligence did not emphasize the development of the self, it would appear that conscious intention and selfawareness are necessary aspects of the various representational abilities that emerge during the second year of life. For example, the capacity for recall memory indicates the presence of conscious intention and selfawareness in that the individual has to know that it remembers that it remembers an event. Similarly, the capacity for both object and person permanence requires the representation of the self for it to know the existence of objects and other people. In this regard, Bertenthal and Fischer (1978) found a strong correspondence between the development of object permanence and visual selfrecognition in the mirror. These researchers, in fact, had expected considerable décalage in development across the two domains. This strong relation is consistent with a central, perhaps organizing role for selfknowledge in representational functioning (cf. Mounoud & Vinter, 1981). Thus, we would expect relations of selfrecognition to various milestones in the
Page 92
development of representation in addition to the present one between selfrecognition and pretend play. References Alessandri, S. M., Sullivan, M. W., & Lewis, M. (1990). Violation of expectancy and frustration in early infancy. Developmental Psychology, 26, 738744. Amsel, A. (1958). The role of frustrative nonreward in noncontinuous reward situations. Psychological Bulletin, 55, 102119. Amsel, A. (1992). Frustration theory—Many years later. Psychological Bulletin, 112, 396399. Asendorpf, J. B., & Baudonnière, P.M. (1993). Selfawareness and otherawareness: Mirror selfrecognition and synchronic imitation among unfamiliar peers. Developmental Psychology, 29, 8895. Asendorpf, J. B., Warkentin, V., & Baudonnière, P.M. (1996). Selfawareness and otherawareness II: Mirror selfrecognition, social contingency awareness, and synchronic imitation. Developmental Psychology, 32, 313321. Bargh, J. A. (1990). Goal intent: Goaldirected thought and behavior are often unintentional. Psychological Inquiry, 1, 248250. Bertenthal, B. I., & Fischer, K. W. (1978). Development of selfrecognition in the infant. Developmental Psychology, 14, 4450. BrooksGunn, J., & Lewis, M. (1980, March). Selfrecognition in handicapped infants and toddlers. Paper presented at the Second International Conference on Infant Studies, New Haven, CT. Butterworth, G. (1992). Origins of selfperception in infancy. Psychological Inquiry, 3, 103111. Dantzer, R. (1991). Stress and disease: A psychobiological perspective. Annals of Behavioral Medicine, 13, 205210. Darwin, C. (1872). The expression of the emotions in man and animals. London: Murray. Dennett, D. C. (1987). The intentional stance. Cambridge, MA: MIT Press. DiBiase, R., & Lewis, M. (1997). The relation between temperament and embarrassment. Cognition and Emotion, 11, 259271. Duval, S., & Wicklund, R. A. (1972). A theory of objective self awareness. New York: Academic. Fenson, L., & Ramsay, D. S. (1980). Decentration and integration of the child's play in the second year. Child Development, 51, 171178. Fenson, L., & Ramsay, D. S. (1981). Effects of modeling action sequences on the play of twelve, fifteen, and nineteenmonthold children. Child Development, 52, 10281036. Fodor, J. A. (1975). The language of thought. New York: Crowell. Fodor, J. A. (1981). Representations: Philosophical essays on the foundation of cognitive science. Cambridge, MA: MIT Press. Freud, S. (1959). Instincts and their vicissitudes. In J. Riviere (Trans.), Collected papers (Vol. 4, pp. 6083). New York: Basic. (Original work published 1915) Freud, S. (1961). The ego and the id. In J. Strachey (Ed. & Trans.), The standard edition of the complete psychological works of Sigmund Freud (Vol. 19, pp. 366). London: Hogarth. (Original work published 1923) Gallup, G. G., Jr. (1991). Toward a comparative psychology of selfawareness: Species limitations and cognitive consequences. In G. R. Goethals & J. Strauss (Eds.), The self: An interdisciplinary approach (pp. 121135). New York: SpringerVerlag. Gardner, H. (1985). The mind's new science. New York: Basic. Harris, P. L., Kavanaugh, R. D., & Meredith, M. C. (1994). Young children's comprehension of pretend episodes: The integration of successive actions. Child Development, 65, 1630.
Page 93
Harter, S. (1983). Developmental perspectives on the selfsystem. In E. M. Hetherington (Ed.), Handbook of child psychology (4th ed.). Socialization, personality, and social development (Vol. 4, pp. 275385). New York: Wiley. Jaynes, J. (1977). The origins of consciousness in the breakdown of the bicameral mind. Boston: Houghton Mifflin. Largo R. H., & Howard, J. A. (1979). Developmental progression in play behavior of children between nine and thirty months, I: Spontaneous play and imitation. Developmental Medicine and Child Neurology, 21, 299310. LeDoux, J. E. (1990). Cognitive and emotional interactions in the brain. Cognition and Emotions, 3, 265289. LeDoux, J. E. (1993). Emotional networks in the brain. In M. Lewis &J. M. Haviland (Eds.), Handbook of emotions (pp. 109118). New York: Guilford. Leslie, A. M. (1987). Pretense and representation: The origin of "Theory of Mind." Psychological Review, 94, 412426. Lewis, M. (1990a). The development of intentionality and the role of consciousness. Psychological Inquiry, 1, 231248. Lewis, M. (1990b). Intention, consciousness, desires and development. Psychological Inquiry, 1, 278283. Lewis, M. (1992). Shame, the exposed self: New York: The Free Press. Lewis, M. (1995). Aspects of self: From systems to ideas. In P. Rochat (Ed.), The self in early infancy: Theory and research (pp. 95115). Amsterdam, The Netherlands: Elsevier. Lewis, M. (1997). Altering fate: Why the past does not predict the future. New York: Guilford Press. Lewis, M., Alessandri, S. M., & Sullivan, M. W. (1990). Violation of expectancy, loss of control, and anger in young infants. Developmental Psychology, 26, 745 751. Lewis, M., & BrooksGunn, J. (1979). Social cognition and the acquisition of self. New York: Plenum. Lewis, M., & Ramsay, D. S. (1997). Stress reactivity and selfrecognition. Child Development, 68, 621629. Lewis, M., Sullivan, M. W., Stanger, C., & Weiss, M. (1989). Self development and selfconscious emotions. Child Development, 60, 146156. Loveland, K. A. (1987a). Behavior of young children with Down syndrome before the mirror: Exploration. Child Development, 58, 768778. Loveland, K. A. (1987b). Behavior of young children with Down syndrome before the mirror: Finding things reflected. Child Development, 58, 928936. Lowe, M. (1975). Trends in the development of representational play in infants from one to three years—an observational study. Journal of Child Psychology and Psychiatry, 16, 3347. Mans, L., Cicchetti, D., & Sroufe, L. A. (1978). Mirror reactions of Down's syndrome infants and toddlers: Cognitive underpinnings of selfrecognition. (Child Development, 49, 12471250. McCune, L. (1995). A normative study of representational play at the transition to language. Developmental Psychology, 31, 198206. Meltzoff, A. N. (1995). Understanding the intentions of others: Reenactment of intented acts by 18monthold children. Developmental Psychology, 31, 838850. Mounoud, P. (1990). Consciousness as a necessary transitional phenomenon in cognitive development. Psychological Inquiry, 1, 253258. Mounoud, P., & Vinter, A. (Eds.). (1981). La reconnaissance de son image chez l'enfant et l'animal. Paris: Delachaux et Niestlé. Newell, A. (1982). The knowledge level. Artificial Intelligence, 18, 81132. Nicolich, L. (1977). Beyond sensorimotor intelligence: assessment of symbolic maturity through analysis of pretend play. MerrillPalmer Quarterly, 23, 89102.
Page 94
Papousek, H. (1967). Experimental studies of appetitional behavior in human newborns and infants. In H. W. Stevenson, E. H. Hess, & H. L. Rheingold (Eds.), Early behavior: Comparative and developmental approaches (pp. 249277). New York: Wiley. Perner, J. (1991). Understanding the representational mind. Cambridge, MA: MIT Press. Piaget, J. (1952). The origins of intelligence in children (M. Cook, Trans.). NewYork: International Universities Press. (Original work published 1936) Piaget, J. (1954). The construction of reality in the child (M. Cook, Trans.). New York: Basic Books. (Original work published 1937) Piaget, J. (1962). Play, dreams and imitation in childhood (C. Gattegno & F. M. Hodgson, Trans.). New York: Norton. (Original work published 1951) Pipp, S., Fischer, K. W., & Jennings, S. (1987). Acquisition of self and mother knowledge in infancy. Developmental Psychology, 23, 8696. Povinelli, D.J., & Eddy, T.J. (1996). What young chimpanzees know about seeing. Monographs of the Society for Research in Child Development, 61(3, Serial No. 247). Rescorla, R. A. (1987). A Pavlovian analysis of goaldirected behavior. American Psychologist, 42, 119129. Rochat, P. (Ed.). (1995). The self in early infancy. Theory and research. Amsterdam, The Netherlands: Elsevier. Roitblat, H. L. (1990). Causation, intentionality, and Cognitive Action Theory. Psychological Inquiry 1, 263265. Rosen, C. S., Schwebel, D. C., & Singer, J. L. (1997). Preschoolers' attributions of mental states in pretense. Child Development, 68, 11331142. Searle, J. (1983). Intentionality: An essay on the philosophy of mind. Cambridge, England: Cambridge University Press. Searle, J. (1984). Minds, brains and science. Cambridge, MA: Harvard University Press. Spiker, D., & Ricks, M. (1984). Visual selfrecognition in autistic children: Developmental relationships. Child Development, 55, 214225. Stipek, D. J., Gralinski, J. H., & Kopp, C. B. (1990). Selfconcept development in the toddler years. Developmental Psychology, 26, 972977. Sugarman, S. (1987). Piaget's construction of the child's reality. Cambridge, England: Cambridge University Press. Taylor, M., & Carlson, S. M. (1997). The relation between individual differences in fantasy and theory of mind. Child Development, 68, 436455. Watson, M. W., & Fischer, K. W. (1977). A developmental sequence of agent use in late infancy. Child Development, 48, 828836. Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception. Cognition, 13, 103128.
Page 95
Chapter 6— Language, Levels of Consciousness, and the Development of Intentional Action Philip David Zelazo University of Toronto In the beginning was the Word. —John, 1:1
In the Second Act of Faust (Part 1), Goethe (1808/1972) has his hero revise the first verse of John's Gospel to read, "In the beginning was the deed! (Im Anfang war die Tat!)." Deed, as the term is used here, does not refer merely to the behavior of organisms, but to action—behavior for which one may be held responsible (e.g., Meldon, 1964).1 Because action is, by definition, the proper subject of moral evaluation, it is natural to suggest that it sets the stage for the entire drama of human life; defines the players, ushers in the audience. And indeed, most psychologists (including Baldwin, 1894/1968, Piaget, 1936/1952, and Vygotsky, 1934/1962, p. 153) have followed Faust in emphasizing the (temporal) priority of action over language in ontogeny. Although there is debate about the exact age at which there is evidence of intentionality in infant behavior, there is now near consensus that infants act intentionally on their environment well before they utter their first word (e.g., see Bruner, 1973; Meltzoff, Gopnik, & Repacholi, chap. 2, this volume; Meltzoff & Moore, 1995; Tomasello, chap. 4, this volume; Trevarthen, 1979). In this chapter, I argue instead that the genesis of action actually does consist in language, because labeling an 1
The German word Tat is sometimes translated as act (e.g., by Barker Fairley; Goethe, 1808/1972).
Page 96
experience is a precondition for the control of behavior by a conscious representation. Moreover, I argue that in addition to this initial, constitutive function, language plays an executive function in the development of action control (because this development depends on discursive reasoning, or rule use). The former arguments follow from the Levels of Consciousness (LOC) model of the role of reflection in the control of thought and action (P. D. Zelazo, in press; P. D. Zelazo &Jacques, 1997; P. R. Zelazo & Zelazo, 1998), whereas the latter derive from the Cognitive Complexity and Control (CCC) theory of the development of deliberate reasoning and intentional action (Frye, Zelazo, & Palfai, 1995; P. D. Zelazo & Frye, 1997). Consciousness and Action The importance of considering consciousness in a scientific examination of action is partly a matter of definition and partly a matter of epistemological prudence. On the one hand, if we are interested in consciousness, then we would be wise to proceed from an examination of action, because action, unlike consciousness, can be observed from an objective, thirdperson perspective. On the other hand, if we are interested in action as consciously controlled behavior, then we are obliged to pursue the problem of consciousness even if it leads us deep into the domain of subjectivity. This definitional interdependence between consciousness and action is underscored by the fact that both are intentional in a similar, if not identical, sense. For Brentano (1874/1973), intentionality (from the Latin intendere, meaning "to stretch") captures the fact that any subjective, conscious experience, no matter how minimal, is an experience of something.2 Unlike things that are merely objective (e.g., rocks), conscious experiences always have intentional objects, whether those intentional objects be simple pain or pleasure, or something more complex, like a proposition. Intentionality is thus a groundlevel characteristic of consciousness, which aims beyond itself at an object, much in the same way that intentional actions aim at something (viz., their goals). Indeed, the object of an intentional action necessarily corresponds to the object of (at least one of) its accompanying conscious states (i.e., it corresponds to the object of a desire [e.g., Baldwin, 2
Brentano (1874/1973) evidently understood mentality to be coextensive with consciousness (see his description of "presentations," p. 78 ff.). This is important because it underscores the fact that the (real) problem of intentionality is not restricted to propositions, which exhibit referential opacity. The fundamental problem of intentionality (i.e., how to cash in Brentano's of) is more closely related to questions concerning the perspectival character of subjective experience (e.g., Eilan, 1995) than it is to questions concerning sense and reference (e.g., Frege, 1892/1949; Russell, 1905/1949). The latter questions (at least in their psychological applications) presuppose solutions to the former (e.g., they take for granted the possibility of acquaintance, in Russell's sense of the term).
Page 97
1892] or an intention [e.g., Adams, 1986]), and action would thus seem to inherit its intentionality from the consciousness that controls it. Behavior that is unaccompanied by conscious states may well be goal directed (in the sense that a thermostat is goal directed), but it is not intentional on this account (and, as a result, it is properly exempt from moral evaluation). The LOC Model The LOC model is an information processing model that is designed to address the interdependence between consciousness and action in development. As an information processing model, it traces the flow of information through a functional system—in this case, illustrating the way in which primitive representations (intentional objects) are operated on as they come to figure in the conscious control of behavior. As a developmental model, the LOC model depicts the way in which this functional system changes in the course of ontogeny, and the consequences of these changes for action control. A central claim of the model is that agerelated changes in action control from infancy through the preschool years (for reviews, see P. D. Zelazo, Carter, Reznick, & Frye, 1997; P. D. Zelazo & Jacques, 1997) are causally dependent on agerelated increases in the highest "level of consciousness" that children can attain when attempting to solve a problem. Levels of consciousness correspond to degrees of recursion in a recurrent or reentrant system.3 The functional process of recursion permits the contents of consciousness at one level of consciousness to be considered in connection with other contents of consciousness at that same level. In this section I review the basic claims of the LOC model, summarizing material that is presented more fully elsewhere (P. D. Zelazo, 1998; for similar summaries, see P. D. Zelazo, in press; P. R. Zelazo & Zelazo, 1998), and emphasizing the way in which this model assigns language a crucial role in the development of consciousness and action. Minimal Consciousness The first, and most important, theoretical primitive in the LOC model is minimal consciousness (cf. Armstrong, 1980), which is meant to be the simplest, but still conceptually coherent, kind of consciousness that we can imagine. From this foundation, it is possible to trace the development of adultlike consciousness using very few theoretical tools (mainly recursion), and to explain a wide range of agerelated changes in behavioral control. 3
The term recursion is used here in the sense of a computer program that calls itself.
Page 98
As argued elsewhere (P. D. Zelazo, 1996), minimal consciousness must be characterized by intentionality in Brentano's sense (i.e., if one is conscious in any sense then one must be conscious of something). In addition, minimal consciousness is conative and functional insofar as it motivates approach and avoidance behavior on the basis of pleasure and pain (see Baldwin, 1894; Dewey, 1896, for discussion). Moreover, contra Carruthers (1989, 1996), who argues that infants (and perhaps even preschoolers) lack subjective experience, there would be something that it is like to be a minimally conscious infant (cf. Nagel, 1974). It should be noted, however, that the infant would not know this, and he or she would be unable later to recall what it was like because, in contrast to more sophisticated forms of consciousness, minimal consciousness is unreflective, presentoriented, and makes no reference to an explicit sense of self that contains or explains subjective experience. While minimally conscious, an infant is conscious of what he or she sees (the object of experience), but he or she is not conscious of seeing what he or she sees, let alone that he or she (as an agent) is seeing what he or she sees. And subsequently, he or she could not remember seeing what he or she saw. Processing at a minimally consciousness level is often described as implicit or unconscious (cf. Gray, 1998), but these terms are potentially misleading because the processing is in fact conscious in an important sense; it is just that the processing is unreflective, and unavailable for subsequent recollection (for reasons that will be addressed). In addition to minimal consciousness, which is depicted in Fig. 6.1 as a geometric plane, the LOC model assumes the existence of longterm memory (abbreviated LTM), which has both semantic and procedural functions, or "systems" (Tulving, 1985). The semantic system stores potentially declarative knowledge, whereas the procedural system stores sensorimotor schemata.
Fig. 6.1. A process model of minimal consciousness. (Adapted from "Children's Rule Use: Representation, Reflection, and Cognitive Control" by P. D. Zelazo & S. Jacques, 1997, in R. Vasta (Ed.), Annals of Child Development, Vol. 12, p. 163). An object in the environment (objA) triggers an intentional representation of that object (IobjA) that is stored in semantic long term memory (LTM); this lobjA, which is causally connected (cc) to objA, becomes the content of minimal consciousness (minC), by way of which it triggers an associated action program stored in procedural LTM.
Page 99
Consider how an infant can act on the basis of minimal consciousness. An actual object in the environment (objA) triggers an intentional object (IobjA) that is "stored" in LTM (within the information processing system, there exist only intentional objects, or representations; objects as things in themselves are noumenal, beyond the reach of phenomenal awareness). This IobjA, with its particular aspectual shape (Searle, 1990), then becomes the content of minimal consciousness. In Fig. 6.1, the IobjA is depicted as causally connected (cc) to a bracketed (i.e., noumenal) objA. When the IobjA becomes the content of minimal consciousness, it can trigger an associated (reflexive or acquired) action program that is "stored" in procedural memory. This program is then executed as a response. Because it mediates responding to the environment, minimal consciousness plays a functional role in the production of behavior. However, behavior at a minimally conscious level is necessarily stereotypical (and hence, lacks key defining features of intentional action). The presence of a rattle, for instance, elicits an habitual stereotypical response—the infant puts the rattle into his or her mouth, assimilating it to the intentional object "suckable thing," which is (reflexively) associated with a sucking schema. Although sensorimotor schemata are modified through practice and accommodation, and coordinated into higher order units, as Baldwin (1894/ 1968) and Piaget (1936/1952) described in detail (see also Cohen, 1998), a minimally conscious infant would not represent these schemata in minimal consciousness (the infant would only be aware of the stimuli that trigger them). A review of the literature (e.g., P. R. Zelazo & Zelazo, 1998) reveals that minimal consciousness manages to account for infant behavior reasonably well until about the end of the first year of life. Onset of Recursive Consciousness At the end of the first year (9 to 12 months), a large number of new abilities appear with highinterval synchrony, which suggests some sort of underlying central, and probably maturational, determinant, as Kagan (1972), P. R. Zelazo (1982), P. R. Zelazo & Leonard (1983), and others (e.g., Frye, 1981; Moore & Corkum, 1994; Tomasello, chap. 4, this volume) have pointed out. Within the span of a few months, infants speak their first words, point protodeclaratively, search for hidden objects, use objects in a functional way, and display deferred imitation, social referencing, and joint attention, among many other major behavioral milestones. According to the LOC model, the synchronous emergence of these behaviors can be parsimoniously explained by the onset of the first new form of consciousness (see Fig. 6.2). When the contents of minimal consciousness are reentered into minimal consciousness so that they can be combined with other contents of minimal
Page 100
Fig. 6.2. Model of recursive consciousness. (Adapted from ''Children's Rule Use: Representation, Reflection, and Cognitive Control" by P. D. Zelazo & S. Jacques, 1997, in R. Vasta (Ed.), Annals of Child Development, Vol. 12, p. 163). When the entire contents of minimal consciousness (minC) are fed back into minC via a recursive process, a higher level of consciousness is achieved, namely recursive consciousness (recC). The contents of recC can be related (Rel1) to a corresponding description (descA) or label, which can then be deposited into working memory where it can serve as a goal (G1) to trigger an action program (stored in procedural LTM). See text and Fig. 6.4 for additional definitions.
consciousness via an identity ("is" or "is a") relation (abbreviated Rel1), then the (first) contents of minimal consciousness become available to the infant in what is called recursive consciousness, or RecC. (Note that RecC is just minimal consciousness depicted in another psychological moment, or functional phase). Naming (or labeling) is the prototype of this recursive process. When a 12monthold looks at his or her father and says, "Daddy," he or she says (effectively), "[Daddy] is Daddy." A perceptual experience is associated with a description from memory, and this description is brought to bear on the perceptual experience. According to the LOC model, there must be two things, the perception and the label, in order for one of them, the perception interpreted in terms of the label, to become an object of conscious experience (note the nontrivial redundancy of this expression). In the absence of a label or description, the contents of consciousness are fleeting and unrecoverable; they are immediately replaced by new stimulation. However, because a label can be "decoupled" (cf. Leslie, 1987) from the experience that is labeled, the label provides a potentially enduring trace of that experience. This has consequences for the qualitative character of experience, and, more important for the present purpose, it has consequences for memory. Specifically, recursive consciousness allows a description to be deposited into longterm memory (so that the experience can subsequently be remembered under that description), and it allows the description to be deposited into working memory, which is understood here simply to be a shortterm buffer for maintaining contents of consciousness in an activated state (e.g., GoldmanRakic, 1990).
Page 101
According to the model, recursive consciousness (at least) of a representation is a necessary precondition for the deposition of a representation into working memory. The contents of working memory (e.g., representations of hidden objects) can serve as goals to trigger action programs indirectly so that the infant is not restricted to responses that are triggered directly, or "suggested" (Baldwin, 1891), by the (minimally conscious) perception of an immediately present stimulus. It is at this point, when the infant is capable of recursive consciousness and able to keep an (absent) object in mind as a goal, that the new behaviors at end of the first year become possible. Now when the infant is presented with a stimulus, such as a rattle or an object hidden at a new location (as in an AnotB task; Piaget, 1936/1952), he or she is able to act mediately, in light of a description of the stimulus, rather than immediately, according to a prepotent, default action program that is triggered directly by a superficial, perceptual experience. For example, the recursively conscious infant may see a rattle, describe it as a rattle, and respond to it under that description, shaking it appropriately (functional play; P. R. Zelazo & Kearsley, 1980). Or, in the case of the AnotB task, the recursively conscious infant may search successfully at location B based on a representation of the hidden object (in working memory), rather than respond directly and perseveratively to the perception of location A (see Marcovitch & Zelazo, in press; P. D. Zelazo, Reznick, & Spinazzola, 1998, for further discussion). In terms of the model (see Fig. 6.2), an objA triggers IobjA, which then becomes the content of minC. However, instead of triggering an associated action program directly (as in a minimally conscious infant), the IobjA is fed back into minimal consciousness (which is called recursive consciousness after one degree of recursion) where it is related via an identity relation to a description (descA) from semantic LTM. This descA is decoupled from the IobjA and deposited into working memory where it serves as a goal (G1) and can elicit an action program even in the absence of objA, and even if the IobjA would otherwise trigger a different, prepotent action program. The Emergence of SelfConsciousness The yearold infant behaves in a way that is considerably more controlled than, say, the 6monthold. However, in the second half of the second year of life, there is a transition in children's behavior that is nearly as dramatic as the one at the end of the first year. In fact, it is so dramatic that Piaget (1936/1952) attributed it to the emergence of symbolic thought. More recent accounts (Kagan, 1981; Lewis & BrooksGunn, 1979; Lewis & Ramsay, chap. 5, this volume) have tended to focus on the implications of this
Page 102
transition for children's consciousness of self, emphasizing children's first use of personal pronouns, their selfrecognition in mirrors, and their display of selfconscious emotions such as shame. Although interpretation of any one of these behavioral developments may be underdetermined, considering them together (i.e., "horizontally"; Fischer & Bidell, 1992) provides interpretive leverage. The LOC model is consistent with these recent accounts, but it also follows Piaget in considering the implications for reasoning and action. According to the model, the behavioral changes during the second year occur because of an additional degree of recursion in consciousness, which yields a new level of consciousness referred to (unsurprisingly) as selfconsciousness. This further reflection upon the contents of consciousness allows children to relate different conscious experiences to a concept of self, which is invoked in order to explain those experiences. One particularly suggestive piece of evidence comes from Kagan (1981), who noted the way in which 2yearolds respond when shown a complex series of steps in the context of an imitative routine. Kagan found that children at this age (but not before) sometimes exhibited signs of distress, as if they knew that the series of steps was beyond the range of their behavioral repertoire, and was not among the means that they had at their disposal. In light of the other behavioral changes mentioned above, and consistent with research on the development on toddlers' action and their understanding of intention (e.g., Frye, 1981, 1991; Meltzoff, 1995, Meltzoff, Gopnik, & Repacholi, chap. 2, this volume; Tomasello, chap. 4, this volume), it seems reasonable to suppose that the child may be considering his or her own capabilities for the first time (Kagan, 1981). Knowledge of one's own capabilities in a situation—that is, consideration of available means as well as desired ends—marks the beginning of a new phase of action control. Keeping a means in mind as a prior intention (Searle, 1983; see Astington, chap. 15, this volume, for discussion) and then translating it into action is an instance of rule use insofar as a conditionally specified means is considered relative to the goal that occasions it. (Note that the means are conditionally specified because they can only be executed when certain antecedent conditions are satisfied.) According to the LOC model, selfdescriptive rules are formulated in potentially silent selfdirected speech, as when we say to ourselves, "If I see a mailbox, then I will mail this letter." In order to control behavior consciously, these rules must be maintained in working memory so that they can constrain responding regardless of fluxional environmental stimulation, which may pull for inappropriate responses. As shown in Fig. 6.3, a selfconscious child can take as an object of consciousness a conditionally specified selfdescription (SdescA) of his or her behavioral potential. This SdescA can then be maintained in working memory as a single rule (R1), considered
Page 103
Fig. 6.3. Model of selfconsciousness. (Adapted from "Children's Rule Use: Representation, Reflection, and Cognitive Control" by P. D. Zelazo & S. Jacques, 1997, in R. Vasta (Ed.), Annals of Child Development, Vol. 12, p. 163).See text for definitions.
against the background of the goal that occasions it. Note that the relevant selfdescriptions can either be recalled from LTM or provided to the child in the form of instructions. In the former case, when the descriptions are recalled from LTM, specific rules may be formulated in an ad hoc fashion from knowledge of relatively general regularities—the variables that figure in representations of regularities may be instantiated with values appropriate to a specific situation. The latter case, in which children are presented with instructions, corresponds to the ruleuse paradigm pioneered by Vygotsky and Luria (see P. D. Zelazo & Jacques, 1997, for a review). In any ruleuse task (unlike rulelearning tasks, such as the Wisconsin Card Sort Test [WCST]; Grant & Berg, 1948), participants are presented with rules and required to use them to guide their behavior. If we assume (rather uncontroversially) that people are conscious of information that they can verbally report, then rule use can be seen as a clear case of the use of a conscious, actionoriented plan to guide behavior. That is, it can be seen as a clear case of intentional action (based on a prior intention; Astington, chap. 15, this volume; Searle, 1983). Recent research on card sorting has confirmed Luria's (1961) finding that even 2yearolds can use a single excitatory (vs. inhibitory) rule systematically to control their behavior (e.g., Jacques, 1995; P. D. Zelazo, Reznick, & Piñon, 1995). Card sorting by an arbitrary rule provides a clear demonstration of rule use because systematic responding is underdetermined by nonlinguistic aspects of the task. In a series of studies, Steve Reznick and I (P. D. Zelazo & Reznick, 1991; P. D. Zelazo et al., 1995) presented 2.5 to 3yearolds with a pair of ad hoc rules for sorting pictures. Children were told, for example, "If I show you something that goes outside the house, put it here; things that go inside, go over here. Here's a snow
Page 104
man, which box does that go in? How about this refrigerator? Etc." The 2.5yearolds started to use these rules, but then they perseverated on one of them. For example, they might put the snowman in the right box, but then assimilate the refrigerator to that same rule. In one study (P. D. Zelazo, Reznick, & Piñon, 1995, Experiment 1), 71% of children's errors were perseverative in this way, which is significantly more than the 50% that would be expected if children were sorting randomly. This failure to switch between two rules is indicative of a limit on their level of reflection. If the 2.5yearolds had been able to step back from their knowledge (via recursion), and consider one rule in relation to the other, then they would recognize the need to choose carefully between them. Reflective Consciousness 1: Contrastive Relations Among Rules In contrast to 2.5yearolds, 3yearolds can successfully employ a pair of arbitrary rules: inside/outside, things that make noise/are quiet, animals that can fly/run, etc. In terms of the levels of consciousness model, this new level of action control is explained by the acquisition of a higher level of consciousness, referred to as reflective consciousness 1 (RefC1; see Fig. 6.4). Threeyearold children can reflect upon a SdescA of a rule (R1) and consider it in relation to another Sdesc (SdescB) of another rule (R2). This relation (Rel2) is a secondorder contrastive relation (as opposed to an identity relation). Both of these rules can then be deposited into working
Fig. 6.4. Model of reflective consciousness 1. (Adapted from "Children's Rule Use: Representation, Reflection, and Cognitive Control" by P. D. Zelazo & S. Jacques, 1997, in R. Vasta (Ed.), Annals of Child Development, Vol. 12, p. 163). See text for definitions.
Page 105
memory where they can be used contrastively to control the elicitation of action programs. As a result, 3yearolds do not perseverate on a single rule when provided with a pair of rules to use. Of course, there are still limitations on children's level of consciousness and consequent selfcontrol. The limitations of reflective consciousness 1 become obvious when children are required to switch flexibly between two conflicting perspectives on a single situation. A perspective on a situation is (minimally) established by a discriminative judgment—a pair of rules—because the rules can vary while the situation remains constant. A single rule fails to constitute a perspective because the rule cannot be differentiated from its condition of application—the situation. The dimensional change card sort, or DCCS (see Fig. 6.5), provides an illustration. In the colorshape version of the DCCS, children are given target cards and then test cards that would be sorted differently by color and shape. They are first told two rules for separating test cards by one dimension (e.g., "All the red ones go here, but all the blue ones go there"), shown test cards, and asked, "Where does this go in the (e.g., color) game?" Then, they are told to switch and sort the cards by the other dimension. The DCCS thus involves conflicting perspectives because a discrimination must be reversed. Threeyearolds can successfully use a single pair of rules, such as "Red here; blue there," in the absence of interference. However, once they have used these rules, they have difficulty switching to the new, incompatible pair of rules ("Rabbits here, boats there") in that same situation despite being told the new rules on every trial. Regardless of which rules are presented first, when 3yearolds are told to switch, they typically perseverate on the first pair of rules. Confirmation that 3yearolds perseverate on a pair of rules, as opposed to learning what to do with specific stimuli, or learning to attend selectively to one dimension (e.g., color), comes from research designed to test predictions derived from the CCC Theory. In a study involving two new versions of the DCCS (see Fig. 6.6), target cards were taken down between the pre and postswitch phases, and replaced with the same or with new cards.
Fig. 6.5. Stimuli from the dimensional change card sort (DCCS).
Page 106
Fig. 6.6. Stimuli from the standard, partial change, and total change versions of the DCCS.
The particular cards only stayed the same in the standard task, so if the children learned responses to particular stimuli, then it was expected they should only perseverate on this version. If they learned to attend selectively to color, then they were expected to perseverate on all three versions, even the total change version. However, if the children persisted in using specific rules (red here, blue there), then they were expected to perseverate in the partial change task, but not in the total change task, which is essentially what was found (although there seemed to be some general disruption of children's set in the partial change version; see Table 6.1). The LOC model is also supported by the presence of an abulic dissociation between knowing rules and using them. In one study (P. D. Zelazo, Frye, & Rapus, 1996, Experiment 4), children who perseverated on the DCCS were then asked a series of knowledge and action questions pertaining to the postswitch rules. One box was designated Ernie's box and the other Big Bird's. Children who were sorting perseveratively according to color were asked, "Whose box do the rabbits go in, in the shape game? Whose box do
Page 107 TABLE 6.1 Numbers of 3 and 4YearOlds Failing and Passing the Standard, Partial Change, and Total Change Versions of the Dimensional Change Card Sort (DCCS)
Card Sort Condition 3yearolds Standard
Performance
Fail
Pass
12
8
Partial Change 10 Total Change 4yearolds Standard
10
4
16
6
14
Partial Change 7
13
Total Change
17
3
the boats go in?" and then they were told (for example), "Ok, now, play the shape game. Here's a red rabbit. Whose box does this go in, in the shape game?" Remarkably, almost all children who perseverated on the DCCS proper (i.e., most 3yearolds) answered the knowledge questions correctly, but they then proceeded to answer the action question incorrectly. Thus, 3yearolds appear to represent the postswitch rules consciously, which is why they can answer the knowledge questions, and they obviously know the color rules because they use them. However, they do not seem to reflect further on their conscious representations of the two pairs of rules, which is why they cannot make a deliberate decision to use the postswitch rules in contradistinction to the preswitch rules (which are now prepotent because they have been invoked and used in this context). This situation is illustrated in Fig. 6.7a. A rule such as A, which links antecedent 1 (a1) to consequent 1 (c1), stands in a contrastive (exclusiveor, Vx) relation with rule B, which links antecedent 2 (a2) to consequent 2 (c2). This contrastive relation defines a universe of discourse (U1), which in this case amounts to a pair of rules. Note that U1 is incompatible with U2, comprising rules C and D, because the relations between antecedents and consequents are reversed. As shown in Fig. 6.7a, both rule pairs are represented, but children cannot formulate a higher order rule that allows them to integrate both rule pairs into a single universe of discourse. Only when the rule pairs are considered in relation to one another can the appropriate rule pair be selected to control action. Otherwise, a rule pair will be selected on the basis of relatively local considerations (i.e., they will select whichever rule pair is most strongly associated with the situation). Confirmation that 3yearolds' difficulty is with selecting the appropriate rules, not executing those rules, comes from a recent study conducted by
Page 108
Fig. 6.7. (a) Unintegrated tree structure showing two pairs of rules. (b) Hierarchical tree structure showing how one rule can be embedded under another and controlled by it. Note: U = universe of discourse; V= exclusiveor, a contrastive relation; s and s2 = setting conditions; a2 and a2 = antecedent 1
conditions; c1 and c2 = consequences.
Jacques, Zelazo, Kirkham, and Semcesen (in press) in which 3yearolds observed a puppet playing versions of the DCCS, and were required to evaluate the puppet's behavior as right or wrong. Thus, the action component was removed from the task altogether. In one experiment, the puppet either perseverated, switched correctly, or switched gratuitously even though it was not instructed to do so. When the puppet perseverated, children incorrectly said it was right, and when the puppet switched correctly, children incorrectly said it was wrong. However, when the puppet switched gratuitously, they correctly said it was wrong. Furthermore, their evaluation performance in the first two conditions was related to their own sorting performance when they were given the standard (Child) DCCS. Together, these findings indicate that 3yearolds select the wrong (i.e., preswitch) rules for the purpose of evaluating the puppets' behavior and for the purpose of guiding their own behavior. It does not appear to be the case that on their own DCCS, they select the right rules but simply cannot inhibit responding according to the wrong pair (for further discussion of the role of inhibition in executive function, see Frye, chap. 7, this volume; Perner, Stummer, & Lang, chap. 8, this volume; P. D. Zelazo & Frye, 1998; P. D. Zelazo, Carter et al., 1997). Reflective Consciousness 2: Putting Rules Into Perspective By 4 or 5 years of age, children switch flexibly on the DCCS, and in terms of the LOC model, children at this age attain a higher level of reflective consciousness (see Fig. 6.8), reflective consciousness 2 (RefC2). At this level, an additional degree of recursion allows children to consider the entire
Page 109
Fig. 6.8. Model of reflective consciousness 2. (Adapted from "Children's Rule Use: Representation, Reflection, and Cognitive Control" by P. D. Zelazo & S. Jacques, 1997, in R. Vasta (Ed.), Annals of Child Development, Vol. 12, p. 163). See text for definitions.
contents of RefC1 in relation to a selfdescription (Sdesc) of comparable complexity. As a result, they see that they have one pair of rules for responding to a situation, and that they have another, incompatible pair of rules for responding to the same situation. From this perspective, they are able to formulate and use a higher order rule for selecting which pair of rules to use. The single system of rules depicted in Fig. 6.7b characterizes the conscious inferences that children at this age are capable of making. A higher order rule (rule E) can now be used to select rules A and B, as opposed to rules C and D. To use rule E, one must consider both the setting condition (s1 or s2) and the lower order antecedent conditions (a1 or a2), in order to determine which consequence follows (c1 or c2): "If playing color, then if red rabbit, then here." Thus, a higher level of consciousness allows the formulation of a more powerful and flexible system of rules to be maintained in working memory.4 As explained by the CCC theory, children's ability to use rule systems at this level of complexity has consequences for their deliberate reasoning about a variety of topics, including physical causality (Frye, Zelazo, Brooks, & Samuels, 1996), "theory 4
Note that the contents of working memory are not identical to the occurrent contents of consciousness (e.g., Baars, 1988). For example, when we rehearse a telephone number, all seven digits remain in an activated state (i.e., in working memory) but we are only aware of subset of them at any given instant.
Page 110
of mind'' (Frye et al., 1995), morality (P. D. Zelazo, Helwig, & Lau, 1996), and external representations (P. D. Zelazo, Sommerville, & Nichols, in press). Summary of the LOC Model The LOC model starts with a characterization of minimal consciousness and charts four major agerelated increases in the level of consciousness that children are able to muster in response to situational demands. These increases are brought about by a functional process of recursion whereby the contents of consciousness are fed back into consciousness so that the contents of consciousness at one level (or moment) become available to consciousness at a higher level, with various consequences for the quality of subjective experience, the potential for recall, and the possibility of action control. Recursion adds depth to subjective experience because more details can be integrated into the experience before the contents of consciousness are replaced by new environmental stimulation. In general, each degree of recursion causes information to be processed at a deeper, less superficial level (Craik & Lockhart, 1972), which increases the likelihood of retrieval (Craik & Tulving, 1975). However, the consequences of this first degree of recursion are perhaps the most categorical in nature, involving, as they do, not merely better recall, but the very possibility of recall. Moreover, with each new level of consciousness, the child is able to exercise a new degree of control over his or her environment and behavior because he or she is able to formulate rules of greater complexity and maintain those rules in working memory. Still, behavior at a particular level of consciousness is subject to limitations that cannot be overcome until yet another level is achieved. Dissociations between knowledge and the ability to use that knowledge to control behavior will occur until incompatible pieces of knowledge are integrated into a single, more complex system via another degree of recursion. These dissociations are important theoretically because they show that children sometimes have conscious knowledge that they are unable to bring to bear on a particular situation under conditions of interference from prepotent responses or representations. In the absence of integration, the particular conscious knowledge that controls behavior is determined by relatively local associations. For example, the rules that are selected and stored in working memory may depend on the way in which a test question is asked, or on what children have done previously in that (selfdescribed) same situation. The LOC model is meant to address the changes in rule use that are described by the CCC theory (Frye et al., 1995; P. D. Zelazo & Frye, 1997), and to illustrate the causal (enabling) relation between changes in consciousness and changes in rule complexity. In general, the relation between
Page 111
levels of consciousness and degree of embedding can be described in terms of the tree structure in Fig. 6.7b: Twoyearolds with selfconsciousness can formulate a single conscious rule, 3yearolds with reflective consciousness 1 can consider a pair of rules at a time, and 5yearolds with reflective consciousness 2 can reflect upon two incompatible pairs of rules that apply to the same situation (i.e., they can consider two different perspectives). Recursion moves consciousness further away from the exigencies of environmental stimulation in what might be called psychological distance (cf. DeLoache, 1993; Dewey, 1931/1985; Sigel, 1993), and this allows for the formulation of increasingly complex discursive reasoning. The Role(S) of Language in Consciousness (and Action) A desire is "conscious" when we have told ourselves that we have it. —B. Russell (1921, p. 31)
Language has long been believed to play an important role in consciousness. Indeed, the ability (at least potentially) to talk about one's experiences has been taken to be a criterion of consciousness by theorists as (otherwise) dissimilar as Freud (1901/1938), Skinner (1969), and Piaget (1964/1967). Piaget (1964/1967), for example, writes that "thought becomes conscious to the degree to which the child is able to communicate it" (p. 19). For Freud (1901/1938), the translation of a drive from a primary to a secondary process is typically achieved through verbal interaction (e.g., with an analyst). Similarly, within a Skinnerian framework, the verbal description of the relation between an environmental condition and an action is tantamount to conscious representation (see Skinner, 1969, p. 245; see also Vygotsky, 1934/1962). On the latter two accounts, consciousness creates the possibility of action control (or action in light of a "precurrent" description; in Skinner's [1969, p. 155] terms). The LOC Model likewise assigns language an instrumental role in consciousness and action, although it does so in two different ways. According to the model, language has both a constitutive function and an executive function in the conscious control of behavior. The Constitutive Function Recursion allows the contents of minimal consciousness to be related to a semantic description (i.e., a label) and interpreted in terms of the description. This description can then be decoupled from the perception described, and maintained in working memory. Unlike minimal conscious
Page 112
ness, whose contents are continually replaced by new perceptual input, recursive consciousness therefore provides a potentially enduring trace of one's experience, and allows one to act in the absence of, or in spite of, direct environmental stimulation. Without language, recursive consciousness would be impossible, and so would the control of behavior by the contents of working memory. Although this initial degree of recursion creates the most categorical changes in the characteristics of consciousness, and has the most dramatic consequences for action, it should be noted that it is only the first step towards the development of adultlike consciousness. Accounts of consciousness (even developmental accounts) typically assume that consciousness is a unitary phenomenon, and that the emergence of consciousness necessarily involves, say, the acquisition of selfconsciousness (Lewis & Ramsay, chap. 5, this volume), or a theory of mind (Carruthers, 1996). However, as argued above, one need not be selfconscious in order to be conscious of objects in the environment or the objects of one's desires. Moreover, although the functional process of recursion eventually (with the acquisition of RefC2) allows children to engage in metarepresentation (i.e., to represent their representations as representations; Perner, 1991), it does not demand that they do so. Metarepresentation, and "theory of mind" in general, involves the acquisition of contingent knowledge (or beliefs) about minds and the way they work, whereas recursion is a functional process within an information processing system. The suggestion that metarepresentation, and an understanding of the appearancereality distinction, is required for conscious experience (Carruthers, 1996) would seem to be a rather severe case of the psychologist's fallacy (James, 1890/1950)—the confusion of the psychologist's standpoint with that of the mental fact about which he or she is making his or her report. As James (1890/1950) writes, "What the thought sees is its own object; what the psychologist sees is the thought's object, plus the thought itself, plus possibly the rest of the world" (p. 197), and he warns us to "avoid substituting what we know the consciousness IS, for what it is a consciousness OF . . ." (p. 197). One need not have a concept of pain, for example, in order to feel it (Anand & Hickey, 1987). A developmental perspective is instructive in helping us to appreciate this point, which is supported empirically by the presence of agerelated dissociations between different levels of consciousness. The Executive Function Language is required for the development of recursive consciousness, and it therefore plays a constitutive role in the emergence of action control, but it also has an executive function. When a description is deposited in working memory, it can be used as a representational proxy for the per
Page 113
ception of a stimulus in order to guide responding indirectly. The possible contents of working memory become more sophisticated with each additional level of consciousness. Unlike infants, toddlers are able to reason discursively, formulating selfinstructions that are instrumental in allowing them to exert conscious control over their environment and over themselves. This kind of reasoning is tantamount to rule use, and there is now strong support for the suggestion that children routinely represent, select, and consult rules in order to guide their behavior (see P. D. Zelazo & Jacques, 1997). Indeed, over the course of the preschool years, there are reliable increases in the complexity of the rule systems that children are able to use. These increases in complexity allow children to bring the right knowledge to bear on their behavior (or their inferences) at the right time in specific situations. Thus, for example, when presented with the DCCS, a 3yearold with reflective consciousness 1 can say to him or herself, "Okay, red ones go here and blue ones go there. This red rabbit is a red one, so I'd better put it over here." In contrast, a 5yearold with reflective consciousness 2 can say, "Before we were playing by color, but now we're playing by shape, so this red rabbit has to go over here now because it's a rabbit." The same child presented with a representational change task (Gopnik & Astington, 1988) may say something like this: "There are sticks in the box, not crayons, but I'm being asked about before I saw that. so the answer is crayons, not sticks." In these examples, children must formulate rules in natural language (based on both their knowledge and the task demands) that allow them to access and focus on particular pieces of knowledge at the time of responding. The LOC model describes the functional processes in consciousness that permit (or constrain) the complexity of the discursive reasoning that children can engage in at different ages. However, performance in any particular situation would also be expected to vary as a function of experience and language proficiency (and especially, proficiency with certain aspects of language, e.g., conditionals; Bowerman, 1986). Recently, correlations with language proficiency have indeed been found for both the DCCS (Bialystok, in press, who compared the performance of mono and bilingual children) and theory of mind (Jenkins & Astington, 1996). This view of the role of language in action is very Vygotskyian: According to Vygotsky and Luria (Luria, 1961; Vygotsky, 1934/1962), language initially serves a communicative function, but it later acquires semantic, syntactic, and directive functions. This directive function allows children to organize and plan their behavior, essentially rendering them capable of voluntary, purposeful behavior. Thus, despite Vygotsky's (1934/1962, p. 153) insistence that "the word was not the beginning— action was there first," he maintained that there was a close connection between language and conscious control. The LOC model follows Vygotsky in this regard. However,
Page 114
in contrast to Vygotsky, for the LOC model, language really is required for action. On this model, the onset of naming signals a split between namer and named, and in an important sense, it brings into being the world as we know it (as opposed to the world as it is experienced). The model attempts to show how this split makes intentional action possible by following the role of language from its initial constitutive function in recursive consciousness through its continued executive function in discursive reasoning. Acknowledgments Preparation of this chapter was supported in part by a grant from NSERC of Canada to P. D. Zelazo. I would like to thank D. Frye and P. R. Zelazo for their valuable contributions to the ideas contained in this chapter, and J. Astington, M. Greif, S. Jacques, and S. Marcovitch for very helpful comments on an earlier draft. References Adams, F. (1986). Intention and intentional action: The simple view. Mind and Language, 1, 281301. Anand, K. J., & Hickey, P. R. (1987). Pain and its effects in the human neonate and fetus. New England Journal of Medicine, 317, 13211329. Armstrong, D. M. (1980). The nature of mind and other essays. Ithaca, NY: Cornell University Press. Baars, B. J. (1988). A cognitive theory of consciousness. New York: Cambridge University Press. Baldwin, J. M. (1891). Suggestion in infancy. Science, 18, 113117. Baldwin, J. M. (1892). Origin of volition in childhood. Science 20, 286288. Baldwin, J. M. (1894). Imitation: A chapter in the natural history of consciousness. Mind, 3, 2655. Baldwin, J. M. (1968). Mental development in the child and the race (3rd ed.). New York: Augustus M. Kelley. (Original work published 1894) Bialystok, E. (in press). Cognitive complexity and attentional control in the bilingual mind. Child Development. Bowerman, M. (1986). First steps in acquiring conditionals. In E. C. Traugott, A. ter Meulen, J. S. Reilly, & C. A. Ferguson (Eds.), On conditionals (pp. 285307). Cambridge, England: Cambridge University Press. Brentano, F. (1973). Psychology from an empirical standpoint. (O. Kraus, Ed.; A. C. Rancurello, D. B. Terell, & L. L. McAlister, Trans.). London: Routledge & Kegan Paul. (Original work published 1874) Bruner, J. S. (1973). Organization of early skilled action. Child Development, 44, 111. Carruthers, P. (1989). Brute experience. Journal of Philosophy, 86, 258269. Carruthers, P. K. (1996). Language, thought, and consciousness: An essay in philosophical psychology. New York: Cambridge University Press.
Page 115
Cohen, L. B. (1998). An information processing approach to infant perception and cognition. In G. Butterworth & F. Simion (Eds.), The development of sensory, motor, and cognitive capacities in early infancy: From perception to cognition (pp. 277300). Hove, UK: Taylor & Francis. Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11, 671 684. Craik, F. I. M., & Tulving, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104, 268294. DeLoache, J. (1993). Distancing and dual representation. In R. R. Cocking & K. A. Renninger (Eds.), The development and meaning of psychological distance (pp. 91107). Hillsdale, NJ: Lawrence Erlbaum Associates. Dewey, J. (1896). [Review of Studies in the evolutionary psychology of feeling]. Philosophical Review, 5, 292299. Dewey, J. (1985). Context and thought. In J. A. Boydston (Ed.) & A. Sharpe (Textual Ed.), John Dewey: The later works, 19251953 (Vol. 6, 19311932; pp. 3 21). Carbondale, IL: Southern Illinois University Press. (Original work published 1931) Eilan, N. (1995). Consciousness and the self. In J. Bermúdez, A.J. Marcel, & N. Eilan (Eds.), The body and the self (pp. 337357). Cambridge, MA: MIT Press. Fischer, K. W., & Bidell, T. R. (1992, Winter). Ever younger ages: Constructive use of nativist findings about early development. SRCD Newsletter, 1, 1011, 14. Frege, G. (1949). On sense and nominatum (H. Feigl, Trans.). In H. Feigl & W. Sellars (Eds.), Readings in philosophical analysis (pp. 85102). New York: Appleton, Century, Crofts. (Original work published 1892) Freud, S. (1938). The interpretation of dreams. In S. Freud, The basic writings of Sigmund Freud (A. A. Brill, Trans.). New York: The Modern Library. (Original work published 1901) Frye, D. (1981). Developmental changes in strategies of social interaction. In M. E. Lamb & L. R. Sherrod (Eds.), Infant social cognition (pp. 315331). Hillsdale, NJ: Lawrence Erlbaum Associates: Frye, D. (1991). The origins of intention in infancy. In D. Frye & C. Moore (Eds.), Children's theories of mind: Mental states and social understanding (pp. 15 38). Hillsdale, NJ: Lawrence Erlbaum Associates. Frye, D., Zelazo, P. D., Brooks, P., & Samuels, M. (1996). Inference and action in early causal reasoning. Developmental Psychology, 32, 120131. Frye, D., Zelazo, P. D., & Palfai, T. (1995). Theory of mind and rulebased reasoning. Cognitive Development, 10, 483527. Goethe, J. W. (1972). Faust (B. Fairley, Trans.). Toronto: University of Toronto Press. (Original work published 1808) GoldmanRakic, P. S. (1990). The prefrontal contribution to working memory and conscious experience. In J. C. Eccles & O. Creutzfeldt (Eds.), The principles of design and operation of the Brain (pp. 389407). New York: SpringerVerlag. Gopnik, A., & Astington, J. W. (1988). Children's understanding of representational change and its relation to the understanding of false belief and the appearance reality distinction. Child Development, 59, 2637. Grant, D. A., & Berg, E. A. (1948). A behavioral analysis of degree of reinforcement and ease of shifting to new responses in a Weigltype cardsorting problem. Journal of Experimental Psychology, 38, 404411. Gray, J. A. (1998). Abnormal contents of consciousness: The transition from automatic to controlled processing. In H. H. Jasper, L. Descarries, V. F. Castellucci, & S. Rossignol (Eds.), Consciousness: At the frontiers of neuroscience, advances in neurology (Vol. 77, pp. 195212). New York: LippincottRaven Press. Jacques, S. (1995). The development of rule use during the preschool period. Unpublished master's thesis, University of Toronto, Ontario, Canada.
Page 116
Jacques, S., Zelazo, P. D., Kirkham, N. Z., & Semcesen, T. K. (in press). Rule selection and rule execution in preschoolers: An errordetection approach. Developmental Psychology. James, W. (1950). The principles of psychology, Vol. 1. New York: Dover. (Originally published 1890) Jenkins, J. M., & Astington, J. W. (1996). Cognitive factors and family structure associated with theory of mind development in young children. Developmental Psychology, 32, 7078. Kagan, J. (1972). Do infants think? Scientific American, 226, 7482. Kagan, J. (1981). The second year. Cambridge, MA: Harvard. Leslie, A. M. (1987). Pretense and representation: The origins of "theory of mind." Psychological Review, 94, 412426. Lewis, M., & BrooksGunn, J. (1979). Social cognition and the acquisition of self New York: Plenum Press. Luria, A. R. (1961. The role of speech in the regulation of normal and abnormal behaviour (J. Tizard, Ed.). New York: Pergamon Press. Marcovitch, S., & Zelazo, P. D. (in press). The AnotB error: Results from a logistic metaanalysis. Child Development. Meldon, A. I. (1964). Action. In D. F. Gustafson (Ed.), Essays in philosophical psychology (pp. 5876). Garden City, NY: Doubleday. Meltzoff, A. N. (1995). Understanding the intentions of others: Reenactment of intended acts by 18monthold children. Developmental Psychology, 31, 838850. Meltzoff, A. N., & Moore, M. K. (1995). Infants' understanding of people and things: From body imitation to folk psychology. In J. Bermúdez, A. J. Marcel, & N. Eilan (Eds.), The body and the self (pp. 4369). Cambridge, MA: MIT Press. Moore, C. & Corkum (1994). Social understanding at the end of the first year of life. Developmental Review, 14, 349372. Nagel, T. (1974). What is it like to be a bat? The Philosophical Review, 83, 435450. Perner, J. (1991). Understanding the representational mind. Cambridge, MA: MIT Press. Piaget, J. (1952). The origins of intelligence in children (M. Cook, Trans.). New York: Vintage. (Original work published 1936) Piaget, J. (1967). Six psychological studies (D. Elkind, Ed.; A. Tenzer, Trans.). NewYork: Vintage. (Original work published 1964) Russell, B. (1921). Analysis of mind. London: Allen & Unwin. Russell, B. (1949). On denoting. In H. Feigl & W. Sellars (Eds.), Readings in philosophical analysis (pp. 103115). New York: Appleton, Century, Crofts. (Original work published 1905) Searle, J. (1983). Intentionality. Cambridge, England: Cambridge University Press. Searle, J. (1990). Consciousness, explanatory inversion, and cognitive science, Behavioral and Brain Sciences, 13, 585642. Sigel, I. (1993). The centrality of a distancing model for the development of representational competence. In R. R. Cocking & K. A. Renninger (Eds.), The development and meaning of psychological distance (pp. 91107). Hillsdale, NJ: Lawrence Erlbaum Associates. Skinner, B. F. (1969). Contingencies of reinforcement: A theoretical analysis. Englewood Cliffs, NJ: Prentice Hall. Trevarthen, C. (1979). Communication and cooperation in early infancy: A description of primary intersubjectivity. In M. Bullowa (Ed.), Before speech: The beginning of interpersonal communication (pp. 321347). Cambridge, England: Cambridge University Press. Tulving, E. (1985). How many memory systems are there? American Psychologist, 40, 385398. Vygotsky, L. S. (1962). Thought and language (E. Hanfmann & G. Vakar, Trans.). Cambridge, MA: MIT Press. (Original work published 1934) Zelazo, P. D. (1996). Towards a characterization of minimal consciousness. New Ideas in Psychology, 14, 6380. Zelazo, P. D. (1998). The development of consciousness. Manuscript in preparation.
Page 117
Zelazo, P. D. (in press). Selfreflection and the development of consciously controlled processing. In P. Mitchell & K.J. Riggs (Eds.), Children's reasoning and the mind. Hove, UK: Psychology Press. Zelazo, P. D., Carter, A., Reznick, J. S., & Frye, D. (1997). Early development of executive function: A problemsolving framework. Review of General Psychology, 1, 198226. Zelazo, P. D., & Frye, D. (1997). Cognitive complexity and control: A theory of the development of deliberate reasoning and intentional action. In M. Stamenov (Ed.), Language structure, discourse, and the access to consciousness (pp. 113153). Amsterdam & Philadelphia: John Benjamins. Zelazo, P. D., & Frye, D. (1998). II. Cognitive complexity and control: the development of executive function in childhood. Current Directions in Psychological Science, 7. Zelazo, P. D., Frye, D., & Rapus, T. (1996). An agerelated dissociation between knowing rules and using them. Cognitive Development, 11, 3763. Zelazo, P. D., Helwig, C. C., & Lau, A. (1996). Intention, act, and outcome in behavioral prediction and moral judgment. Child Development, 67, 24782492. Zelazo, P. D., & Jacques, S. (1997). Children's rule use: Representation, reflection, and cognitive control. In R. Vasta (Ed.), Annals of child development, vol. 12. (pp. 119176). London: Jessica Kingsley Press. Zelazo, P. D., & Reznick, J. S. (1991). Agerelated asynchrony of knowledge and action. Child Development, 62, 719735. Zelazo, P. D., Reznick, J. S., & Piñon, D. E. (1995). Response control and the execution of verbal rules. Developmental Psychology, 31, 508517. Zelazo, P. D., Reznick, J. S., & Spinazzola, J. (1998). Representational flexibility and response control in a multistep, multilocation search task. Developmental Psychology, 34, 203214. Zelazo, P. D., Sommerville, J. A., & Nichols, S. (in press). Agerelated changes in children's use of external representations. Developmental Psychology. Zelazo, P. R. (1982). The yearold infant: A period of major cognitive change. In T. Beaver (Ed.), Regressions in mental development: Basic phenomena and theoretical alternatives (pp. 4779). Hillsdale, NJ: Lawrence Erlbaum Associates. Zelazo, P. R., & Kearsley, R. (1980). The emergence of functional play in infants: Evidence for a major cognitive transition. Journal of Applied Developmental Psychology, 1, 95117. Zelazo, P. R., & Leonard, E. L. (1983). The dawn of active thought. In K. Fisher (Ed.), Levels and transitions in children's development (pp. 3750). San Francisco: JossesBass. Zelazo, P. R., & Zelazo, P. D. (1998). The emergence of consciousness. In H. H. Jasper, L. Descarries, V. F. Castellucci, & S. Rossignol (Eds.), Consciousness: At the frontiers of neuroscience: Advances in neurology (Vol. 77, pp. 149165). New York: LippincottRaven Press.
Page 119
Chapter 7— Development of Intention: The Relation of Executive Function to Theory of Mind Douglas Frye University of Pennsylvania There are many reasons to think that theory of mind and executive function would be unrelated. They seem very different. It is not apparent why the development of mental state understanding would be connected to the control of one's own actions. Theoretical grounds exist for expecting them to be independent. Indeed, the rationale of domain specific accounts (BaronCohen, 1994; Gopnik & Wellman, 1994) is that the development of theory of mind should be autonomous. Nonetheless, relations between the development of theory of mind and executive function have been detected for individuals with autism (Ozonoff, Pennington, & Rogers, 1991), in studies of deception (Russell, Mauthner, Sharpe, & Tidswell, 1991), and for typically developing preschoolers (Frye, Zelazo, & Palfai, 1995). Recent studies (Carlson, 1997; Hughes, 1998; Ozonoff & McEvoy, 1994) have confirmed these connections. The confirmation of an empirical connection between theory of mind and executive function invites the question of the nature of the theoretical link between the two. This question is of interest because all of the logical possibilities are available. Theory of mind could subsume executive function, executive function could subsume theory of mind, or the two could be distinct and irreducible to each other. As the example of domain specificity has already shown, these possibilities carry different implications for what theory of mind itself is understood to be. They also carry different implications for executive function. Historically, the construct of executive function has suffered from its ready use as a covering term for a hetero
Page 120
geneous collection of different abilities (Zelazo, Carter, Reznick, & Frye, 1997). Clarifying the relation between executive function and theory of mind could help to reveal the relative importance of these different abilities. The view taken here is that theory of mind and executive function are distinct, but so closely related that each essentially implies the other. This view has the consequence of drawing theory of mind much closer to action, and of shifting the emphasis in executive function from response inhibition to planning. Proposals for the Relation It is fortunate that all of the logical possibilities for the relation between theory of mind and executive function exist—because they have all been claimed. Perner (1998) has advocated the position that the relation obtains because the relevant executive function depends on theory of mind. Perner's (1991b) metarepresentational account is well adapted to its original purpose, the explanation of theory of mind and especially false belief. It is easy to see how understanding representation (as representation) applies to understanding how we form beliefs about the world. The account applies less naturally to desire given that our desires do not represent the world as much as indicate how we would like the world to be (Astington, 1991; Perner, 1991a). The same complication arises with executive function in that control of action involves altering the world rather than simply representing it. However, Perner (1998; Perner, Stummer, & Lang, chap. 8, this volume) has theorized that metarepresentation is necessary for executive function, particularly when a new action must be taken in a situation in which another action has become well established. He argues that the child must be able to understand how the old action is represented (i.e., metarepresent it) in order to lessen its probability so that the new action can be carried out. Russell (1996) has taken the opposite stance, speculating that changes in executive function may explain improvements in theory of mind. Russell reached this conclusion from the study of deception. Deception depends on action because we must do something (or refrain from acting when it is expected) in order to mislead someone. Early on, it was recognized that deception and theory of mind are empirically related (Wimmer & Perner, 1983); however, the conceptual similarity that they both involve false belief could be responsible, given that deception can be defined as acting to give someone a false belief. Russell et al. (1991) devised a straightforward test in which children could deceive a competitor merely by pointing to the incorrect location of a candy. Threeyearolds produced a high frequency of perseverative pointing to the actual location, a response pattern that is
Page 121
characteristic of failures in executive function. The results of a subsequent study (Russell, Jarrold, & Potel, 1994) opened the question of whether deception was indeed the problem because perseveration occurred even without the presence of a competitor. These results favor the possibility that young children's difficulty with deception is not conceptual but the consequence of poor executive control. In this regard, Russell (1996) introduces a valuable distinction between executivecompetence versus executiveperformance failures. Executive competence refers to actual abilities to control action that may be lacking because they have not yet been developed. Executive performance, in contrast, refers to more superficial difficulties in carrying out actions that may consequently mask children's real abilities. Russell proposes an executivecompetence explanation of the relation between executive function and theory of mind. In both deception and false belief, young children's inability to control their own thinking sufficiently makes it difficult for them to pull their thoughts away from reality in order to determine how someone else can be made to or will mistakenly see the situation. Carlson, Moses, and Hix (1998) have recently formulated an executive performance account of the relation between theory of mind and executive function. They verified that 3yearolds have difficulty pointing away from the location of an object even when the person to be deceived was not present. They also found, however, that the children were more successful when they just had to attach a pictorial cue to the empty location rather than point to it. This finding indicated that pointing, which is well established as a way of indicating the location of something, may be a prepotent response. If young children lack the response inhibition to refrain from pointing to the object's actual location, they will appear to be poor at deception even though they may understand it perfectly well. This executiveperformance failure may also extend to theory of mind because many children respond by pointing in the changeoflocation false belief task. This account takes the position, then, that theory of mind and executive function each exist in their own right but that theory of mind success may be masked by the executive performance failure of inadequate behavioral inhibition. One last possibility is set out by Cognitive Complexity and Control (CCC) theory (Frye et al., 1995; Zelazo & Frye, 1997). This account accepts that theory of mind and executive function are distinct. One cannot be reduced to the other. However, it suggests that it may be possible to discover a formal relation between the two. Although it is not apparent in the characterizations they previously have been given, the two may rest on the same set of embedded rules. Rule use develops over the course of the preschool period (Zelazo &Jacques, 1996) with embedding first occurring at about 4 years of age. The same embedded rules, employed in slightly different ways, guide the inferences necessary for theory of mind and allow
Page 122
the formulation of action that results in improved executive control. As a consequence, the approach takes an executivecompetence view that places more emphasis on planning and deliberative action in executive function (Zelazo, Carter, et al., 1997) rather than on just response inhibition. In line with that emphasis, it proposes that young children's difficulty with deception may relate more to planning than to executing actions (Brooks, Frye, & Samuels, 1997). The approach also furnishes an interpretation of action that gives a different insight into what it could share with theory of mind. Similarity between Theory of Mind and Executive Function Much of the research on preschool executive function within the CCC approach has investigated a seemingly uncomplicated card sorting task (Frye et al., 1995; Zelazo, Frye, & Rapus, 1996). The task taps executive function because it requires children to act according to stated rules. Children are shown cards with pictures that vary on two dimensions. Thus, for the dimensions of color and shape, there might be cards picturing a yellow flower and others showing a blue car. The children are asked to match these to two target cards (a blue flower and a yellow car), first for one dimension and then the other. They are explicitly told the relevant dimension and correct rule every time. (''Remember this is the color game. In the color game, if it is yellow, it goes here. If it is blue, it goes there. Where does this yellow flower go?") Because of the cards used, there is always a conflict present. The target that a test card matches for one dimension will not be the one it matches for the other dimension. The procedure produces a very clear age pattern. Threeyearolds can sort by whichever dimension is presented first, but then continue to sort by that dimension even when they are asked to employ the other one. Their perseverative responding is characteristic of a failure in executive function. Fiveyearolds typically sort without errors, switching easily from one dimension to the other. As was indicated previously, children's success in switching is correlated with success in understanding false belief and other theory of mind developments even after the effects of age have been removed (Frye et al., 1995; Carlson, 1997). A plausible interpretation of these results is that 3yearolds have weaker response inhibition. When they sort by one dimension several times, those responses become prepotent. Switching to a new dimension is difficult because the prepotent responses must be inhibited before the new, correct ones can be made. Fiveyear olds with better developed response inhibition are able to refrain from making the old responses, and consequently are
Page 123
able to switch to sorting by the new dimension. Other explanations can be given for this change. CCC theory offers an alternative that is described in more detail later. A series of studies designed in part to compare the two accounts is also reviewed. However, a different strategy for trying to understand how young children approach this problem is to consider how the executive function aspects of card sorting might be like theory of mind, given the two are empirically related. All of the characteristic 3 to 5year theory of mind problems involve a situation or object that can be seen in two different ways and, as Flavell (1988) has argued, the two views are in conflict with each other. The clearest example may be the familiar rocksponge task in which children must contend with a sponge that has been painted to resemble a rock (Gopnik & Astington, 1988). The two views or descriptions of the object, rock versus sponge, are always each associated with a different perspective in the theory of mind problems—other versus self in false belief, earlier self versus present self in representational change, and how it looks now versus what it really is now for appearance—reality. Of course, the pattern in each of these contrasts is for 3yearolds to adopt only one of these descriptions, usually what the object actually isbut note, for example, the occurrence of phenomenalism errors in the appearancereality task (Flavell, Flavell, & Green, 1983)—and give it for both perspectives. Fiveyearolds, in contrast, change their judgment depending on which perspective is being considered. It is easy to see that the rocksponge task reveals something important about the child's understanding because it involves the identity of the object. The child is being asked "What do you think it is?" and "What will your friend think it is?" In comparison, the card sorting task has more of the flavor of a mechanical procedure. The child merely has to get the cards in the right place. However, it is possible to imagine there is more involved in the card sort. When the task begins, the sorting procedure requires the children to make a judgment about some aspect of the object—for example, for color, they have to identify each picture as being yellow or blue. When the change in dimension is announced, the sorting procedure now requires a judgment about a different aspect of the card. They must judge whether the picture is a flower or a car. Even though they must judge shape, young children's sorting response, because they continue to use the old dimension, indicates a verdict of something like "yes, it is a yellow one." This mistake seems much like what happens when young children repeat the reply "it is a sponge" to the questions of what their friend will think the rock/sponge is, what they thought it was, and what it looks to be. The theory of mind and card sorting results can be given a similar reading because they have the same formal structure. Both involve something that can take two different descriptions—rock or sponge for the
Page 124
object in the theory of mind problems and a shape or a color for a given test card in the sorting task. In order to judge which description is appropriate, the choice has to be conditionalized on the additional determinant of the specific perspective in the theory of mind problems and the relevant dimension in the card sort. Fiveyear olds are able to incorporate the additional condition to choose between descriptions appropriately. Threeyearolds appear to have much more difficulty with the ambiguity of the descriptions. In their characteristic pattern of errors, they seem to ignore the additional condition and consistently adopt one description at the expense of the other. This analysis is not limited to these examples. The important features of other common theory of mind problems can be captured with the same terminology. For example, the original changeinlocation version of the false belief problem (Wimmer & Perner, 1983) can be accommodated by treating location as an aspect or part of the description of the hidden object. It may be more interesting to consider the approach's implications for executive function, however. The card sorting task assesses executive function because it involves action. The child is given the goal of sorting the cards on a particular dimension and must choose the actions that will fulfill that goal. Perhaps the ambiguity in descriptions that characterizes the card sort does not just apply to describing the cards themselves, but also extends to the underlying actions that are involved. There is a wellestablished recognition in philosophy that intentional actions take different descriptions. In Searle's (1983) memorable example, Gavrilo Princip's action can be variously described as "he pulled the trigger," "he shot the Archduke Franz Ferdinand," "he started World War I," "he disturbed a number of air molecules," and so on. Although all of these descriptions apply to the action (just as rock and sponge, yellow and flower all also apply in the previous examples), not all of them fit as a description of what Gavrilo Princip was trying to do. In other words, the action is intentional only under certain descriptions because the actor had only some of them as a goal, or could only reasonably be expected to anticipate some of them as an outcome. Gavrilo Princip was trying to shoot the Archduke, but he probably was not trying to start World War I or disturb air molecules. This example illustrates that there is always a potential ambiguity in the description of intentional actions. What if young children have trouble appreciating potentially conflicting descriptions of intentional action, just as they have trouble appreciating conflicting descriptions of what an object is thought to be? This problem would tend to have the effect of making young children think that an action has only one purpose in a given situation. Older children, in contrast, will appreciate that the same action can be used for different purposes in the same situation. In the card sort, this interpretation would imply that
Page 125
when young children have to change goals from matching to shape rather than color, their previously successful matching makes it so that they have difficulty seeing that sorting could be used to produce a different outcome. Fiveyearolds, on the other hand, recognize that sorting can lead to different outcomes in this situation, just as they see that objects can be identified differently, and they are able to reconstrue the action to meet the new goal. Another situation in which this approach to action can be informative is for deception. Deception requires the intentional action of misleading someone. The tests done with young children (e.g., Russell et al., 1991) typically ask the child to make a response like pointing to mislead someone to search in the wrong place for an object. Looking at it as an instance of intentional action, deception involves taking a normally helpful action (pointing usually has the goal of showing where something is) and putting it to the new purpose of sending someone to the wrong place. This conflict should be precisely what is problematic for young children if they have difficulty seeing that the same action can be put to different purposes within the same situation. Older children, in contrast, ought to be able to appreciate that the intentional action of pointing can be put to different uses in order to produce different outcomes. Reinterpreting some of young children's executive function difficulties in this fashion gives a different slant to the problems involved. It is not hard to imagine that theory of mind would be a challenge for young children given the amount of thinking that is demanded. Action, on the other hand, has the connotation, similar to the tendency to see the card sort as a mechanical procedure, of being something that is relatively straightforward, perhaps automatic—a response. However, the necessity for planning and problem solving becomes apparent when the relevance of intentional action to executive function is acknowledged (Zelazo, Carter, et al., 1997). Finding an action that can be put to a new use to produce an intended outcome can require every bit as much thought as inferring how someone else is looking at a situation, and may be subject to the same errors. Cognitive Complexity and Control (CCC) Theory The preceding discussion helps to show that there is an equivalence between theory of mind and executive function in preschool, and that the two may be similar because the difficulties young children have understanding ambiguity in descriptions of situations can also apply to intentional action. For this account to be useful, it is important to be able to model the source of these difficulties. The embedded rules in CCC theory provide
Page 126
one characterization of how 5yearolds are able to contend successfully with the identified ambiguities. Theory of mind and executive function, or inference of mental states and planning action, are related in that they depend on different applications of these same reasoning rules. Threeyearolds err in characteristic ways when inferring mental states and planning action because they rely on a simpler version of the rules. Embedded rules allow the solving of the preschool theory of mind problems because they govern the switching between ambiguous descriptions of the object according to setting conditions. So, for example, in the rocksponge false belief task, the object can be seen as either thing, but which is chosen for self is different from which has to be chosen for other. Embedded rules on the order of "if self, if judging what object is, then sponge" and "if other, if judging what object is, then rock" are needed. Embedding within the setting conditions of self and other is what makes it possible for the judgment about the object to produce conflicting answers. Although 3yearolds can make rulebased judgments, they are not yet able to embed them within another rule. Hence, when they are confronted with a problem as complex as false belief, they reduce the complexity by ignoring the setting condition and judge according to simple ifthen rules (see also Halford, Wilson, & Phillips, in press). This simplification results in the same judgment being given for self and other. The theory of mind problems are inference problems because children use the embedded rules to reason from the antecedents to the consequent. In false belief, they have to consider which person (antecedent) and the object or situation (antecedent) in order to infer the resulting mental state (consequent). The rules make it possible to be more explicit about the relation between theory of mind and executive function. In the executive function or action tasks that have been discussed, children must solve for a different term in the embedded rules. They are given the consequent, and one of the antecedents, but must solve for the other. Thus, in the card sort, they are given the consequent of having the cards match, the antecedent of the relevant dimension (e.g., shape), but they have to determine the action on the situation (put it here) that will produce the consequent. Similarly, in deception, the consequent of influencing the search of the other person is specified, and so is the fact that the other person is a competitor, but the child must formulate the action (e.g., point here) that will be effective. Although the embedded rule term being solved for is different in the inference versus action problems, 3yearolds' use of simple ifthen rules has the same detrimental effect. In the card sort, absent embedding, young children will ignore the setting condition of the new dimension and satisfy the consequent of matching the cards by selecting the (now incorrect) action that had previously succeeded. In deception, when the setting con
Page 127
dition that the other person is a competitor is ignored, the consequent of influencing the other's search will be accomplished by the usual rule of pointing where the object is. Just as in the inference problems, considering conflicting descriptions is not sustainable in the missing term of the action problems without setting conditions. It will not be possible to put a normally helpful action like pointing to the opposite purpose, unless that choice can be embedded under a setting condition to partition it from the way it is typically understood. Finally, the assumption that 3yearolds employ simple rules in these situations predicts that their performance will have a bias (Frye et al., 1995, p. 504; Frye, Zelazo, Brooks, & Samuels, 1996, p. 130; Zelazo & Frye, 1997, pp. 129, 141). The bias is a result of the use of the rules when they are no longer appropriate. It is assumed that any bias comes strictly from experience. The card sort is the most interesting case in this regard because it is designed so that the bias is established within the task. Threeyearolds can sort by either color or shape, but once they have been given experience with one, they are then predisposed to it in that situation. Children arrive at the theory of mind and deception tasks—and certain causal reasoning problems (Frye et al., 1996)—with biases that depend on preexperimental experience. They are accustomed to naming objects according to what they actually are and pointing to direct people to things. Thus, contrary to some readings (Russell, 1996, p. 273), the theory expects there to be a reality bias in theory of mind, and suggests that it will only be overcome when the setting conditions of embedded rules give a means of isolating exceptions to the way things usually occur. Empirical Implications of the Approach for Executive Function Embedded rules make explicit how theory of mind could be related to executive function. The two are distinct and neither underlies the other, yet in a sense they are two sides of the same thing, because they both rely on different use of the same set of rules. A clear implication is that theory of mind and executive function ought to be empirically related. Moreover, the relation should not be bounded by conceptual similarity (e.g., a linking of theory of mind and deception because they both involve false belief), but should extend to developments that rely on the same rules. The empirical relations that have been found between theory of mind and other executive function problems like the card sort establish this point. These problems are related to theory of mind, do not involve the understanding of mental states, and yet arguably both rely on a single set of rules. Indeed, Carlson's (1997) results show that the executive function tasks that would
Page 128
best fit the specification given here, those that present a response conflict, were the ones that were most strongly correlated with theory of mind. In addition to the empirical predictions it makes about the relation of theory of mind and executive function, the current approach also maintains some different expectations for the development of executive function. Gains in executive function in childhood have often been explained in terms of improvements in response inhibition (e.g., Carlson, Moses, & Hix, 1998; Dempster, 1992; Diamond & Gilbert, 1989; Harnishfeger & Bjorklund, 1993; Hughes, 1998; Russell et al., 1991; Welsh & Pennington, 1988). Response inhibition is undoubtedly important, but it may not be what is predominantly involved in the preschool developments under discussion. The heavier emphasis that CCC theory places on intentional action and planning suggests changes may occur because children become able to employ embedded rules to select the appropriate actions, rather than just increasing their ability to inhibit inappropriate ones. This distinction carries with it differences in empirical predictions for preschool executive function. Card Sort As was outlined earlier, young children's perseverative responding in the dimensional change card sort could be the result of inadequate response inhibition. Sorting the cards repeatedly by one dimension should produce prepotent responses. Difficulty inhibiting these responses would prevent executing the new ones when the other dimension came into effect. One test of this possibility is to determine whether extensive training with the preswitch rules is required to produce perseveration. If inhibition of prepotent responses is the problem, then the likelihood of perseveration ought to increase as a function of the number of preswitch trials. In contrast, if 3 yearolds are simply unable to represent a higher order rule for selecting between rule pairs, then they ought to perseverate even in the presence of minimal interference. Zelazo, Frye, and Rapus (1996, Experiment 2) found that a single successful sorting trial on the first dimension was sufficient to cause the majority of 3 yearolds to perseverate when the dimension changed, indicating that overlearning is not required. The response inhibition and CCC accounts also make differential predictions for the importance of the response itself. According to an inhibition approach, specific responses presumably become prepotent, whereas CCC theory would suggest that the lack of embedded rules could affect a variety of different responses. To explore this question, Zelazo et al. (1996, Experiments 3 & 4) investigated the importance of response modality for perseveration in the card sort. A new version of the task was presented in which children indicated their sorting responses verbally rather than manually. Classic inhibition approaches (e.g., Luria, 1961) have theorized that
Page 129
there is more flexibility in linguistic responses than motoric ones. Nonetheless, 3yearolds perseverated in this purely verbal version. The strategy of manipulating the response to evaluate the contribution of inhibition in the card sort has been taken one step further. Jacques, Zelazo, Kirkham, and Semcesen (in press) tested whether 3yearolds could master the card sorting task if they did not have to sort at all. To eliminate the need to execute a sorting response, children were simply asked to evaluate the sorting of a puppet. In one instance, the puppet perseverated on the postswitch trials. In another, the puppet switched to sorting correctly by the new dimension. Threeyearolds judged that the puppet was right when perseverating, and wrong when sorting in accordance with the new dimension. In addition, their own sorting on an independent test closely agreed with their appraisals of the puppet's performance. With the need to make and change the sorting response removed, it is hard to escape the impression that 3yearolds are thinking about the task in the wrong way, rather than being caught in the wrong response. If their lack of embedded rules leaves them thinking that matching on the first dimension is the correct action to take even after the change, then they will think the perseverative responses are correct whether they or another make them. Deception Carlson, Moses, and Hix's (in press) new data appear to favor a response inhibition explanation of the executive function difficulties in deception. Children were found to be worse at deception when an empty location had to be indicated by the familiar response of pointing, than when the more novel responses of a pictorial cue or pointer had to be used. The authors concluded that deception was more difficult with the familiar gesture because it had a prepotent response history, and hence more inhibition would be necessary to prevent its typical use. They reasoned that these results would not be consistent with the approach taken by CCC theory because it ought not to predict differences for these different responses. The present review of CCC theory indicates that the Carlson et al. (1998) findings may not distinguish the two explanations sharply. CCC theory depends on there being biases in conflict situations, and those biases are stipulated to come from experience, so familiar gestures should present more of a problem. Consistent with CCC theory, the age group comparison in Carlson et al. showed that although 3yearolds were better at deception with pictures compared to pointing, they were still significantly worse than 4yearolds on both. This finding suggests that prepotency of the pointing response cannot be the entire explanation of 3yearolds' difficulty with deception.
Page 130
Finally, Brooks et al. (1997) provided data that seem to favor CCC theory's approach over response inhibition. Brooks et al. tested a strategy judgment procedure for deception that bears some resemblance to the errorjudgment procedure Jacques et al. (1997) instituted for card sorting. Rather than having children produce their own deceptive strategy in a given situation, a deceptive and nondeceptive response was demonstrated, and the children merely had to choose the one that would be effective. Despite not having to make the response, 3yearolds were still not able to select the deceptive strategy reliably. This finding again argues that young children's problems with deception are not just in executing a response, but that they have difficulty choosing an action for a new purpose when it has previously been used for a conflicting one. Conclusion Sufficient empirical results now exist to press the question of the relation between theory of mind and executive function. The importance of intention for theory of mind is apparent. If there is any point to understanding belief and desire, it is to make sense of the others' intentions. The role of intention in preschool executive function has less often been recognized, presumably because of the prevailing emphasis on response inhibition. When executive function is cast in terms of intentional action, its relation to theory of mind becomes much more comprehensible. In particular, the characteristic errors children make are seen to be very similar. Young children have difficulty appreciating that conflicting descriptions can apply to objects and they have similar difficulty grasping their application to intentional actions. CCC theory offers a specific explanation for this correspondence in terms of how embedded rules are used in inferring mental states on one hand and formulating actions on the other. This explanation has the advantage of maintaining the distinctions between theory of mind and executive function while bringing them both under the same theory. Acknowledgment Thanks to P. D. Zelazo for his editorial comments. References Astington, J. W. (1991). Intention in the child's theory of mind. In D. Frye & C. Moore (Eds.), Children's theories of mind (pp. 157172). Hillsdale, NJ: Lawrence Erlbaum Associates.
Page 131
BaronCohen, S. (1994). How to build a baby than can read minds: Cognitive mechanisms in mindreading. Cahiers de Pyschologie Cognitive, 83, 513552. Brooks, P. J., Frye, D., & Samuels, M. C. (1997). The comprehension and production of deception for self and other. Manuscript in preparation. Carlson, S. M. (April, 1997). Individual differences in inhibitory control and children's theory of mind. Meetings of the Society for Research in Child Development, Washington, DC. Carlson, S. M., Moses, L. J., & Hix, H. R. (1998). The role of inhibitory processes in young children's difficulties with deception and false belief. Child Development, 69, 672691. Dempster, F. N. (1992). The rise and fall of the inhibitory mechanism: Toward a unified theory of cognitive development and aging. Developmental Review, 12, 45 75. Diamond, A., & Gilbert, J. (1989). Development as progressive inhibitory control of action: Retrieval of a contiguous object. Cognitive Development, 4, 223249. Flavell, J. H. (1988). The development of children's knowledge about the mind: From cognitive connections to mental representations. In J. W. Astington, P. L. Harris, & D. R. Olson (Eds.), Developing theories of mind (pp. 244267). New York: Cambridge University Press. Flavell, J. H., Flavell, E. R., & Green, F. L. (1983). Development of the appearancereality distinction. Cognitive Psychology, 15, 95120. Frye, D., Zelazo, P. D., & Brooks, P. J., & Samuels, M. C. (1996). Inference and action in early causal reasoning. Developmental Psychology, 32, 120131. Frye, D., Zelazo, P. D., & Palfai, T. (1995). Theory of mind and rulebased reasoning. Cognitive Development, 10, 483527. Gopnik, A., & Astington,J. (1988). Children's understanding of representational change and its relation to the understanding of false belief and the appearancereality distinction. Child Development, 59, 2637. Gopnik, A., & Wellman, H. (1994). The theory theory. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture (pp. 257293). New York: Cambridge University Press. Halford, G., Wilson, W. H., & Phillips, S. (in press). Processing capacity defined by relational complexity: Implications for comparative, developmental, and cognitive psychology. Behavioral and Brain Sciences. Harnishfeger, K. K, & Bjorklund, D. F. (1993). The ontogeny of inhibition mechanisms: A renewed approach to cognitive development. In R. Pasnak & M. Howe (Eds.), Emerging themes in cognitive development (pp. 2849). Lincoln: University of Nebraska Press. Hughes, C. (1998). Executive function in preschoolers: Links with theory of mind and verbal ability. British Journal of Developmental Psychology, 16, 233253. Jacques, S,. Zelazo, P. D., Kirkham, N. Z., & Semcesen, T. (in press). Rule selection versus rule execution in preschoolers: an errordetection approach. Developmental Psychology. Luria, A. R. (1961). The role of speech in the regulation of normal and abnormal behavior (J. Tizard, Trans.). New York: Pergamon. Ozonoff, S., & McEvoy, R. E. (1994). A longitudinal study of executive function and theory of mind development in autism. Development and Psychopathology, 6, 415431. Ozonoff, S., Pennington, B., & Rogers, S. (1991). Executive function deficits in highfunctioning autistic individuals: Relationship to theory of mind. Journal of Child Psychology and Psychiatry, 32, 10811105. Perner, J. (1991a). On representing that: The asymmetry between belief and desire in children's theory of mind. In D. Frye & C. Moore (Eds.), Children's theories of mind (pp. 115138). Hillsdale, NJ: Lawrence Erlbaum Associates. Perner, J. (1991b). Understanding the representational mind. Cambridge, MA: MIT Press. Perer, J. (1998). The metaintentional nature of executive functions and theory of mind. In P. Carruthers & J. Boucher (Eds.), Language and thought (pp. 270 283). Cambridge, England: Cambridge University Press.
Page 132
Russell, J. (1996). Agency. Its role in mental development. Hove, England: Erlbaum (UK) Taylor & Francis. Russell, J., Jarrold, C., & Potel, D. (1994). Executive factors in preschoolers' strategic deception. British Journal of Developmental Psychology, 9, 301314. Russell, J., Mauthner, N., Sharpe, S., & Tidswell, T. (1991). The 'windows task'. as a measure of strategic deception in preschoolers and autistic subjects. British Journal of Developmental Psychology, 9, 331350. Searle, J. R. (1983). Intentionality: An essay in philosophy of mind. Cambridge, England: Cambridge University Press. Welsh, M., & Pennington, B. (1988). Assessing frontal lobe function in children: Views from developmental psychology. Developmental Neuropsychology, 4, 199 230. Wimmer, H., & Perer, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception. Cognition, 13, 103128. Zelazo, P. D., Carter, A., Reznick, J. S., & Frye, D. (1997). Early development of executive function: A problemsolving framework. Review of General Psychology, 1, 198226. Zelazo, P. D., & Frye, D. (1997). Cognitive complexity and control: A theory of the development of deliberate reasoning and intentional action. In M. Stamenov (Ed.), language structure, discourse, and the access to consciousness (pp. 113153). Amsterdam & Philadelphia: John Benjamins. Zelazo, P. D., Frye, D. & Rapus, T. (1996). An agerelated dissociation between knowing rules and using them. Cognitive Development, 11, 3763. Zelazo, P. D., & Jacques, S. (1996). Children's rule use: Representation, reflection, and cognitive control. In R. Vasta (Ed.), Annals of child development (Vol. 12, pp. 119176). London: Jessica Kingsley Press.
Page 133
Chapter 8— Executive Functions and Theory of Mind: Cognitive Complexity or Functional Dependence? Josef Perner Sandra Stummer Birgit Lang University of Salzburg There is the intuition that better understanding of one's inner life will give better selfcontrol. On this intuition one expects that children's progress in developing a theory of mind should show itself in increased selfcontrol (Wimmer, 1989). Ten years ago there was just circumstantial evidence that theory of mind achievements at particular ages were paralleled by signs of greater selfcontrol (Perner, 1991). Now there is rapidly increasing evidence that theory of mind development, especially around the age of 4 years, is related to gains in executive control on socalled executive function tasks at this age (e.g., Carlson, 1997; Frye, Zelazo, & Palfai, 1995; Hughes, 1998, in press; Russell, Mauthner, Sharpe, & Tidswell, 1991; Russell, Jarrold, & Potel, 1994). This, however, cannot be interpreted as evidence for the original intuition because there are now also several alternative explanations for this developmental relation. In this chapter we evaluate some of these alternative accounts. The most frequently used measure on which these correlations are based is the false belief test. The standard version is the unexpected transfer story (Wimmer & Perner, 1983) in which a protagonist puts an object into location A. In his absence the object is unexpectedly transferred into location B, so that he mistakenly believes that the object is still in location A. To assess their understanding of the protagonist's belief, children are either asked where the protagonist thinks the object is, or where the protagonist will look for the object. The typical developmental trend is that at 3 years almost all children answer wrongly with actual location (B) of
Page 134
the object, whereas most children older than 4 years, and definitely a clear majority of 5yearolds, answer correctly with location A. Another version is the deceptive container task (Hogrefe, Wimmer, & Perner, 1986; Perner, Leekam, & Wimmer, 1987). For instance, children are shown a Smarties® container and asked about its content. Practically all children say, ''Smarties." They are then shown that it contains a pencil. With the pencil back in the closed container they are asked what another person, who has not yet seen what is inside, would think (or say) was inside. A small but important twist on this procedure turns the task into the memory for earlier false belief test (Gopnik & Astington, 1988). The child is simply asked what he or she herself had thought was in the box when first shown. Finally, Flavell, Flavell, and Green (1983) pioneered the appearancereality test. For instance, children are shown a trick object that they identify reliably as a rock. Manual manipulation reveals that it is just sponge (that looks like a rock). When 3yearolds are asked the reality question about what the object really is they tend to answer "a sponge," and when asked about what the object looks like they tend to answer wrongly with "a sponge." The executive function tasks on which children around this age markedly improve typically require the inhibition of some response strategy or cognitive set. The tasks repeatedly used to establish the link with theory of mind are a version of Luria's (Luria, Pribram, & Homskaya, 1964) hand game in which the child has to first repeat two hand gestures (e.g., fist or flat hand on table) shown by the experimenter, that is, fist/fistflat/flat, and then switch to doing the opposite gesture as the experimenter, that is, fist/flatflat/fist. Another paradigm is the daynight Stroop task (Gerstadt, Hong, & Diamond, 1994) where children have to say "day" when the dark picture with the moon is shown, and "night" when the bright, sunlit scene is shown. The most frequently used task is a modification of the Wisconsin Card Sorting Test (WCST; used to diagnose executive dysfunction in frontal lobe patients) by Frye et al. (1995). Children are given two target cards, for instance, a green car and a yellow flower, and they have to place a pack of 5 cards showing either a yellow car or a green flower according to one of two pairs of rules. For instance, if the color rules are used then the yellow car has to be placed with the yellow flower target and the green flower with the green car target. Children above the age of 3 years usually have no problem following these instructions. Their executive problems emerge when the rules change and they are instructed to sort by shape, that is, cars to car and flowers to flower despite their different colors (see Frye, chap. 7, this volume; see Zelazo, chap. 6, this volume). Several studies (e.g., Carlson, 1997; Frye et al., 1995; Hughes, 1998, in press; see Perner & Lang, in press, for review) have shown that the ability to master these executive function (EF) tasks emerges at the same time as children pass the mentioned theory of mind (ToM) tasks. This is re
Page 135
flected in correlations between EF tasks and ToM tasks in the age range of 3 to 6 years of r = .30 to .60. And even after partialling out age, these correlations tend to remain positive between .15 to .50. We look at two classes of explanations for this relation. One type of explanation assumes a common logical structure underlying the relevant ToM and EF tasks that demands a certain cognitive capacity (cognitive complexity and control: CCC theory). The other class sees a functional relation between theory of mind and executive function. The CCC Theory Zelazo and Frye (1997) suggested that children's control problems are a consequence of the cognitive complexity posed by the executive function tasks in terms of doubly embedded conditional action rules. Moreover, the same complex rules are required for some theory of mind tasks like the false belief task. For this reason children pass all these tasks at the same age. All these tasks are characterized by conditional relations from different antecedents (a1) to consequents (c1) that change according to setting condition (sk). The classic task on which to illustrate this analysis is the dimensional change card sort (DCCS) used by, for instance, Frye et al. (1995). In this task, in one setting (s1: sort by color) the green car (a1) goes into the box marked with a green flower (c1) and the red flower (a2) into the box with the red car (c2), whereas under the other setting (s2: sort by shape) the green car (al) goes into the box marked with a red car (c2) and the red flower (a2) into the box with the green flower (c1). The task can be schematically rendered thus: IF color game (sl)
THEN IF green car (a1)
THEN put it to green flower (cl),
IF red flower (a2)
THEN put it to red car (c2);
IF shape game (s2)
THEN IF green car (a1)
THEN put it to red car (c2),
IF red flower (a2)
THEN put it to green flower (c1).
A similar task used by Frye et al. (1995) involved a covered ramp and a ball that could be put at the top into either a left or right hole from where it then rolled down into one of two boxes (left or right). At one setting (s1), visibly marked by a light turned on, the ramp connected upper left hole (a1) with lower left box (c1) and upper right hole (a2) with lower right box (c2), that is, leftleft, rightright. At the other setting (s2: light off), the connections were crossed over, that is, leftright, rightleft. Children were shown the mechanism and then had to predict in which box
Page 136
the ball could be found. To solve this task the same cognitive complexity in terms of understanding embedded conditionals is required. These two tasks evidently also qualify as executive function tasks (the DCCS is in fact a modification of the WCST often used on frontal lobe patients) because the antecedent conditionconsequent action links (a1cl, a2c2) established under setting s1 interfere with the links required under s2, that is, a1c2, a2c1. Hence the association established under one setting leading from a1 to c1 has to be monitored and inhibited under the other setting. Frye et al. (1995) argued that the false belief task itself requires reasoning with embedded conditionals and, therefore, it is the developing ability to engage in such reasoning that underlies the developmental link between false belief test performance and performance on executive function tasks like card sorting and the ramp task. Frye (in press) has given the clearest analysis for the traditional false belief task, where the chocolate is transferred from here to there and the child is to predict where Maxi will go for his chocolate: IF me (s1)
THEN IF looking for chocolate (a1)
THEN there (c1),
IF .. . (a2)
THEN. . . (c2);
IF Maxi (s2)
THEN IF looking for chocolate (a1)
THEN here (c2),
IF . . . (a2)
THEN . . . (c1).
The a2 condition has been left out, but could be easily filled in with the assumption that two objects, chocolate and something else, are being switched from here to there and vice versa. This is not the crucial difficulty with this analysis. There are three other nontrivial difficulties: (a) the analysis does not capture the rules by which the task is solved, (b) the analysis seems often arbitrary (too many free parameters), and (c) the application to the false belief case also fits other tasks on which children excel at a much earlier age. A Priori Problems With the CCC Analysis 1. What Does the Analysis Capture? The ififthen structure in the DCCS and the ramp problems clearly describes the rules that one has to know in order to solve the task. In the case of the false belief task the stated conditionals surely cannot be the rules that one brings to bear on the task. I cannot have a rule "If Maxi, then if looking for chocolate, then here." Such a rule seems possible only if children were given repeated exposures to the same behavioral sequence of Maxi going to the empty cupboard.
Page 137
Only then could they form such a rule that captures such a specific behavioral regularity. The belief task, however, is to capture a deeper understanding that allows behavioral predictions without knowledge of behavioral regularities in specific situations. That is, the child who understands belief has to apply knowledge of the sort, "People look for objects where they think the object is." Therefore, the relevant rules to apply to the false belief task would be: IF asked where I will look for the chocolate, THEN give that location where I think it is; IF asked where Maxi will look for the chocolate, THEN give that location where he thinks it is. This constitutes only a simple pair of rules. Evidently more reasoning is required. Whether the spelling out of this further reasoning results in the required embedded conditional structure remains to be shown. 2. Arbitrariness The application of the ififthen analysis to the false belief task highlights a wider problem with this kind of analysis, namely that it suffers from a dangerous arbitrariness in how one chooses to describe the problem. For instance, it is unclear why one could not propose the following analysis: IF I am looking for the chocolate (al) THEN there (cl); IF Maxi is looking for the chocolate (a2) THEN here (c2). Put in this more natural way, the task reduces to simple conditionals, which ought to be within the ability of children one year younger, that is, 3yearolds (Zelazo & Frye, 1997), and should be noticeably easier than the false belief task. No principled rule has been stated for why Frye's (in press) analysis of the false belief task is to be preferred over this alternative analysis. 3. Conflicting Data Even if principles of analysis could be specified, we suspect that its application would lead to wrong predictions on other tasks. For instance, Repacholi and Gopnik (1997) showed that as early as 18 months, but not before, infants understand divergent food preferences (see also Meltzoff, Gopnik, & Repacholi, chap. 2, this volume). An ififthen analysis closely modeled on Frye's (in press) analysis of the false belief problem can be given of this task:
THEN IF wanting to eat (a1)
THEN crackers (cl),
IF ... (a2)
THEN . . . (c2);
IF me (sl)
Page 138
IF Other (s2)
THEN IF wanting to eat (a1)
THEN broccoli (c2),
IF . . . (a2)
THEN . . . (c1).
Because children are able to respond correctly at the age of 18 months, either the task analysis must be flawed or there is something seriously wrong with Frye et al.'s (1995) developmental claim that the ability to reason with embedded conditionals emerges at the age of 4 years (the time they typically master the false belief task). Another task that young children tend to be fairly good at and which is not only in Frye's analysis but also on intuitive grounds very similar to the false belief problems is children's understanding of pretense. In particular, Harris and Kavanaugh (1993) and Leslie (1994) investigated children's ability to draw consequences from pretend assumptions. For instance, teddy bear pretends that there is tea in the cup. Teddy pours the tea over a piece of chocolate. Children are asked what teddy did, and then whether the chocolate is now wet or dry. Harris and Kavanaugh (1993) found that at the age of about 2 1/2 years there was marked improvement in answering that the chocolate is now wet. An analysis in terms of doubly embedded conditionals suggests itself. Reality and pretense are two different setting conditions and within each the same conditional antecedent (a1: tilt the cup) leads to different consequences: chocolate being dry (c1) in reality and being wet (c2) in pretense. As in the false belief scenario, reality is opposed to something that isn't true (false belief, pretense); hence one would expect children of up to the age of 4 years to respond in terms of reality that the chocolate is still dry. Again, CCC theory doesn't provide a clear explanation for why children respond according to the pretend assumptions over a year earlier than they are able to respond according to someone's false belief. Empirical Tests of CCC Theory Apart from these a priori problems in applying CCC theory, we also tried to assess it with empirical tests. Perner, Lang, and Stummer (1998) tested 57 children on 4 relevant tasks: a false belief task; the DCCS task used by Frye et al. (1995); Luria's hand game (after mimicking the experimenter's sequence of flat handfistflat hand, etc., the children were instructed to always produce the opposite hand shape, i.e., make a fist when the experimenter puts the flat hand down); and a test of understanding the involuntary nature of the kneejerk reflex. This last task is particularly relevant. Shultz, Wells, and Sarda (1980) reported that 3yearolds, who are otherwise quite proficient in judging accidents as unintentional, had a hard time realizing that their reflexive knee movements are involuntary. Perner (1991, pp. 217 220) speculated that the onset of this realization should relate to
Page 139
children's mastery of false belief because both tasks require an understanding of mental states as causally effective representational states: The belief makes the actor look in the wrong place even though he doesn't want to look in an empty place (unlike in pretend play) and in the case of the kneejerk reflex, the knee moves in the absence of an intention to move it. When scoring children as passing if they performed perfectly (5 of 5 trials) on the EF tasks (DCCS and hand game), then all four tasks were of comparable difficulty: between 52% and 59% of children passed. Table 8.1 presents an analysis of how much variance of false belief performance is explained by the various other tasks. The first two rows of Table 8.1 show that all these tasks are strong predictors of performance on the false belief task: Age and verbal intelligence (the KABC test) explain 1/4 of the variance. The two EF tasks both explain highly significant amounts of variance. The kneejerk reflex test, however, explains a good half of all the variance—considerably more than the EF tasks. The second row shows that after introducing age and verbal intelligence into the model, each of the two EF tasks and the kneejerk reflex still explain a significant amount of additional variance. The third row brings out an interesting asymmetry. When the kneejerk reflex is introduced into the model, the additional variance explained by the EF tasks sinks to below 10%, whereas if the two EF tasks are introduced first (row 6), understanding the kneejerk reflex still explains a respectable 14% of the variance of false belief test performance. Under the assumption that the kneejerk reflex task does not contain any obvious conditional reasoning requirements, it is likely that a conceptual advance in understanding the causal significance of mental states is reflected in the relation between false belief task and kneejerk reflex task. This understanding (or whatever else might make these tasks so similar in difficulty) accounts for 30% of the false belief variance and most of the covariance of false belief with EF tasks (after age and verbal intelligence). TABLE 8.1 Percent of Variance of False Belief Performance Explained by Last Factor in Model (SPSS Logistic Regression With Nagelkerke's R2)
Initial Factors in Model
Age + KABC
KR
DCCS
HG
25.5**
55.8**
41.4**
20.3**
Age + KABC
—
30.4**
21.6**
12.3*
Age + KABC + KneeReflex
—
—
8.4*
5.2*
Age + KABC + DCCS
—
17.0**
—
6.5*
Age + KABC + HandGame
—
23.3**
15.8*
—
Age + KABC + HG + DCCS
—
14.0**
—
—
None
Note. Data from Perner, Lang, and Stummer (1998). *p < .05. **p < .01.
Last Factor Added to Model
Page 140
Less than 10% might be attributable to problems of conditional reasoning that the kneejerk reflex task cannot account for. Moreover, even if one allows for the possibility that the kneejerk reflex task also requires embedded conditionals, this pattern of results still suggests that the common variance between false belief task and kneejerk reflex task cannot be explained with a common conditional reasoning structure. Why would this relation survive when all variance due to conditional reasoning structure (DCCS and hand game) has been accounted for? These data, of course, do not rule out that mastery of embedded conditionals is a prerequisite for the false belief task and the executive function tasks. It only shows that a particular conceptual advance underlies the false belief task, the kneejerk reflex task, and the EF tasks, and can account for more of the common variance than the CCC theory. One should also point out that the kneejerkreflex task, too, may require the conditional reasoning structure posited by CCC theory (P. D. Zelazo, personal communication, July 11, 1998). To see this, one needs to think of the setting conditions as how bodily movements of actions are normally linked to intentions (s1) and how they are linked in this special case (s2) of a reflex movement. Hence we can see the same ififthen structure: IF normal case (sl)
THEN IF I intend to move leg (a1)
THEN leg moves (c1),
IF I do not intend to move
THEN leg doesn't move (c2);
leg (a2)
IF reflex involved (s2)
THEN IF . . . (a1)
THEN . . . (c2),
IF I do not intend to move
THEN leg moves (c1).
leg (a2)
This proposal suffers from the problems outlined above. It is hard to see how this rule structure can underlie the child's understanding of the problem. At best it can be the result of an understanding of reflexes. Also, there are conflicting data. Shultz et al. (1980) showed that children were quite able to differentiate intended from nonintended actions when they could compare what they were instructed to achieve with what they actually achieved. For instance, if when interlacing one's fingers one is instructed to move a particular finger, then one tends to make errors. Threeyearolds, who all misjudged their kneereflex movement as intentional, judged the movement of the wrong finger correctly as unintentional. However, according to the knee reflex analysis, they should not be able to do this. If you substitute "that finger" for "leg," the normal case (s1) makes the child judge movement of that finger as intended. One can redescribe the problem in terms of whether the finger movement matches or mismatches the instructions so that the required complexity fits
Page 141
the observed development (i.e., so that the required structure involves only simple conditionals, which should be mastered at the age of 3 years): IF finger moved matches (a1)
THEN I meant to move it (c1);
IF finger moved mismatches (a2)
THEN I didn't mean to move it (c2).
This, however, raises the question of arbitrariness. A priori rules are needed for problem analysis. Judging by the example of the knee reflex and finger movement, an embedded conditionals analysis has to interact with what the child knows about the domain of knowledge in question, for instance, their theory of mind development. There is another set of data, originally carried out from a quite different vantage point, that speaks to the issue of theory of mind, executive function, and CCC theory. Perner and Stummer (1998) followed up a paradigm used by Doherty and Perner (1998) to test the theory that understanding false belief and understanding synonyms relate developmentally because both tasks require an understanding of representation. In the "production" version of the synonyms paradigm, children are first tested for their knowledge of synonyms; for instance, they are tested to see if they identify the correct referent (female person) under the description "woman" as well as under "lady." Then a puppet names the item using one of the synonyms and children are instructed to use the other name for it. In the test phase assessing children's grasp of these instructions, younger children tend simply to repeat what the puppet had said. The ability to use the synonym emerged with the ability to pass the false belief task. The two tasks correlated strongly, r = .71 in one experiment, and r = .65 in a second experiment. Even after verbal intelligence, age, and other control variables had been partialled out, these coefficients stayed high, .68 and .60, respectively. After varying the material of this task from synonyms to superordinatebasic categories (e.g., animaldog), we realized that the original synonyms task and these new variants, which can be called "saysomethingdifferent" tasks, can also be described as executive function tasks and as tasks requiring the ififthen structure suggested by Frye for the false belief task. For instance: IF puppet says "lady" (sl)
THEN IF asked to name picture (a1)
THEN say "woman" (c1),
IF . . . (a2)
THEN . . . (c2);
IF puppet says "woman" (s2)
THEN IF asked to name picture (al)
THEN say "lady" (c1),
IF . . . (a2)
THEN . . . (cl).
The tasks can also require executive control on the following grounds. We can assume that there is a wellestablished questionanswering schema that, once triggered by the question, looks automatically for a suitable
Page 142
answer. If there are several possible answers, it will produce the one that first comes to mind. For instance, if asked what this (female person) is, the answer will be more likely "a woman" than "a lady," because woman is the more common expression and is thus more prevalent in the child's mind. This prevalence can be changed if the picture is labeled by the experimenter "a lady" just before. Then the answer schema is more likely to encounter "lady" as a first suitable answer. The saysomething different task requires executive control in order to modulate the response given by the automatic answering schema, because the answer that the schema would choose by itself is not to be given, but the alternative. To test among alternative theories, we introduced further variations of the saysomethingdifferent task besides synonyms (ladywoman) and categories (animaldog). In the color/color task, the puppet named one of the two colors of an object and the child had to name the other. In the color/name task, if the puppet named the color of an object, the child had to say what the object was or vice versa. Table 8.2 summarizes children's performance on the false belief task and these four say somethingdifferent tasks. Across all four experiments, performance on three of the four saysomethingdifferent tasks was very similar to performance on the false belief task in terms of percentage of children passing each task (upper panel) and in terms of correlation with the false belief task (lower panel). The one task that did not conform to this pattern was the color/name TABLE 8.2 Results From Four Experiments by Perner and Stummer (1998)
Experiment, Age Range and Sample Size
Experiment 1
Tasks
Experiment Experiment 2 3
Experiment 4
3;06;1
3;44;9
2;95;7
3;15;0
(n = 36)
(n = 40)
(n = 30)
(n = 39)
50
54
% Children Correct on All Items False belief (FB) Saysomethingdifferent
61
73
Synonyms
67
—
53
—
Categories
64
73
—
62
Color/color
—
—
50
64
Color/name
94
95
83
90
Correlations FB × synonyms
.64
—
.93
—
FB × categories
.59
.77
—
.74
FB × color/color
—
—
1.00
.70
FB × color/name
.38
.37
.39
.36
*p < .05. **p < .01.
Page 143
task because it was mastered by even most of the young 3yearolds and performance by all children was close to ceiling (above 90%) when performance on the other tasks hovered around 60%. The good performance on the color/name task poses a problem for CCC theory because the conditional structure for the other saysomethingdifferent tasks seems to be the same—in Frye's (in press) analysis—as the conditional structure for the false belief task. If requested to say something (antecedent), the answer (consequent) differs depending on the setting condition (what the puppet chooses to say). It is important to see that in all these tasks the stimulus prompt for the child remained the same in both setting conditions: After the puppet gave its answer, the child was prompted with "Now say the other thing!" That is, in the case of a yellow bird: IF puppet says "yellow" (s1)
THEN IF asked to say other thing (a1)
THEN say "bird" (c1),
IF. . . (a2)
THEN . . . (c2);
IF puppet says "bird" (s2)
THEN IF asked to say other thing (a1)
THEN say "yellow" (c2),
IF . . . (a2)
THEN . . . . (c1).
The only difference in the color/color tasks was that the prompt in both cases was "Say the other color!" And yet, the change in task had a large impact on task difficulty. The executive function account of the original synonyms task outlined previously can account for this finding. The color/name task can be easy according to EF theory because the answer highlighted by puppet (e.g., "It's black") is not an admissible response for the child who is asked what kind it is or to name it. In other words, there is no automatic answering routine triggered that needs executive control to prevent it delivering the first suitable answer that comes to mind. In sum, we have highlighted several theoretical and empirical problems with the CCC theory as an account for why executive function tasks are mastered at the same age as false belief tasks and why they are related to theory of mind tasks. We now discuss briefly three theories that attribute the observed developmental link to functional dependencies between theory of mind and executive control. Functional Dependency between Theory of Mind and Executive Function Executive Components in Theory of Mind Tasks Russell et al. (1991) and Hughes and Russell (1993) suggested that the typical theory of mind tasks used at the age of four years contain an
Page 144
executive component. Although there are direct data suggesting that this is indeed the case for deception as a measure of understanding false belief (Carlson, Moses, & Hix, 1998), where the natural tendency of pointing to the object's actual location has to be inhibited, the claim that similar inhibition is required for the false belief task has received little empirical confirmation (see Perner, 1995, and Perner & Lang, in press, for more extensive discussion). For instance, one line of evidence supporting this claim was that the explanation task by Bartsch and Wellman (1989), which does not require inhibition of a natural answering tendency, is easier than the traditional prediction task. However, contrary to these authors' original claims, there is little evidence for this claim once methodological problems and ambiguity in interpretation of children's answers are cleared up (Wimmer & Mayringer, 1998). Most relevantly for present purposes, Hughes (1998) also found no difference in difficulty between prediction and explanation tasks and found that the prediction task (which lacks the alleged executive component of the prediction task) correlated with a measure of inhibitory ability as strongly as the prediction task. Executive Function as a Prerequisite for Developing a Theory of Mind Russell (1996) argued that action monitoring is a prerequisite for a rudimentary selfawareness, which in turn is a necessary requirement for building a theory of mind. Pacherie (1998) has made a similar proposal emphasizing the role of motor images (à la Jeannerod, 1997). These proposals leave it open why specific theory of mind tasks (i.e., false belief tasks) are expected to be mastered at the same age as specific executive function tasks (those requiring inhibition of existing response tendencies). However, the general message is clear: Executive control is a prerequisite for theory of mind; that is, without executive control there cannot be a theory of mind. Theory of Mind as an Integral Part of Executive Control Perner (1998) has argued that the essence of executive control is to direct (excitatory or inhibitory) control to action schemata via a specification of what these schemata are to achieve (their representational content). This distinguishes it from automatic control, or in Norman and Shallice's (1986) terms, contention scheduling, which does not require control via representational content. In order to enable control by content one has to represent the content of the schemata as something desired or undesired. Because such representation involves concepts such as desire, it involves a rudimentary theory of mind. The argument continues: To understand
Page 145
the need for inhibiting unwanted interference from existing schemata, the child has to understand not only that particular actions are unwanted (in that case, do nothing—do not want to act that way), but that there are existing schemata that make one act in that unwanted way unless one actively inhibits those actions. That makes clear why for some unwanted actions one has to actively inhibit them and not just refrain from initiating these actions. Now, to understand the existence of schemata as entities that make you act a certain way is to understand them as representations (representational vehicles) with causal powers. This understanding can, therefore, be dubbed metarepresentational Because inhibitory executive control is required for those executive function tasks that pose a problem for children up to 4 or 5 years, we can understand why they correlate with the false belief task. Perner (1991) has argued that understanding internal states as representations with causal power is what is required for understanding falsebeliefbased action; that is, the false belief makes one look in the wrong place even though that action does not serve one's goal and one does not want to look in an empty place. It also helps one understand the good correlation with the kneejerk reflex task, because that task requires understanding that reflex movements differ from intentional action by their lack of an internal representation (intention) that initiates one's movement. A central feature of this theory is that inhibitory control is not achieved until the causal/representational nature of mental states is understood. As stated, this is too simple. As Zelazo, Carter, Reznick, and Frye (1997) point out, inhibitory abilities develop at different ages and one should resist identifying executive functioning with a general inhibitory ability. Indeed, what should—according to our theory—develop around 4 years of age is not a general ability to inhibit but executive inhibition, that is, inhibitory control directed at action schemata on the basis of their representational content. Critically important questions, therefore, are which tasks require executive inhibition and when is automatic inhibition (contention scheduling) sufficient. Executive Inhibition Automatic inhibition is a natural consequence of competing action schemata. That is, a more highly activated schema inhibits less activated competitors. This can explain subsidence of the A error (searching for an object where it had previously been hidden rather than where it has been visibly put on this trial) in the AnotB task in infancy. Although Diamond (1991) claimed that this is due to improved inhibition of the prepotent response of searching in location A, this is not obvious. The Aerror does not occur immediately but only when the infant is restrained from searching correctly under B. Moreover, Harris (1989) pointed out how shifting attention from the B to the A location can explain
Page 146
the findings in a variety of different experimental conditions. Thus, the developmentally increasing resistance to committing Aerrors with longer and longer delays may reflect an increase in the ability to sustain attention on the B location (executive control in terms of boosting desirable action schemata), which then automatically inhibits search in A, rather than the onset of executive inhibition of the undesirable search in A. As executive control of boosting desirable action schemata increases, children become increasingly able to follow verbal instructions (Zelazo & Jacques, 1997). In particular by the age of 3 years children are able to follow a pair of rules (Zelazo & Reznick, 1991): ''If it is something found inside the house, put it into this box; if it is found outside the house, put it into that box." After sorting, for instance, an indoors item correctly into this box, one needs to inhibit the tendency to put the next item into the same box. However, this inhibition can be achieved automatically by the desired schema of putting the outdoors item into that box. No executive inhibition is required. Now consider Luria's two light go/nogo task (see Zelazo &Jacques, 1997, pp. 129137), which structurally also requires but a pair of rules: "If the red light comes on, press the bulb; if the green light goes on, do not press." One important difference is that only one response schema is defined ("if red—press"). The other rule does not establish a competing action schema but designates the lack of a schema for the other stimulus. So in this case, when the green light comes on, because it is a light it will activate to some degree the positive action schema "if red—press," but it will not activate a competing action schema (e.g., "if green—do X"), which would then automatically inhibit the undesirable action schema "if red—press." Because there is no automatic inhibition, executive inhibition is needed. The original Russian studies and Western replications in the 1960s and 1970s do show that the tendency to press on the wrong stimulus declines between 3 and 5 years (see review by Zelazo & Jacques, 1997). This analysis raises the question why the ififthen tasks used by Frye et al. (1995) need executive inhibition as claimed by the theory. In order to answer this question, we need to look briefly into the logical structure of these tasks. According to Zelazo and Frye (1997), these tasks are characterized by their logical structure as embedded conditionals: "IF s THEN (IF a THEN c)." On standard propositional calculus this is logically equivalent (same truth function) to "IF (s AND a) THEN c." Because AND is commutative and because it can be turned back into an embedded conditional, the original embedded conditional rule is equivalent to "IF a THEN (IF s THEN c)." In other words, there are no logical grounds for designating one kind of condition as setting s and the other as antecedent a. Additional criteria for this designation are required. Intuitively, the difference is that the setting conditions are in the background whereas
Page 147
the antecedents are in the foreground. They are the conditions that currently vary and to which prime attention is drawn. For instance, in the case of the ramp task used by Frye et al. (1995), the information of whether the ramp connects straight down or crosses over is the background set for several trials, whereas the choice of input hole (left or right) varies and is brought prominently into the child's attention. As a result, these kinds of tasks pose the following problem: They highlight stimuli (e.g., input hole) that trigger two incompatible action schemata, the wrong one better established through previous use and the correct one depending exclusively on recent verbal instructions. The additional stimulus information that is required to trigger the correct schema has been backgrounded. Hence the conditions for automatic inhibition of the previously used schema are insufficient. Additional, executive inhibition is needed.1 This analysis makes clear that executive inhibition is not used so much for replacing a previous action by an alternative action; rather, it is used when there is either no clear alternative action yielding automatic inhibition of the original action (see Luria's two light go/nogo task) or when the critical trigger for the new action has been backgrounded. A good illustration of this can be found in the finding by Frye, Zelazo, Brooks, and Samuels (1996) with the ramp task. When, for instance, only the left input hole is used, 3yearolds can inhibit the natural tendency of predicting that the marble will roll straight down to the left output hole. That is, when the light indicates a crossover connection, they predict correctly that the marble is to be found in the right output hole. In this case, however, the light indicating a straightdown or a crossover connection becomes the direct focus and the input hole (due to no variation) becomes completely backgrounded. No executive inhibition is needed to suppress backgrounded stimulus information. The cases discussed previously concern inhibition in order to prevent stimuli from triggering unwanted schemata. A somewhat different case obtains when an existing schema is being triggered but its execution has to be influenced. A case in point may be deception. The question, "Where is X?" triggers a questionanswering schema that, when left to its own devices, will produce a veridical verbal answer or a pointing to where X 1
We have so far emphasized executive inhibition for which an understanding of mental functioning (the causal/representational nature of mental states) is required. This analysis of the ififthen tasks presents another way in which this understanding might help. Instead of having to understand that an existing interfering schema needs to be inhibited, it may be equally helpful to understand something about how stimuli trigger and boost the desired action schemata. In this particular case, it is helpful to understand that backgrounded stimuli do not play as much a role in triggering and boosting the desired schema as is required. This understanding makes it possible to put special effort into highlighting (e.g., consciously rehearsing) the backgrounded information.
Page 148
actually is. In order to deceive, one has to give an answer but has to change that answer from the usual veridical answer. Carlson, Moses, and Hix (1998), Moore, Barresi, and Thompson (1998), and Russell et al. (1991) report that this may be indeed a serious problem for children younger than 4 years (but see Samuels, Brooks, & Frye, 1996, for a much weaker trend around this age). The need of executive control for steering triggered schemata is also indicated in the earlier mentioned finding by Perner and Stummer (1998) that 3yearolds find it difficult to provide alternative answers to the same question (e.g., color/color task) but have no difficulty answering an alternative question (color/name task). A final question that needs to be raised in this context is why executive inhibition is not required for the personal preference task used by Repacholi and Gopnik (1997) with 18monthold babies. Earlier in the chapter, we used the early mastery of this task to argue against the suggestion by Frye et al. (1995) that ififthen tasks (together with the false belief task) are not mastered before the age of 4 years. Now we argue that ififthen tasks do require executive inhibition, which is not acquired before the false belief task is mastered around 4 years of age. The solution is that the Repacholi and Gopnik task is not an ififthen task if one assumes that children's understanding of people's preferences changes at 18 months. The younger child may have an action rule based on "objective goodness": "If asked by a person to hand her food (a1), then hand her the goodtasting one (c1)." If by experience the child found out that that particular person wants to be handed the bad tasting food, then different setting conditions for self (and other normal people) and for this aberrant person would have to be introduced. The natural interpretation of the data, however, is that children acquire a new understanding of subjective preference. With this their action rule changes: ''If asked by a person to hand her food (a1), then hand her the one she likes (c1)." With this change in understanding the task by Repacholi and Gopnik does not require embedded conditionals, does not require executive inhibition, and consequently can—according to theory—be mastered by children well before they pass the false belief task. With this lengthy discussion of executive inhibition, we have shown that the inhibition required by those EF tasks that are mastered around the age of 4 years is executive inhibition, or at least that these tasks are more seriously dependent on executive inhibition than tasks mastered earlier. Evaluating the Functional Relatedness Theories Perner and Lang (in press) reviewed much literature from normal and abnormal development and found that most of the available data are compatible with these last two theories. There is some tentative suggestion
Page 149
from data on children with Williams and PraderWilli syndrome that some of these children consistently fail executive function tasks but pass false belief tasks (Tager Flusberg, Sullivan, & Boshart, 1997). As it stands, the interpretation of these data hinges on how pass criteria for EF tasks are set. If, however, the existence of such cases showing intact theory of mind in the absence of executive control can be confirmed, then this would pose a serious problem for the position that executive function is a prerequisite for a theory of mind. One very recent study by Hughes (in press) appeared to provide difficult data for the position that theory of mind is an integral part of executive functioning. A group of 50 children was tested at the average age of 3 years 7 months and again at the age of 5 years on several executive function tasks (e.g., Luria's hand game) and several false belief tasks. The finding of interest was that performance on the EF tasks at time 1 correlated with performance on the false belief tasks at time 2 (e.g., for the hand game, r = .41) somewhat more strongly than false belief at time 1 correlated with EF tasks at time 2 (e.g., for the hand game, r = .26). For the other executive function tasks there was the same asymmetry indicating that executive function is primary to theory of mind because early variations in executive control predict later theory of mind competence but not the other way around. This asymmetry would only speak against the position that theory of mind is an integral part of executive functioning under the assumption that ToM tests provide privileged access to a child's theory of mind. However, that is not so. They only demonstrate the existence of certain ToM capabilities. It remains an open question whether the methods used in ToM tasks (e.g., prediction of how people—including oneself—would act) tap the existence of ToM more directly and efficiently than executive functioning (provided a ToM is indeed an integral part of such functioning). In fact, if theory of mind does not just mature as a functional module (as Leslie, 1987, 1994, suggested) but needs constant exercising to develop further, then control of one's own action (inhibition of intruding action sequences) is by far the more important exercise ground for building a representational/causal understanding of mind than trying to understand other people's and one's own false beliefs. Although, on this view, it is in principle possible to develop a theory of mind without executive control in some pathological cases, in normal development the theory of mind capacity may show itself in executive control before it emerges in socalled theory of mind tests. There is also another reason for this timing. Clements and Perner (1994, 1997) showed that an "implicit" understanding of false belief, which shows itself in looking behavior and in spontaneous actions (moving a mat to where the mistaken character will appear to search for an object), precedes by about a year explicit understanding in terms of answering questions as to where the
Page 150
character will appear. If this earlier implicit understanding of another person's false belief can guide action, then similar implicit understanding of one's own action tendencies might provide an earlier form of executive control. Conclusion We looked at different explanations for the finding that theory of mind development around the age of 4 years relates to gaining executive control on tasks that require inhibition of interfering response tendencies. CCC theory explains this developmental relation by a common cognitive complexity required for executive control and understanding false belief. Our conclusion is that although the analysis in terms of embedded conditionals may capture some commonality across these different domains, the rules of applying the analysis to specific problems needs to be sharpened. As it stands it suffers from too many degrees of freedom for fitting the analysis in order to suit the developmental predictions. We then looked at three explanations in terms of direct functional dependencies between theory of mind and executive function. One of them locates the dependency at the superficial level of theory of mind tasks as a measuring instrument. A review of available data makes it unlikely that all theory of mind tasks that show a relation with executive function tasks have themselves a strong executive component, although some (deceptive pointing) may have a substantial component. Available data are by and large compatible with the position that executive functioning is a prerequisite for developing a theory of mind and with the position that a theory of mind is an integral (constitutive) part of executive functions. In fact the latter theory, in all plausibility, implies a weak version of the former. If theory of mind is integral to executive functioning and theory of mind needs to be formed by exercising it, then executive functioning is probably essential because it is the main grounds for developing a theory of mind. Hence executive function is a requirement for normal development of theory of mind. Executive problems should lead to delayed acquisition of a theory of mind. However, the possibility remains that delayed development is possible despite complete lack of executive control. References Bartsch, K., & Wellman, H. M. (1989). Young children's attribution of action to beliefs and desires. Child Development, 60, 946964. Carlson, S. M. (1997, April). Individual differences in inhibitory control and children's theory of mind. Poster presented at the biennial meeting of the Society for Research in Child Development, Washington, DC.
Page 151
Carlson, S. M., Moses, L. J., & Hix, H. R. (1998). The role of inhibitory processes in young children's difficulties with deception and false belief. Child Development, 69, 672691. Clements, W. A., & Perner, J. (1994). Implicit understanding of belief. Cognitive Development, 9, 377397. Clements, W. A., & Perner, J. (1997). When actions really do speak louder than words—but only implicitly: Young children's understanding of false belief in action. Unpublished manuscript, University of Sussex, Falmer, East Sussex, UK. Diamond, A. (1991). Frontal lobe involvement in cognitive changes during the first year of life. In K. R. Gibson & A. C. Peterson (Eds.), Brain maturation and cognitive development. (pp. 127180). New York: Aldine De Gruyter. Doherty, M. J., & Perner, J. (1998). Metalinguistic awareness and theory of mind: Just two words for the same thing? Cognitive Development, 13, 279305. Flavell, J. H., Flavell, E. R., & Green, F. L. (1983). Development of the appearancereality distinction. Cognitive Psychology, 15, 95120. Frye, D. (in press). Theory of mind, domain specificity, and reasoning. In P. Mitchell & K. J. Riggs (Eds.), Children's reasoning and the mind. Hove, England: Psychology Press. Frye, D., Zelazo, P. D., Brooks, P.J., & Samuels, M. C. (1996). Inference and action in early causal reasoning. Developmental Psychology, 32, 120131. Frye, D., Zelazo, P. D., & Palfai, T. (1995). Theory of mind and rulebased reasoning. Cognitive Development, 10, 483527. Gerstadt, C. L., Hong, Y. J., & Diamond, A. (1994). The relationship between cognition and action: Performance of children 3andahalf to 7 years on a Stroop like daynight task. Cognition, 53, 129153. Gopnik, A., & Astington,J. W. (1988). Children's understanding of representational change and its relation to the understanding of false belief and the appearance reality distinction. Child Development, 59, 2637. Harris, P. L. (1989). Object permanence in infancy. In A. Slater & G. Bremner (Eds.), Infant development (pp. 103121). Hillsdale, NJ: Lawrence Erlbaum Associates. Harris, P. L., & Kavanaugh, R. D. (1993). Young children's understanding of pretence. Society for Research in Child Development Monographs, 58 (1, Serial No. 237). Hogrefe, J., Wimmer, H., & Perner,J. (1986). Ignorance versus false belief: A developmental lag in attribution of epistemic states. Child Development, 57, 567582. Hughes, C. (1998). Executive functions in preschoolers: Links with theory of mind and verbal ability. British Journal of Developmental Psychology, 16, 233253. Hughes, C. (in press). Finding your marbles: Does preschoolers' strategic behavior predict later understanding of mind? Developmental Psychology. Hughes, C., & Russell, J. (1993). Autistic children's difficulty with mental disengagement from an object: Its implication for theories of autism. Developmental Psychology, 29, 498510. Jeannerod, M. (1997). The cognitive neurosciences in action. Oxford: Blackwell. Leslie, A. M. (1987). Pretense and representation: The origins of "theory of mind." Psychological Review, 94, 412426. Leslie, A. M. (1994). Pretending and believing: issues in the theory of ToMM. Cognition, 50, 211238. Luria, A. R., Pribram, K. H., & Homskaya, E. D. (1964). An experimental analysis of the behavioural disturbance produced by a left frontal arachnoidal endothelioma (meningioma). Neuropsychologia, 2, 257280. Moore, C., Barresi, J., & Thompson, C. (1998). The cognitive basis of futureoriented prosocial behavior. Social Development, 7, 198218. Norman, D. A., & Shallice, T. (1986). Attention to action. Willed and automatic control of behavior. In R. J. Davidson, G. E. Schwartz, & D. Shapiro (Eds.), Consciousness and selfregulation (Vol. 4, pp. 118). New York: Plenum. Pacherie, E. (1998). Motorimages, self consciousness and autism. In J. Russell (Ed.), Autism as an executive disorder (pp. 215255). Oxford, England: Oxford University Press.
Page 152
Perner, J. (1991). Understanding the representational mind. Cambridge, MA: MIT Press. Perner, J. (1995). The many faces of belief: Reflections on Fodor's and the child's theory of mind. Cognition, 57, 241269. Perner, J. (1998). The metaintentional nature of executive functions and theory of mind. In P. Carruthers & J. Boucher (Eds.), Language and thought (pp. 270 283). Cambridge, England: Cambridge University Press. Perner, J., & Lang, B. (in press). Theory of mind and executive function: Is there a developmental relationship? In S. BaronCohen, H. TagerFlusberg, & D. Cohen (Eds.), Understanding other minds: Perspectives from autism and developmental cognitive neuroscience. Oxford, England: Oxford University Press. Perner, J., Lang, B., & Stummer, S. (1998). Theory of mind and executive function: Which depends on which? Unpublished manuscript, University of Salzburg, Salzburg, Austria. Perner, J., Leekam, S. R., & Wimmer, H. (1987). Threeyear olds' difficulty with false belief: The case for a conceptual deficit. British Journal of Developmental Psychology, 5, 125137. Perner,J., & Stummer, S. (1998). Say something different—ToM, Metalinguistic awareness, embeded conditionals or executive function? Unpublished manuscript, University of Salzburg, Salzburg, Austria. Repacholi, B. M., & Gopnik, A. (1997). Early reasoning about desires: Evidence from 14 and 18montholds. Developmental Psychology, 33, 1221. Russell, J. (1996). Agency: Its role in mental development. Hove: Erlbaum (UK) Taylor & Francis. Russell, J., Jarrold, C., & Potel, D. (1994). What makes strategic deception difficult for children the deception or the strategy? British Journal of Developmental Psychology, 12, 301314. Russell, J., Mauthner, N., Sharpe, S., & Tidswell, T. (1991). The "windows task" as a measure of strategic deception in preschoolers and autistic subjects. British Journal of Developmental Psychology, 9, 331349. Samuels, M. C., Brooks, P. J., & Frye, D. (1996). Strategic game playing in children through the windows task. British Journal of Developmental Psychology, 14, 159172. Shultz, T. R., Wells, D., & Sarda, M. (1980). The development of the ability to distinguish intended actions from mistakes, reflexes, and passive movements. The British Journal of Social and Clinical Psychology, 19, 301310. TagerFlusberg, H., Sullivan, K., & Boshart, J. (1997). Executive functions and performance on false belief tasks. Developmental Neuropsychology, 13, 487493. Wimmer, H. (1989). CommonSense Mentalismus und Emotion: Einige entwicklungspsychologische Implikationen [Commonsense mentalism and emotions: Some implications for developmental psychology]. In E. Roth (Ed.), Denken und Fühlen [Thinking and feeling] (pp. 5666). Berlin: Springer Verlag. Wimmer, H., & Mayringer, H. (1998). False belief understanding in young children: Explanations do not develop before predictions. International Journal of Behavioral Development, 22, 403422. Wimmer, H., & Perner,J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception. Cognition, 13, 103128. Zelazo, P. D., Carter, A., Reznick, J. S., & Frye, D. (1997). Early development of executive function: a problemsolving framework. Review of General Psychology, 1, 198226. Zelazo, P. D., & Frye, D. (1997). Cognitive complexity and control: A theory of the development of deliberate reasoning and intentional action. In M. Stamenov (Ed.), Language structure, discourse and access to consciousness (pp. 113153). Amsterdam & Philadelphia: John Benjamins. Zelazo, P. D., &Jacques, S. (1997). Children's rule use: representation, reflection and cognitive control. In R. Vasta (Ed.), Annals of child development (Vol. 12, pp. 119176). London: Jessica Kingsley. Zelazo, P. D., & Reznick, J. S. (1991). Agerelated asynchrony of knowledge and action. Child Development, 62, 719735.
Page 153
Chapter 9— The Theory of Ascriptions David R. Olson Deepthi Kamawar Ontario Institute for Studies in Education of the University of Toronto The title of this chapter is a play on the famous RussellFrege theory of descriptions that attempted to account for naming (Kripke, 1972) but that also laid the foundation for much of the current interest in representations. Cognitive science, as is well known, is the science of mental representations, their structure, their role in behavior, and for us, their development. In this chapter we explore the role of ascriptions in the development of mental representation and the relation between ascription of intention and epistemological understanding. By ascription, we mean the ability or practice of attributing particular states or properties to persons, objects, or events. Ascriptions differ from descriptions in their ontological commitments. Descriptions purport to be objective; that is, they assume the reality of the thing described and hence its availability to any observer. Ascriptions are describerrelative: They may or may not assume the reality of the property or its availability to other observers. Ascriptions, then, are characterizations of things from an ascriber's point of view. Recall the unfortunate Polonius' eagerness to accommodate to Hamlet's ascriptions: "Do you see yonder cloud that's almost in shape of a camel?" In the current literature, the issue of the nature and function of mental representations, specifically beliefs, desires, and intentions, is taken to be a different enquiry than the issue of the ascriptions of these states to others. This latter issue is seen as the problem of metarepresentation, of the recognition and understanding of beliefs quite independently of the possession of those beliefs. Ascription of beliefs is seen as an aspect of social
Page 154
behavior that presupposes the reality of the beliefs ascribed. The alternative, which we shall consider here, is that ascription of mental states is a critical ingredient in the possession of those states. What begins as ascription results, through socialization, as describable mental states. Of course, that is easier said than done. Ascribing mental states to computers has little effect on the computers; ascribing them to children, it can plausibly be argued, does. So what makes the uptake of such ascriptions possible? We return to this issue in the second part of the chapter. Note that our point is not the equally important point, namely, that the ascription of mental states to others is closely tied to the ascription of those states to oneself. That concepts of belief apply equally to self and other is both an empirical fact (Gopnik & Astington, 1988) and a theoretical necessity (Baldwin, 1897/1973; Barresi & Moore, 1996). Here, we explore the possibility that the ability to ascribe mental states is relevant to the possession of these states. We second the suggestion earlier made by Davidson (1975, 1985) that to have a belief one must have a concept of belief. Davidson (1985) made three points in his argument for this claim. He began by claiming that a single belief cannot exist in isolation; that is, a creature cannot be said to possess just one belief. Rather, in order for something to be a belief, that something must be situated within a complex set of interconnected beliefs; it is the belief's position within this larger logical network that provides it with identity and content. For example, Davidson describes a situation in which a dog is observed chasing a cat that it loses sight of. This results in the dog barking up the wrong tree, an oak tree in this case (the cat is in a nearby tree). The casual observer might want to attribute to the dog the belief that the cat is in the oak tree (in order to explain why he is barking up that tree). However, in order to attribute that belief to the dog it would be necessary to attribute other beliefs to the dog (e.g., that the cat went up the same tree the last time the dog chased it, that trees are alive, etc.). It does not make sense to say that the dog has a belief about a tree if we cannot attribute any other beliefs to the dog. Although Davidson did not explicitly extend this example to cover young children, it seems natural to do so. The second point Davidson made concerns the relation between the phenomenon of surprise and possession of the concept of belief. Davidson pointed out that in order to be surprised, one must be able to compare one's present beliefs (based on reality) with one's prior beliefs (which turned out to be wrong). That is, in order to be surprised, one must have a belief about a belief (e.g., I now believe that my previous belief was false) and be able to understand that beliefs can be true or false (i.e., have a concept of objective truth). Therefore, in order to be surprised, one must possess the concept of belief.
Page 155
Thus far it is easy to agree with Davidson: Beliefs don't exist in isolation and to be surprised involves the concept of belief. The next step requires a bit of a leap. The third point of Davidson's argument held that "one cannot have a general stock of beliefs of the sort necessary for having any beliefs at all without being subject to surprises that involve beliefs about the correctness of one's own beliefs" (p. 479). In other words, once one's network of beliefs is complex enough for the individuated items within it to count as beliefs, one is subject to being surprised about at least some of those beliefs. Therefore, one cannot have a belief without already having the concept of belief. Put another way, to have a belief is to be able to represent psychological states and processes as beliefs and to be able to ascribe beliefs (e.g., surprise) to oneself and others. Beliefs and Intentions as Ascribable Mental States Ascription theory is traceable to Sellars' (1963) famous story "The Myth of Jones." According to that story, Jones discovers that he can better anticipate the actions of another by attributing or ascribing such mental states as beliefs and feelings to that person. But having worked out that that characterization is useful for understanding that other person, Jones discovers that the same account may be usefully applied to himself. In recent years, the story has been improved on by, among others, Barresi and Moore (1996) who abandoned the ordering of acquisition assumed by Sellars: Namely, first understand others and then understand self by means of the same model, and offer in its place the compelling notion that combining information about the self and other is necessary to form a concept of belief. The two capture different sorts of information. The firstperson information about the self must be coordinated with the thirdperson information about the other in order to derive the firstperson information for the other and hence allow prediction of the other's actions. Consider how one would come to understand how perception leads to knowledge. Children, whose mental lives we are attempting to explain, already know a procedure for gaining knowledge. If they want to know what is in a box, they look into the box. But they can do so without understanding the causal relations between looking and knowing. To understand the relation between looking and knowing, children have to coordinate information about their own firstperson experience, namely, "I know what is in the box," with the thirdperson information about what the other is doing, "He is looking into the box." Putting the two together allows them to grasp the causal relation between the two, namely, that it is the "looking" that causes the "knowing.'' This knowledge then allows the ascription of firstperson information to the other: "He now knows what is in the box [because he looked into the box]." It also allows the complementary attitude to oneself,
Page 156
what Barresi and Moore called the thirdperson stance to oneself, namely, "The reason I know what is in the box is because I looked into the box." To understand false belief one has, in addition, to recognize the implication of the negation, "He did not look so he does not know," and draw the inference that because he did not see, he will base his judgment or action on what he saw previously (Artuso, 1998). The standard counterargument to the ascriptivist proposal is that children may have beliefs even if they cannot ascribe them. In some sense children may know what is in the box even if they are unable to ascribe the appropriate belief to the other, and if they know, one may argue they have a belief. We do not wish to deny they are in a cognitive state when they act; the question is the status of that knowledge that the subject can neither ascribe, acknowledge, nor avow. This point could easily turn into a semantic quibble over the use of the word belief. Our purpose is not to deny internal cognitive states to infants and nonhuman animals but rather to argue that to ascribe such states as beliefs in cases in which they are incapable of ascribing them to themselves is suspect. Although there are cognitive states, they may be more appropriately characterized in nonintentionalist terms such as expectancies or attunement, namely, those available in subsymbolic connectionist models (Shultz & Mareschal, 1997). The problem is in the peculiarity of the claim that the child knows. That claim is, obviously, an ascription. We adults ascribe knowledge to the child that the child cannot ascribe to himself: "Do you know what is in the box [when you have not looked]?" Children often claim to know in cases where we adults ascribe ignorance. And conversely, when asked, "What did you think was in the [Smarties] box?" children answer, incorrectly, "Pencils.'' In such cases, it is we, the ascribers, who attribute to the child the belief that he thought there were Smarties in the box, because he had earlier acknowledged that fact, even if now, when asked, he says, "Pencils." That is, we adults characterize the child's actions in terms of beliefs, desires, and intentions even if the child, at this age, either does not or does so incorrectly. There is a further problem. The child is in just the same position as the computer in the Turing test or the chimpanzee who hides his food in the presence of a dominant male. In each case, it is the theorist who ascribes the belief to child, animal, or machine. The child appears to have a belief; the computer acts as if it had a belief; the chimpanzee acts as if it had a belief. These debates have gone in circles, some claiming that they do, others that they do not. The behavior of chimpanzees and computers can at least plausibly be explained without the necessity of appealing to intentional mental states (see Povinelli, chap. 11, this volume). Ascribing mental states such as belief appears to have become a matter of theoretical
Page 157
preference. We are free to choose to adopt the "intentional stance" as Dennett (1978) calls it (see also Olson, 1993), but there is no fact to be described. There is, rather, a characterization to be ascribed by us as adults, if we choose to. The story changes abruptly when children develop the ability to ascribe mental states to themselves and others. Then the question as to whether the child has beliefs may be answered affirmatively. The child now joins in the game. The child in ascribing beliefs has gained an understanding of and, arguably, some control over the conditions of knowing. That is, once the child understands how knowledge is caused, he is in the position to deliberately alter the knowledge states of himself and others. One manifestation is the now well known case of deception (Peskin, 1992). Another landmark of interest is the child's ability to correctly ascribe propositional attitudes (e.g., She thinks that X; I know that Y). The type of situation we discuss here is a referentially (Quine, 1955) or semantically opaque (Davidson, 1985) context. Such contexts can be defined as those statements that may change from true to false when the words that are used to pick out the object of the attitude are substituted with an alternate referring expression (Davidson, 1985). For example, although it might be true that 'John thinks the bird is in the oak tree," it might not be true that "John thinks the bird is in the oldest tree in sight," even though the oak tree is the oldest tree in sight, because he might not know anything else about the trees. Thus, when attributing propositional attitudes to another, the other's beliefs must be taken into account. Let us turn back to Davidson's earlier example of a dog chasing a cat up a tree. The casual observer in that example was willing to say that the dog believes that the cat was in the oak tree. If the oak tree was the oldest tree in sight, should the observer be willing to say that the dog believes that the cat is in the oldest tree in sight? In the case of a dog (or a young child), such questions seem problematic. But if one is willing to use such attitudes as believe, think, and know while dropping the feature of semantic opacity (that of nonsubstitutability), then one is not using those words to ascribe propositional attitudes (thoughts). Our research (Kamawar & Olson, 1997) into children's sensitivity to such contexts showed that there is a relation between the development of a theory of mind and sensitivity to opaque contexts, even with age removed as a factor. It was found that children who have not yet acquired a theory of mind are the same children who are unable to correctly attribute propositional attitudes (thoughts) to another. Furthermore, those children who are able to correctly ascribe thoughts to another are those that were also successful on tasks taken to measure the concept of belief (i.e., theory of mind tasks). Awareness of the constraints of substitution within semantically opaque contexts thus provides evidence of the acquisition of the concept of belief (and hence beliefs).
Page 158
This understanding of, and hence control of, one's own mental life is a form of metacognition, knowing about "knowing." The quotes here mark the uncertain status of the earlier knowing. However, the landmark, if this view is correct, is the appearance of the child's correct ascription of mental states. Whether or not infants, other animals, or computers "have" mental states, then, is not decideable; it depends upon the habits and preferences of the ascribers. Skinner (1983), it will be recalled, withheld such ascriptions even from human adults. But once the child becomes an ascriber, mental states become part of the cognitive economy. Beliefs, in this view, are abstract characterizations of relations between cognitive mechanisms and external states of world, which are useful for both understanding and organizing those relations. Beliefs, rather than being intrinsic structures of a complex brain, are as we say, "social constructions" that, when grasped or internalized, function cognitively. These are not merely convenient fictions, but fictions, that is, constructions, treated as real in just the same way that, in pretense, a child can "let" a banana be a telephone. Perhaps even when children have become ascribers, they are mistaken. They only think they have beliefs, a view which has come to be called "eliminativism" (Stich, 1984). The criticism cuts too deeply in that although it may be helpful to show that the behavior of young children, like that of computers and some other primates, may be adequately explained without appeal to beliefs and intentions, it fails to acknowledge the role of beliefs and other normative structures in social interactions. Holding and ascribing beliefs allows for a way of thinking of oneself and others as capable of planning, of accepting responsibility, and of treating others as partners in a social world, as well as of creating and understanding public representations. In addition to the logical reasons just offered for granting the "ascriptivist" view of beliefs, there are some empirical ones. If beliefs are part of the intrinsic structure of mind of a certain level of complexity, a view shared by such otherwise discrepant writers as Searle (1983) and Fodor (1992), we should expect concepts of mind to emerge, rather than be learned. If, on the contrary, understanding beliefs is a way of viewing people in relation to a world, then that understanding may have to be taught. There is now a considerable body of evidence linking language development to theory of mind (Artuso, 1998; Fletcher et al., 1995; Frey, 1997; Gale, de Villiers, de Villiers, & Pyers, 1996; Jenkins & Astington, 1996). It appears that the understanding of beliefs is tied extremely closely to the acquisition of a language for attributing, that is, ascribing, beliefs to self and others. Acquiring a theory of mind is not easily distinguished from acquiring a language for talking about the mind, and using the language correctly is a matter of correctly ascribing mental states.
Page 159
The importance of language suggests a link, as well, between ascription and description, mentioned earlier. It is the ascription of beliefs and intentions to the child by the adult and the overhearing of such ascriptions by others to others, that may allow the child to come to consider him or herself as a believer and intender. Once these ascriptions are adopted, the child can be correctly described as a believer. What began as ascription would have become description. As Davidson (1975) suggested, it is a mistake to ask if mental states such as beliefs are real. They are and they are not. They are "representations," not "reals," and yet by ascription we treat them as reals in much the same way as by the ritual of naming we become who we are. Social constructions, too, are a kind of real entity, real by virtue of our agreement to treat them so. There are limits to ascriptivism. To learn to ascribe beliefs to oneself and others requires much more than social learning of a language. One must have some first person experience that one can contribute as a basis for ascriptive practice. These experiences provide the subjective side of the explicit knowledge needed for ascription. Beliefs are the product of combining some firstperson experience with the perception of the thirdperson experience of the other and combining them to produce a concept or an understanding of belief. But if the product is a belief, that firstperson experience, itself, cannot already have been a belief. To describe this firstperson experience we will appeal to the concept of "feeling." Babies may not have beliefs but they do have feelings and although one can sometimes recognize those feelings, one is at a loss to know their thoughts. "One can no more think like a baby, than one can think like a bee," wrote Richard Hughes, author of A High Wind in Jamaica (1928, p. 119). For a learner to understand what the experience of another is like, it is essential that the child have had some similar experience. I know what it is like to get knocked off a bicycle because it has happened to me. If I had no such experience I would have a limited understanding of what it is like. As children get older and develop a life of the imagination, they come to have some idea of what it is like even without the experience, but such fantasies never approach realities. Anthony Burgess (1980) in Earthly Powers creates a character, a writer, who gets mugged. The character then points out that in his writing he had often described in intimate detail fictional muggings and in fact thought he knew just what it would be like. His comment, after he was actually mugged, was that it was nothing like he had imagined it at all. Knowledge of what an experience is like is limited by the experience of the ascriber. Adults may go somewhat beyond experience by means of imagination but this is a route that is limited or unavailable to infants. So what is it that the child brings from his firstperson experience that is usefully added to his perception of the other's thirdperson experience?
Page 160
Our suggestion is that it is a set of feelings, including a feeling of knowing or a feeling of understanding. Because the child knows what it feels like to know something on looking into a box, the child can attribute that feeling to the other when the other peers into a container. Thus the limiting condition on ascription is that the child has to have had some personal firstperson feeling that he then can ascribe to others when the other has access to information. Without that feeling the child would have no idea of "what it is like" to look into a container. This is the sense in which we think Paul Harris is correct in arguing for what he calls the "simulation theory" (Harris, 1992). The notion of "what it is like" is borrowed from Thomas Nagel (1974) who uses it to explore the limits of human knowledge. We have no idea of what it is like to be a bat navigating by a form of radar. And we cannot learn what it is like. Our experience puts a limit on what we can ascribe to others. Learning to ascribe beliefs, then, rests on the ability of the learner to ascribe to others the feeling that he himself has had on occasions similar to those now observed in the other. That feeling, we have urged, was not itself a belief. The belief is the product of relating the feeling that one knows in oneself to the perceived occasion or situation of the other. More concrete, combining the feeling of knowing attributed to the other on the basis of one's own experience with the perception of the other's actions of looking into the container produces the relation between seeing and knowing, itself the fundamental understanding underlying the concept of belief. This is close to the story advanced by Barresi and Moore (1996) with three minor revisions. First, we see no compelling reason to postulate an "intentional schema" with the function of producing the relation between firstperson and thirdperson information for both self and other. Second, more important, we limit what the child can ascribe to what the child knows of "what it is like." What it is like provides a learning mechanism in its own right—learning by relating feelings provoked by similar events. Third, the notion of feelings ties this early learning not exclusively to cognition but to affect. Davidson (1985) offers another way in which to view the importance of first and thirdperson information to the development of beliefs and the concept of belief. He maintains that the interaction of two creatures (people) is necessary for either of them to develop a concept of the way things exist objectively (which in turn is important in understanding that beliefs can be true/false). Davidson, using the metaphor of triangulation (two perspectives and one object), holds that it is through language that two creatures can create a common ground of an objective reality, and that this sharing of a concept of truth is what makes sense of the claim that the creatures have beliefs and can assign objects a place in the public world.
Page 161
Ascription theory is based on the assumption that concepts like belief, desire, and intention, rather than acting as pointers to actual cognitive structures are representational inventions that are acquired because they are handy, that is, instrumental in accounting for talk and action. But in our view, it goes a step further to ask what the limits of such ascription are. Contrary to the strong "theorytheory" view, the view that these concepts are the free invention of young children faced with complex social interactions (Gopnik & Wellman, 1994; Sellars, 1963), it suggests that the concepts are limited by some more fundamental experiences of the child that we have characterized as feelings. Babies may not have beliefs but they do have feelings. Feelings of knowing, of course, are often wrong. But if they were always unreliable, one would have no firstperson feelings to bring to the learning situation. Some indication of the understanding of internal states that children bring to the task of acquiring a theory of mind comes from their learning of modal auxiliaries such as might be, must be, could be, and should be, which Moore, Pure, and Furrow (1990) found to be closely related to the understanding of mental states. Astington and Olson (1990) pointed out that modals used in a firstperson sense are equivalent to the attribution of belief thus: "It must be in the cupboard" is equivalent to "I think it is in the cupboard." But this form is not generalizable to others: "It must be in the cupboard for him" is unacceptable, whereas "He thinks it must be in the cupboard" is acceptable. Only the latter expressions attribute the property to the mind. Modals attribute the property to the structure of the world. A similar point may be made about the appearancereality distinction. If one says that a sponge looks like a rock, the problem is attributed to the world, the rock. But if one says that ''I thought it was a rock," one is attributing it to a mental state. In both cases, it is a move from responding to the world in a particular way, which presumes children know "how it feels," to the attribution of the feeling, now in the form of a belief, ascribed to others. Therefore, ascriptions are the product of representing generally what one already experienced as a feeling. Ascriptions allow us to furnish our social worlds with shared ideas, hypotheses, beliefs, and arguments. In addition, they allow us to form an explicit epistemology, an understanding of the conditions of knowledge, and the ability to contribute to the "archival store" of knowledge (Olson, 1994). To this second role of ascription theory we now turn. How do ascriptions allow us to function epistemologically? Ascription theory specifies not only that one has beliefs but also the causes1 of those 1
By "cause" here, we refer to intentional causes, which is to say "reasons," not the mechanical causes of the sciences.
Page 162
beliefs. Once understood, one is in a position to answer such questions as "How do you know?" and also to arrange things so that one knows, that is, to collect and understand evidence. One is in a position to assess arguments and provide evidence for claims. Acquiring an understanding of the grounds for ascriptions of belief to self and others, then, is central to children's epistemological development. Indeed, the recent concern with metacognition can be seen, in large part, as a concern with the assignment and evaluation of the basis for one's own beliefs. And, as noted earlier, evaluating and assigning beliefs to oneself is closely related to the evaluation and assignment of beliefs to others. That is, questions of how one knows apply equally to self and other. Whereas it may be somewhat easier to criticize the beliefs of others than those of oneself and easier to adjust one's own beliefs than to criticize those beliefs, the procedures for evaluating beliefs are the same whether they are one's own or those of another. Epistemological development is a matter of learning how to recognize and categorize types of claims and other statements and the ways that those statements can be criticized. The critical role falls on the concept of evidence: "How do you know?" Gopnik and Graf (1988) showed that children who lacked a theory of mind had difficulty in correctly attributing knowledge to the appropriate source. Our concern here is with their understanding that knowledge is caused by experience at all, not so much with distinguishing sources. In our laboratory for some years we have pursued the question of children's understanding of evidence for a belief. In a typical study (Kamawar & Olson, 1998), children are told a story about a subject (John) who attempts to open a door but fails and incorrectly infers out loud that the door must be locked. He then goes away to find a key. In his absence, the children are shown that the actual cause of the failure to open the door is a wedge under the door rather than because it is locked. Thus far this is a false belief story. But here the scenario changes. The children are asked three questions: (a) a question about the cause of the state of affairs in the world, for example, "Why will the door not open?" (correct answer: because of the wedge); (b) a question about the character's thoughts, for example, "Why does John think that the door won't open?" (correct answer: because he couldn't open the door); and (c) a question about how the character knows the state of affairs, for example, ''How does John know that the door won't open?" (correct answer: because he tried and it wouldn't open). Although this work is still ongoing, previous findings (Sa & Olson, 1996) were that children under 5 or 6 years conflate the three questions, answering, in each case, "Because of the wedge." In addition, they also often respond with answers we describe as showing "reverse logic"; for example, when asked "Why does John think that the door won't open?" some of the younger
Page 163
children say that John thinks that the door will not open "because it is locked," rather than "He thinks the door is locked because it won't open." We infer that children have an imperfect understanding of the conditions of knowing, treating causes of events as the reasons for knowing about them. Although working in a somewhat different theoretical framework, Ramadas (1994) and Astington and Homer (1995) have both reported similar results. In these cases, theory of mind viewed as an ascriptive theory may be seen as directly tied to this aspect of epistemological development. A second line of research relating theory of mind to epistemological development examines the role of the understanding of intention in conceptual change. Conceptual change is evoked when knowledge acquisition is not only augmented but interpreted within a quite different conceptual framework. Intentionality provides one such framework. It has been observed (Roth & Anderson, 1988; Shapiro, 1994) that when asked how light permits us to see an object, children tend to say, "The sun shines on the tree and we look at the tree," whereas the school attempts to teach them that the sun shines on the tree and the tree reflects the light into our eyes. Furthermore, when asked why they cannot see something behind a brick wall, they say, 'You cannot see through a brick wall," whereas the school attempts to teach them to say that light cannot pass through a brick wall. What is the problem? First, even adults accept both modes of explanation: "We cannot see through the brick wall" and "Light cannot pass through the brick wall." Indeed, most adults see no difference between these two statements. However, in the first case, there is the implicit assumption that in looking and seeing, the light passes from the eye through the wall rather than light failing to pass through the wall to the eye. Consequently, we readily imagine the feasibility of Superman's Xray vision. These directions are indicated in Fig. 9.1 (adapted from Roth & Anderson, 1988, but labeled in terms of the view developed herein). Following Carey and Spelke (1994), we suggest that children are caught between a causal frame of reference, suitable for physical phenomena, and an intentional frame of reference, appropriate for psychological ones. Young children tend to be tied to an intentional mode of explanation in terms of which people and other things do or try to do, whereas schooled and older children and adults can, if required, shift to a causal mode of explanation in terms of mechanical processes and causal regularities of light. The first theory is the intentional theory premised on the actions of the looker, the second, Newton's causal theory of light. Even adults are not immune to the intentionalist theory of light. As mentioned even adults commonsensically say that we cannot "see through" a wall, implying that the seeing begins with the eye. N. Poussin, a renowned 17thcentury French painter, wrote the following: "Once you have received your painting . . . adorn it with some framing, for it needs it; so that when
Page 164
Fig. 9.1. Modes of explanation.
gazing at it in all its parts, the rays of the eye are retained and not scattered outside" (Poussin, 1911). The "rays of the eye" are just what children seem to assume in explaining vision. These supposed rays would also explain why people often feel they can tell when someone is looking at them, even when they cannot see that person, and their credulity when confronted with Superman's "enhanced'' vision. The link to ascriptive theory is as follows. In learning to ascribe beliefs and other intentional states to others, the child is moving from a nonintentional stance to an intentional one. That transition is well studied and well known. In the epistemological case, the transition is just in the opposite direction. Once children understand that intentions are ascribed by the viewer, they are in a position to grasp that such ascriptions are inappropriate in some contexts, such as that of understanding light. Once that is understood, they are in a position to begin a deeper understanding of physical matter including light. This hypothesis needs to be tested experimentally. As mentioned, Skinner's ability to recognize that one could resist the temptation to ascribe intentions to persons was what allowed him to produce a quite novel, if now widely dismissed, explanation of behavior. Modern connectionists have taken a similar goal, albeit with quite different means. We have not ventured far from the wellknown position set out by Dennett (1978) in what has come to be known as instrumentalism. On that view, mental states, such as beliefs, are not real, causal states of organisms, but rather ways of characterizing action and interaction from a particular,
Page 165
outsider's, point of view. Our ascriptions of mental states are useful to us as ascribers in accounting for the talk and action of others, and, following Sellars (1963), of both understanding ourselves and gaining some control of our behavior and, ultimately, our knowledge. References Artuso, M. (1998). Information processing constraints on the acquisition of a theory of mind. Unpublished PhD dissertation, University of Toronto. Astington, J., & Homer, B. (1995, April). Children's understanding of secondorder belief and its relation to their ability to distinguish between causes and reasons. Paper presented at the annual meeting of the American Educational Research Association, San Francisco, CA. Astington, J., & Olson, D. (1990). Metacognition and metacognitive language: learning to talk about thought. Applied Psychology: An International Review, 39, 7787. Baldwin, J. M. (1973). Social and ethical interpretations in mental development. New York: Arno Press. (Original work published 1897) Barresi, J., & Moore, C. (1996). Intentional relations and social understanding. Behavioral and Brain Sciences, 19, 107154. Burgess, A. (1980). Earthly powers. New York: Simon & Schuster. Carey, S., & Spelke, E. (1994). Domainspecific knowledge and conceptual change. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture (pp. 169200). New York: Cambridge University Press. Davidson, D. (1975). Inquiries into truth and interpretation. Oxford, England: Clarendon. Davidson, D. (1985). Rational animals. In E. LePore & B. McLaughlin (Eds.), Actions and events: Perspectives on the philosophy of Donald Davidson (pp. 473480). Oxford, England: Basil Blackwell. Dennett, D. (1978). Brainstorms. Montgomery, VT: Bradford. Fletcher, P. C., Happe, F., Frith, U., Baker, S. C., Dolan, R. ., Frackowiak, R. S. J., & Frith, C. D. (1995). Other minds in the brain: A functional imaging study of 'theory of mind' in story comprehension. Cognition, 57, 109128. Fodor, J. (1992). A theory of the child's theory of mind. Cognition, 44, 283296. Frey, R. (1997). General linguistic competency in the deaf: A prerequisit for developing a theory of mind? Unpublished PhD dissertation, University of Toronto. Gale, E., de Villiers, P., de Villiers, J., & Pyers, J. (1996). Language and theory of mind in oral deaf children. In A. Stringfellow, D. CahanaAmitay, E. Hughes, & A. Zukowski (Eds.), Proceedings of the Boston University Conference on language development (pp. 213224). Boston: Cascadilla Press. Gopnik, A., & Astington,J. (1988). Children's understanding of representational change and its relation to the understanding of false belief and the appearancereality distinction. Child Development, 59, 2637. Gopnik, A., & Graf, P. (1988). Knowing how you know: Young children's ability to identify and remember the sources of their beliefs. Child Development, 59, 13661371. Gopnik, A., & Wellman, H. M. (1994). The theorytheory. In L. A. Hirschfeld and S. A. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture (pp. 257293). New York: Cambridge University Press. Harris, P. (1992). From simulation to folk psychology: The case for development. Mind and Language, 7, 120144. Hughes, R. (1928). A high wind in Jamaica. New York: Harpers.
Page 166
Jenkins, J., & Astington, J. (1996). Cognitive factors and family structure associated with theory of mind development in young children. Developmental Psychology, 32, 7078. Kamawar, D., & Olson, D. R. (1997, June). Theory of mind and referential opacity. Representation and meaning. Paper presented at the annual meeting of the Canadian Psychological Association, Toronto, Canada. Kainawar, D., & Olson, D. R. (1998). Children's understanding of evidence. Manuscript in preparation. Kripke, S. (1972). Naming and necessity. In D. Davidson & G. Harman (Eds.), Semantics of natural language (pp. 254355). Boston: Reidel. Moore, C., Pure, K., & Furrow, D. (1990). Children's understanding of the modal expressions of speaker certainty and uncertainty and its relation to the development of a representational theory of mind. Child Development, 61, 722730. Nagel, T. (1974). "What is it like to be a bat?" Philosophical Review, 83, 43551. Olson, D. R. (1993). The development of representations: The origins of mental life. Canadian Psychology, 34, 293306. Olson, D. R. (1994). The world on paper. Cambridge, England: Cambridge University Press. Peskin, J. (1992). Ruse and representation: On children's ability to conceal information. Developmental Psychology, 28, 8489. Poussin, N. (1911). Correspondence de Nicholas Poussin (C. Jouanny, Ed.). Paris: Ecrits. Quine, W. V. O. (1955). Quanitifiers and propositional attitudes. In J. L. Garfield & M. Kiteley (Eds.), Meaning and truth: The essential readings in modern semantics. New York: Paragon House. Ramadas, 1. (1994). Children's scientific thinking. Vivek, 8, 110. Roth, K., & Anderson, C. (1988). Promoting conceptual change learning from science textbooks. In P. Ramsden (Ed.), Improving learning: New perspectives (pp. 109141). London: Kogan Page. Sa, W., & Olson, D. R. (1996). How do you know? Children's conception of evidence [Mimeo]. Toronto, Ontario: Institute for Studies in Education, University of Toronto. Searle, J. (1983). Intentionality. Cambridge, England: Cambridge University Press. Sellars, W. (1963). Science, perception and reality. London: Routledge & Kegan Paul. Shapiro, B. L. (1994). What children bring to light: A constructivist perspective on children's learning in science. New York: Teachers College Press. Skinner, B. (1983). A matter of consequences: Part three of an autobiography. New York: Alfred Knopf. Shultz, T., & Mareschal, D. (1997). Rethinking innateness, learning, and constructivism: Connectionist perspectives on development. Cognitive Development, 12, 563568. Stich, S. (1984). From folk psychology to cognitive science. Cambridge, MA: MIT Press.
Page 167
PART II COMPARATIVE PERSPECTIVES ON INTENTIONALITY
Page 169
Chapter 10— Primate Representations and Expectations: Mental Tools for Navigating in a Social World Marc D. Hauser Harvard University We are, it seems to me, at an obvious cognitive advantage, not only with respect to rocks and reflex machines, but also with respect to every other kind of creature that we have so far encountered . . . . We are patently far cleverer than anybody else, and that we are cries out for an explanation . . . . [A] good theory of the mind might reasonably be expected to say what it is that's so exceptionally good about our minds . . . . I think there is much to be learned from an account of mental representation that stresses connections between content and causation. —Fodor (1994, pp. 9192)
There is a problem that has been brewing for the last twenty or so years in cognitive science. It is a problem that originally appeared to be like motor oil and water, but has turned out to be more like mixing expensive olive oil and balsamic vinegar, a somewhat soluble solution with tasty consequences. The problem involves integrating the idea that mental states are intentional with the idea that mental states are computational (Fodor, 1994). What is critical to both of these positions, but especially the first, is that mental states have content. Representations of beliefs, desires, and intentions are always about or for something or someone. If we want to develop a theory of mental states, therefore, we need to work out what kinds of representational systems are possible because these form the foundation for the kinds of content that can be operated on by our mental states, or for that matter, any organisms' mental states (Allen & Bekoff,
Page 170
1997; Cheney & Seyfarth, 1990b; Dennett, 1996; Dretske, 1995; Fodor, 1994; Millikan, 1993). In this chapter, I explore the content of primate representation. The argument I develop is that we can learn about the representations of nonlinguistic creatures by creating situations that either violate certain core principles from within the physical or psychological domain, or are consistent with them. This argument has been used in studies of infant cognitive development (Baillargeon, 1995; Carey & Spelke, 1994) with great success, and I see no a priori reason for thinking it would be inappropriately applied to other species (Hauser, 1997b). This is an assumption, one that we can readily test. I start with work on vocal communication, focusing on how expectations are derived from the meaning of the utterance produced. I then examine whether nonhuman primates make a distinction between animate and inanimate objects, and if so, whether the representation underlying this distinction is called on to predict object trajectories. Finally, I explore the possibility that primates represent other minds, and based on these representations, establish expectations for social interaction. Communication, Meaning, and Expectations Expectations about the workings of the world will be specified innately whenever the relevant problem is confronted with sufficient regularity that all members of the species encounter it in the same way, and in approximately the same way that their ancestors did before them (Cosmides & Tooby, 1994; Pinker, 1997). This is not to say that experience plays no role in shaping the representation, but that an innate mechanism is put in place that guides the kind of experience that an individual attends to, as well as the timing of such attentional focus. For example, all animals are likely to be equipped with similar expectations about the physical properties of objects, including: two objects cannot occupy the same space at the same time, objects follow spatiotemporally coherent trajectories as they move, and objects adhere to the principles of gravity (Baillargeon, 1995; Carey & Spelke, 1994; Hauser & Carey, 1998; Hood, 1995; Spelke, 1985, 1994; Spelke, Vishton, & von Hofsten, 1995). Such expectations function as important guidelines for those that are further constructed over the course of a lifetime, as particular experiences forge causal relations between objects, events, and their consequences. For instance, young animals respond differently to animate and inanimate objects in the absence of significant experience (e.g., sensitivity to faces, biological motion, artifactual kinds). During the course of development, however, they learn about more specific aspects of this distinction, including an understanding of dominance relationships, differ
Page 171
ent predatory types, and potentially available, as opposed to unavailable, mating partners (Cheney & Seyfarth, 1990a; Gelman, 1990; Hauser & Carey, 1998; Leslie, 1994; D. Premack, 1990; D. Premack & A.J. Premack, 1995). Developmentally and evolutionarily, the interesting cognitive move comes from the capacity to think about counterfactuals for a given expectation. In essence, to be aware of one's expectations, rather than merely guided by them, one must be able to run a thought experiment, assessing how certain kinds of evidence violate the expectation. Detecting violations is a first step. Understanding why they are violations is a leap, one that our own species may be alone in taking. But that is, as they say, an empirical question, one that we visit over the course of this chapter. Communication provides a useful arena for exploring the nature of animal expectations because the signal produced is designed to manipulate the behavior of receivers, whereas receivers are designed to assess the veridicality of the message conveyed with what they expect in such situations. When an animal produces a signal, it must, at some level, expect the signal to play a functional role. We know, however, from several studies of animal communication, that the consequences of a particular signal are often unpredictable, indicating that there is a degree of uncertainty to all communicative interactions (Caryl, 1979; Dawkins & Krebs, 1978; Krebs & Dawkins, 1984; reviewed in Hauser, 1996). Consequently, there will be strong selection on perceivers to recognize dishonest messages, signals that violate current expectations. This brings us to the topic of deception. To say that one animal deceived another is to say that the deceiver set up a situation that violated what the other expected (Hauser, 1997a; Mitchell, 1986). If alarm calls are typically given when eagles are around, then when an alarm call is produced, listeners will expect to find an eagle. An animal has been deceived if, on hearing the alarm call of the deceiver, they respond as if the eagle were nearby, when in fact it is not. In this situation, the listener's expectations have been violated (Munn, 1986). If we are to make sense of primate deception (Byrne & Whiten, 1990; Whiten & Byrne, 1988), then we need to better understand the meaning of their signals (Cheney & Seyfarth, 1990b). Specifically, when a signal has been emitted and then detected, we need to determine the kinds of representations that are set up, including how such representations lead to expectations about action (D. Premack & Woodruff, 1978). Let me develop these points further. What happens in the children's fable when the boy cries "wolf" but there is no wolf? On hearing the first cry, the townspeople listen and respond with concern because they clearly expect some kind of danger. Wolves are associated with danger and when someone cries out "wolf," it is shorthand for "danger." With repeated false alarms—of "wolf'' without danger—public interest is lost, the orienting response drops from sight,
Page 172
and the child's cry for help falls on deaf ears. If, however, a new boy cries wolf, or if the old boy cries "Crazy Indian Joe," then there certainly will be a revival in response. Just because one boy lies about wolves does not mean all boys lie about wolves. Furthermore, just because the boy lied about wolves does not necessarily mean that he will lie about all sources of danger, although here he would certainly be treading on thin ice. The point then is that because our words have quite specific meanings, and because listeners generally expect an appropriate mapping between word usage, meaning, and the current state of the world, there is a critical connection between communication and truth. Recent playback experiments on vervet monkeys and rhesus macaques have tapped into the logic of the cry wolf fable, using a habituationdiscrimination procedure to establish how call meaning causally connects with listener expectations. In vervets, individuals from one social group produce acoustically distinctive calls (wrrs and chutters) when they detect another group (Struhsaker, 1967). Although chutters appear to be given by animals who are slightly more aggressive than those producing wrrs, both calls appear to share the same general meaning, something like "aggressive neighbor in view." Cheney and Seyfarth (1988, 1990a) designed an experiment to determine whether an individual who unreliably signaled the presence of a neighboring group by giving wrrs would be considered unreliable as well if she switched to chutters. Similarly if one individual unreliably signaled the presence of a neighboring group with wrrs, and then another individual wrred or chuttered, would listeners generalize across callers and call types? Results showed that when caller identity was held constant within a playback session, subjects transferred the level of habituation from wrrs to chutters, as well as from chutters to wrrs. Thus, if monkey A unreliably announces the presence of a neighboring group with wrrs, changing to chutters has little effect on the listener's expectations. A is still unreliable. In contrast, if the test trial involves a shift in caller identity, then listeners respond. Although A may be unreliable with respect to identifying threats from a neighboring group, listeners apparently do not expect all individuals in the group to be unreliable. On the island of Cayo Santiago, Puerto Rico, rhesus produce one or more of five acoustically distinctive vocalizations when they find food (Hauser & Marler, 1993a, 1993b). Three of these calls (warble, harmonic arch, and chirp) are given to highquality rare food items, whereas the remaining calls (coo and grunt) are given to lower quality common food items such as the provisioned monkey chow provided each day. Experiments were conducted to further clarify the notion of unreliability just discussed, using the warble, harmonic arch, and grunt as both habituation and test stimuli; caller identity was held constant within sessions (Hauser, 1998b). Following habituation to warbles, subjects transferred the level of
Page 173
habituation to harmonic arches even though these two call types are acoustically discriminable (Fig. 10.1). Similarly, following habituation to harmonic arches, they transferred habituation to warbles. These results suggest that warbles and harmonic arches are perceived by listeners as conveying similar information. If an individual is unreliably calling about the discovery of highquality rare food, shifting from warbles to harmonic arches does little to change the listener's expectations. When warbles or harmonic arches were used as habituating stimuli, subjects responded to grunts in the test trial, but the magnitude of the response was small (Fig. 10.2). In contrast, when grunts were used as habituating stimuli, the magnitude of the response to warbles or harmonic arches was large. More generally, if an individual has been unreliably calling about the discovery of highquality rare food, a shift to lowquality common food is meaningful, but not that interesting. In contrast, having
Fig. 10.1. Results from a habituationdiscrimination experiment using the rhesus monkey's highquality rare food calls. Subjects were habituated until they failed to respond (i.e., look to the speaker) on two consecutive trials. (a) Habituation to different exemplars of the 'harmonic arch' followed by a test with one exemplar of the warble. Sound spectrograms of each call type illustrated. (b) Habituation to different exemplars of the warble followed by a test with one exemplar of the harmonic arch. Mean response duration (seconds) and standard deviations shown.
Page 174
Fig. 10.2. Results from a habituationdiscrimination experiment using the rhesus monkey's highquality rare and lowquality common food calls. Subjects were habituated until they failed to respond (i.e., look to the speaker) on two consecutive trials. (a) Habituation to different exemplars of the grunt followed by a test with one exemplar of the warble or harmonic arch. Sound spectrograms of each call type illustrated. (b) Habituation to different exemplars of the warble or harmonic arch followed by a test with one exemplar of the grunt. Mean response duration (seconds) and standard deviations shown.
habituated to an individual calling about lowquality common food, a shift to higher quality rare food is not only meaningful, but interesting. The habituation procedure used in these experiments creates an anomalous situation, a violation of the presumed expectation formed by listeners. Every time the call is played back, the putative referent or expected context is missing: intergroup calls without competitors, and food calls without food. Given that the listener's expectations have been violated during the habituation series, test trials are designed to determine whether particular changes in a stimulus class will reinstate interest in the message conveyed. Although the test trial does, of course, represent another violation, it has the potential to capture, albeit momentarily, the subject's attention. And this is sufficient to determine whether the habituating and test stimuli differ in meaning and thus, differ with respect to what listeners expect when they hear such calls. A more formal test of the relation between call meaning and expectation has recently been conducted by Cheney, Seyfarth, and Silk (1995) using the vocalizations of wild baboons and an extremely simple, yet elegant test.
Page 175
Dominant female baboons often grunt when they approach a subordinate mother and infant. In response, the subordinate mother often barks back at the dominant female, a gesture of submission. Thus, the vocal sequence "dominant grunt subordinate bark" should be familiar. In contrast, the sequence "subordinate grunt dominant bark" should be unfamiliar or inconsistent with the dominance relationships of the callers. Playback experiments, using two different test sequences, were run to assess whether baboons generate comparable assessments. The first sequence consisted of a subordinate grunt followed by a dominant bark; this represents a violation of the typical call order of baboon vocalizations. In contrast, the second test sequence, although starting with "subordinate grunt dominant bark," ended with a grunt from a more dominant female; this vocal interaction is now consistent with baboon social relationships because the female's bark is temporally associated with the more dominant female's grunt at the end of the sequence. Subjects consistently looked longer in response to the first test sequence, suggesting that baboons form expectations about vocal interactions based on the meaning of the call, the identity of the caller, and the dominance relationship among the interactants. In all three experiments, the amount of time spent looking at the speaker has been used as a tool to explore the nature of primate representation and expectation. Is looking time a valid measure? Many ethologists believe that it is one of many valid measures, and so do other researchers working with organisms who are equally limited from a linguistic perspective: human infants. To support this claim, I now turn to two other experiments, each designed to provide a more complete picture of the relation between primate representation and expectation. In the final section I return to the looking time measure and critically assess its validity. SelfPropelled Motion and Animacy In the Disney animation Aladdin, there is a segment of technological and psychological genius. Aladdin and his monkey Abu are inside the cave of treasures. Suddenly, they notice that there is a carpet sneaking up on them and then turn to scold it. The carpet sulks, does an aboutface, and moves away. The movement is dramatic. With only tassels and a flexible "body," the carpet slinks away, apparently upset and hurt by Aladdin's verbal chastisement. Although the accompanying music certainly helps, our minds are inevitably led to interpret the carpet's actions as intentional, emotional, and goal directed. We of course know that it is a carpet, but nonetheless automatically attribute complex emotions and mental states. And one of the primary factors guiding our attributions in this case is the carpet's motion, in particular, its selfpropelled motion.
Page 176
Animators have long known about such tricks, of how to turn something that is inanimate into something that appears animate. In experimental psychology, work by Heider and Simmel (1944), as well as Michotte (1962), demonstrated that adults attribute causal properties, including mental states, to simple geometric shapes as long as they move in a particular way. Recently, several cognitive psychologists have argued that before human infants acquire a theory of mind, they must be able to distinguish between objects that require an external agent to move and those that move on their own because they are selfpropelled (Gelman, 1990; Leslie, 1994; D. Premack, 1990). From an adaptive perspective, objects that are selfpropelled are more interesting than those that are not because of their potential for selfinitiated interaction with other objects. Zenon Pylyshyn (cited in Fodor, 1994) captured this intuition when he mused that rocks are smarter than cats because they have the sense to move away when kicked. Of course, nothing could be further from the truth! To determine whether infants generate appropriate predictions about object motion, developmentalists have relied on the expectancy violation procedure (Hauser & Carey, 1998; Spelke, 1985). In parallel with the playback experiments described in the previous section, this procedure assumes that if participants detect a physical or psychological violation, then they will look longer at the event than when there is no violation. To remove the potential effects of novelty, however, subjects are first habituated to, or familiarized with, the objects and events to be displayed in the test trials. This is important because novelty alone can drive the looking time response. This methodological and conceptual approach was implemented by Gergely and colleagues' (Gergely, Nádasdy, Csibra, & Bíró, 1995) experiments with 1yearold human infants. Participants were habituated to an animation that started with a large ball on one end of a table, a small ball at the other end, and a tall barrier in between the two balls (Fig. 10.3). The large ball pulsates and then the small ball pulsates, giving the appearance of an interaction. The small ball then moves up to the barrier and then rolls back, moves up, rolls back, and so on. Eventually it approaches the barrier and hops over, rolling to join the large ball. Once participants habituate to this display, they are exposed to a randomly repeating sequence of test trials. Both test trials consist of the same two balls, but now the barrier has been removed. The only difference between the two test trials is that in one, the ball repeats the same pattern of locomotion as in the habituation trial (even though there is no longer a barrier to hop over) whereas in the other, the small ball rolls straight over to the large ball. The first test sequence is highly familiar, and should be expected if participants are predicting that the small ball will always follow the same path no matter what the path is like (i.e., independently of the barrier). In contrast, the second sequence should be expected if participants predict
Page 177
Fig. 10.3. Methodology used by Gergely et al. (1995) to test for human infants' understanding of object motion and intentionality. Participants were first habituated (top row) to the sequence shown. Then they were presented with either an "irrational" sequence, representing a violation of presumed expectations, or a "rational" sequence, representing a nonviolation.
that the small ball will pursue the shortest path to its target (i.e., the large ball). In other words, because the barrier has been removed, rolling straight ahead should be expected. Infants consistently looked longer when the small ball repeated the motion displayed in the habituation series, even though the alternative pattern of motion was novel. Gergely and colleagues suggest that by the age of 1 year, infants identify objects as rational or irrational based on the patterns of motion they perceive. Thus, in their animations, the small ball is acting irrationally when it appears to jump over an invisible barrier, but rationally when it follows a straight path. This hypothesis, although conceptually fascinating, is undermined by the availability of a simpler interpretation, one that considers infants to be quite sophisticated when it comes to predicting an object's goals, but not when it comes to predicting an object's beliefs (D. Premack & A. J. Premack, 1997a, 1997b). Thus, the small ball is goal directed and simply takes the shortest path to its goal, the large ball sitting
Page 178
at the other end of the table. Under this scenario, which the Premacks have developed in far greater detail, infants are highly sensitive to the patterns of object motion, using them as guidelines for making inferences about goals and possibly, emotions. At some level, all animals must eventually carve a categorical division between objects that are selfpropelled and those that are not. The question is, at what level and when? I now turn to a series of experiments designed to address this problem. An expectancy violation procedure was set up using captive cottontop tamarins as subjects (Hauser, 1998a). Although this New World monkey species is more distantly related to us than vervets, rhesus, or baboons (all Old World monkeys), we were less concerned with the establishment of homologies than with the conditions that favor a particular cognitive process. In particular, the experiment was designed to assess whether the kind of object a tamarin perceives, and then represents in memory as it moves out of sight, affects the kinds of expectations generated with respect to object movement or displacement. Subjects were first presented with the test apparatus, a twochambered box separated by a partition; the center of the partition was cut out, providing a passageway from one side to the other. The tamarins were allowed to enter and explore the test box before the experiment commenced. A session started with four familiarization trials, designed to remove the effects of novelty (Fig. 10.4). In familiarizations 1 and 2, subjects watched as an object was lowered into one chamber and looking time scored during a 10second period. In familiarizations 3 and 4, subjects watched as an object was lowered behind a screen into one side of the test box. When the screen was removed, looking time was scored during a 10second period; the object always appeared on the same side that it was placed. In none of these familiarization trials did subjects see the test object pass from one chamber to the other through the window in the center partition. Having received all four familiarization trials, subjects were presented with two test trials. In test trial 1, an object was lowered into one chamber and then the screen put in place, blocking the tamarin's view of the object and test box. After approximately 5 seconds, the screen was removed revealing the object in the same chamber. Test trial 2 was similar, except that the object appeared in the opposite chamber when the screen was removed. Both test trials require the tamarins to generate an expectation about the object's location before the screen is removed. Only the second test, however, represents a potential violation of expectancy. Specifically, only objects that can move on their own have the potential to travel from the original chamber to the opposite one. In the first experiment, we contrasted the tamarins' lookingtime responses (calculated out of a total of 10 seconds) to a live mouse and a
Page 179
Fig. 10.4. Methodology used to test for the cottontop tamarins' understanding of object motion following occlusion from view. All subjects were presented with the four familiarization trials illustrated; the mouse is illustrated here, but all familiarization trials were the same for the other objects except that the clay ball was rolled into the chamber. Subjects then received two test trials, counterbalanced for order of presentation.
cluster of Froot Loops , the children's breakfast cereal; the tamarins have seen mice before and receive Froot Loops as food rewards in other experiments. If tamarins are like us, they should expect the Froot Loops to remain in the original chamber, rather than move on their own to the opposite chamber. Consequently, they should look longer in test 2 than in test 1. In contrast, because mice can move on their own, there should be no difference in looking time between tests 1 and 2. Results showed that on the first familiarization trial, looking time was longer in response to the mouse than to the Froot Loops (Fig. 10.5a). This makes sense since the mouse is alive and moves, whereas the Froot Loops are neither alive nor do they move. By the fourth familiarization trial, however, there were no differences in looking time between the mouse and Froot Loops. In test 1, when both the mouse and Froot Loops stayed in the original chamber, there was no difference in looking time from the final familiarization trial. Similarly, there was no difference in looking time when the mouse appeared in the opposite chamber from which it was originally placed (test 2). However, when the Froot Loops appeared in the opposite chamber, there was a statistically significant increase in looking time. These results suggest that the pattern of movement observed for the mouse was expected, that its appearance in a novel loca
Page 180
Fig. 10.5. Results from tests on object motion following occlusion. (a) Tests involving a live mouse and a cluster of Froot Loops. (b) Tests involving an externally propelled clay ball and a selfpropelled clay face. (c) Tests involving a live Indonesian tree frog and furry toy monkey that vibrated in place. Open symbols refer to cases where the object remained in the original location; closed symbols refer to cases where the object appeared to move from its original location to the opposite chamber of the box. Mean looking time (yaxis) was scored framebyframe and then converted to seconds; standard deviations shown. Plevels indicated in parentheses.
Page 181
tion following occlusion from view was consistent with the tamarins' expectations. In contrast, when the Froot Loops appeared to move behind the screen, the looking time data suggest that the tamarins' expectations were violated. Although the tamarins are clearly making a distinction between the mouse and Froot Loops, the nature of this distinction is unclear. Consequently, the connection between object representation and expectation remains unclear. To reduce the number of factors that could be guiding the tamarins' expectations, four more conditions were run. The first two involved a contrast between a ball of clay propelled into one chamber by a human experimenter and a clay face that could move on its own due to a magnet embedded inside and a second magnet concealed below the box and operated by a hidden experimenter; the clay face was definitionally selfpropelled because it always moved from a resting position and in the absence of an external agent. Although the clay ball moved, the tamarins' response suggested that they did not expect it to be in the opposite chamber. Specifically, looking time dropped from the first to the fourth familiarization and stayed low when the clay ball appeared in the original chamber during the test trial (Fig. 10.5b). When the clay ball appeared in the opposite chamber, however, looking time increased significantly. Interesting to note, the same pattern of results were obtained for the clay face, an unanticipated finding given that the object was selfpropelled and thus, could readily move from one chamber to the other. Given that the tamarins' expectations were not determined by selfpropelled motion alone, two final conditions were run. Specifically, we were interested in testing the hypothesis that tamarins generate different expectations about animate and inanimate objects. If this hypothesis is correct, than all animate objects should be clustered into one category and all inanimate objects, including those that move on their own, should be clustered into a different category. The two final objects were an Indonesian tree frog and a furry toy monkey that moved in place. The tree frog obviously looked different from the mouse, but importantly, also moved differently, taking advantage of all three planes of the chamber. The toy monkey appeared more animate than the clay face, but rather than moving on its own around the chamber walls, it merely vibrated in place. As pointed out by D. Premack and A.J. Premack (1995), although the object is selfpropelled, it is not goal directed, and thus, should lead observers to different expectations. The tamarins responded to the frog in the same way that they responded to the mouse (Fig. 10.5c). Following familiarization, looking time stayed the same or decreased slightly in response to the frog appearing in either the original or the opposite chamber of the test box. In contrast, looking time increased when the furry toy monkey appeared in the opposite chamber, but stayed the same when it appeared in the original chamber.
Page 182
These experiments represent only a first step in our understanding of the relation between object representation and the generation of expectations concerning displacement, especially given the relatively small number of stimuli used in each object category. Tamarins appear to set up different expectations depending on whether the object is animate or inanimate. If an object is animate, they apparently assume that it can move on its own. In contrast, even if an object is selfpropelled, but inanimate, they do not expect it to move to a new location when out of sight. There are several remaining puzzles, however. For example, what kinds of expectations do they form for an immobile, but animate object, something like an anesthetized mouse? What about an animate object that always moves away from the partition when placed in the test box? In this case, subjects witness an object with animacy, selfpropelled motion, and goal directedness. But if its goal is to run to the wall opposite the partition, then following occlusion, it would be surprising to see the object in the opposite chamber. Conversely, what expectations would be generated for an inanimate object that only moves toward the center partition, and thus, appears to have as its goal moving to the other chamber? Finally, do tamarins generate expectations from motion that is related to particular types of goals? For example, if an observer tamarin witnessed an actor tamarin enter a box with both desirable food and a plastic cube, the observer should expect the actor to remove the food but leave the cube alone. Conversely, if a ball rolled into the box, one would expect both the food and cube to remain in place. Such observations would indicate that the tamarins generate goalappropriate expectations about other animate objects, perhaps especially, members of their own species. Recent experiments by my graduate student, Laurie Santos (1997), provide some support for these predictions. Expectations about Mental States A central problem in current studies of cognitive development and cognitive evolution is when the child acquires the capacity for attributing mental states, as well as when the human species acquired it. As several chapters in this volume point out, there are different components to this problem, with different developmental and evolutionary trajectories (BaronCohen, 1995; Carruthers & Smith, 1996; Gopnik & Meltzoff, 1997; Leslie, 1994; Whiten, 1994). What I focus on in this last section is a methodological point that may have significant conceptual ramifications for research on mental state attribution in humans and animals. Most experimental work on primates has failed to find convincing evidence of mental state attribution (Cheney & Seyfarth, 1990a, 1990b; Heyes,
Page 183
1998; Povinelli & Eddy, 1996; D. Premack & Woodruff, 1978; Tomasello & Call, 1997; for potential exceptions, see Gomez, 1998; Whiten, in press). In contrast, observations from the lab and field suggest that primates, especially the great apes, readily engage in behaviors that appear to require mind reading (Byrne & Whiten, 1990; Cheney & Seyfarth, 1990a). As several authors have argued (Cheney & Seyfarth, 1990a; Heyes, 1998; Povinelli, 1993; Tomasello & Call, 1997; Whiten, in press), however, other experiments are needed before we close the door on this chapter of the human evolution story. To this end, work in our lab has picked up the thread from work on human infants and children, using the expectancy violation procedure as a tool. One potential advantage of this procedure over previously published studies of theory of mind in primates is that it involves no training, and thus, is comparable to studies run on human children; for a recent discussion of this point, and a comparable set of experiments, see Gomez (1998) and Whiten (in press). The classic theory of mind test is Wimmer and Perner's (1983) false belief procedure. Although this task typically requires language, a recent set of studies by Clements and Perner (1994; Perner & Clements, in press) suggest another route into the child's knowledge. Specifically, in the first experiments, Clements and Perner ran several versions of the SallyAnne puppet show (Wimmer & Perner, 1983). In addition to asking the child about Sally's search patterns, they also filmed the child's eyes. The key finding was that 3yearolds looked to the correct box first, but then verbally indicated the wrong box. That is, they looked to the box where Sally left her ball, but then reported that she would search in the opposite box, the one where Anne placed it. Based on these results, and a series of followup experiments, Clements and Perner suggested that at the age of 3 years, children's nonverbal responses (eye gaze, reaching) indicate implicit understanding of false beliefs. It is not until 4 years, however, that they have an explicit understanding as revealed by their linguistic responses. Although it is unclear whether the implicitexplicit distinction represents the most appropriate account of the data, one point is clear: The child's eyes reveal something different from her voice. And given that the eyes reveal something about what the child knows or expects (D. Premack & A.J. Premack, 1995), we now have a tool to use with nonhuman primates, and for that matter, any nonhuman animal that will sit and watch a display. A test using the logic of the expectancy violation procedure was run with 10 adult cottontop tamarins (Hauser & Santos, 1998). The aim of this experiment was first, to follow the key design elements of the SallyAnne test, and second, to determine whether tamarins form different expectations about human action based on what they and the human actor appear to know; knowledge in this situation may come from attributing mental states, from tracking behavioral contingencies, or from some com
Page 184
bination of these and other factors. The experimental room consisted of a Plexiglas test box with a tamarin subject, a table with two boxes, a human experimenter, and a human actor (Fig. 10.6a); a video camera filmed the tamarin's face and thus, the visual record of the trial excluded information about the condition and objects displayed. Looking time was scored out of a 10second period. Two observers scored the tapes using framebyframe analyses, and achieved an interobserver reliability score of 0.92. In all three conditions of the first experiment, the actor walked into the room, sat down at a table with a box on each side, took three bites from an apple and then handed a piece of apple to the tamarin. The actor then placed the apple into one of the boxes. In condition A, the boxes were
Fig. 10.6. Lookingtime experiment of expected action. (a) Experimental setup for the test. (b) Mean lookingtime (seconds) results from first experiment: condition A (transparent box, actor leaves room during apple transfer), condition B (opaque box, actor leaves room during apple transfer), and condition C (opaque box, actor stays in room during apple transfer); standard deviations shown. (c) Mean lookingtime results from second experiment: condition A (opaque box, actor stays in room but turns his back during apple transfer), condition B (opaque box, actor stays in room and closes his eyes during apple transfer), and condition C (transparent box, actor leaves room during apple transfer). Open symbols refer to trials where the actor's search patterns should be considered expected based on either his current beliefs or his prior actions. Closed symbols refer to trials where the actor's search patterns should be considered unexpected.
Page 185
transparent, except for a red cap on one and a blue cap on the other. After the actor placed the apple in, for example, the red box, he left the room. While out of the room, the experimenter removed the apple from the red box and placed it in the blue box; across subjects, we varied the side and box in which the apple was first placed. The actor then returned to the room, sat down, and searched in either the red or blue box. Although the actor failed to witness the apple's transfer, the apple's current location is unambiguous given the transparency of the boxes. There are three possible outcomes for this condition. First, if tamarins are simply interested in the apple, then they might be expected to look longer when the actor searches in the blue box because this search pattern is most likely to lead to food. Second, if the tamarins have formed an expectation about where the actor will search, inferring from previous behavior that he will go to where the apple now is (i.e., the blue box), then looking time should be highest when the actor searches in the red box—a violation of their expectations; the same pattern would also emerge if the tamarins attribute beliefs to the actor. Finally, tamarins may look equally long in response to both search outcomes, possibly for different reasons. Results failed to reveal a statistically significant difference between the two test trials of this condition (Fig. 10.6b). In conditions B and C, opaque boxes (one orange, one pink), replaced the transparent ones. In condition B, the actor left the room after placing the apple in, for example, the orange box. In his absence, the experimenter transferred the apple to the pink box. The actor returned and either searched in the orange box or in the pink box. Using the same logic as for condition A above, if the tamarin is simply interested in the apple's location, then they should look longer when the actor searches in the pink box; this prediction assumes that the tamarin pays no attention to what the actor can see during the test session. The same pattern of looking may, however, emerge if tamarins generate expectations about the actor's goals. Specifically, the actor should search in the orange box because he left it there, and has no knowledge that it was transferred to the pink box. Therefore, when the actor searches in the pink box, the tamarin's expectation should be violated. Finally, the tamarins may show no difference in looking time because both search patterns are equally interesting or equally uninteresting. Results showed that the tamarins consistently looked longer when the actor searched in the box currently holding the apple, even though he could not possibly know that the apple had been moved there (Fig. 10.6b). To reiterate, this pattern is predicted on the basis of both the tamarin's interest in the apple, in addition to the fact that the actor searched in an unexpected location given that they failed to see the transfer. Condition C was the same as B except that the actor stayed in the room and watched as the experimenter transferred the apple from one box to the other. With respect to predictions, this condition parallels, to
Page 186
some extent, condition A. Specifically, if the tamarins simply want the apple, then they should look longer when the actor searches in the box where the apple now is, rather than where it was. In contrast, if the tamarins think that the actor will search in the box currently holding th apple because he saw the transfer, then searching in the original box represents a violation. Thus, the only difference between conditions B and C is that in C, the actor witnesses the apple's transfer and in B he does not. Results showed that the tamarins consistently looked longer when the actor searched in the original box, the box without an apple (Fig. 10.6b). This result appears to rule out the possibility that tamarins simply look longer when the actor searches in the box holding the apple, independently of his prior actions. The differences in looking time noted were statistically significant using both parametric and nonparametric tests. Given the significant differences obtained in conditions B and C, it was important to determine the factors that might cause tamarins to fail, and to determine more precisely the basis for the tamarins' expectations. Results from Povinelli and Eddy's (1996) work on chimpanzees, Cheney and Seyfarth's (1990a, 1990b) studies of macaques, and some of our own unpublished work with the tamarins (Santos, 1997) suggest that primates may not understand the distinction between seeing and knowing, even though they do follow the direction of eye gaze. We thus reran two different versions of condition C. In the first, the actor placed the apple in the orange box and then turned around 180 degrees so that his back faced the tamarins. The apple was then transferred by the experimenter to the pink box, the actor turned around, and then searched in either the original orange box, or the pink box holding the apple. In the second version, the actor placed the apple in the orange box and then closed his eyes as the experimenter transferred the apple to the pink box. The actor opened his eyes and searched for the apple in one of the two boxes. In addition, we ran the transparent box condition (A) again, but after the back turned and eyes closed conditions; our aim here was to determine whether the lack of a significant difference in the previous run was due to a ceiling effect, corresponding to the first two trials of the session. In none of these conditions did we find a significant difference in looking time between tests, although there was a slight tendency in the transparent box condition for subjects to look longer on the test in which the actor searched in the visibly empty box (Fig. 10.6c). How are we to interpret these findings? Let us start with conditions B and C of the first experiment, leaving aside the transparent box condition. To repeat, the only critical difference between these conditions was whether the actor was present or absent when the experimenter transferred the apple from one box to the other. Although the actor's search patterns were always the same, looking in the box where he placed the apple or the alternative,
Page 187
the tamarins showed strikingly different looking times between conditions B and C. In condition B, the actor should expect to find the apple in the box where he placed it. He should thus search in this box. However, because the experimenter moved the apple in his absence, the actor holds a false belief. In contrast, the actor witnesses the transfer in condition C, and thus should alter his belief about the apple's location. Here, we have interpreted the actor's behavior in terms of his mental states, his true and false beliefs. The tamarins' looking time responses in conditions B and C of Experiment 1 are consistent with such an interpretation, but there are at least two alternative accounts, each imputing different computational capacities. Let us rerun the video on the experiment. Here is what the tamarin sees. An actor enters the room and places the apple in one box. In condition B he leaves. When he returns, a plausible expectation would be for the actor to return to the box that was last attended to and physically acted on, in this case, the box where he left the apple. Similarly, in condition C, he should search in the box that was last attended to and physically acted on, but in this case, it is the box that holds the apple because the experimenter moved it while the actor watched. If the actor searches in a different box, then this creates a mismatch based on the actor's prior attentional and behavioral focus, not his mental states. In both cases, expectations are formed, but they are derived from cues that are directly observable; this contrasts with the human adult's capacity to generate expectations based on unobservable mental states. For conditions B and C of the first experiment then, we cannot distinguish between a mentalistic interpretation and a behavioristic one. We are now in the process of running a second set of experiments where attentional and behavioral cues are generally ruled out. Specifically, after the actor places the apple in one box, he physically touches and looks at both boxes before the apple is transferred. Thus, the tamarins can no longer use the actor's last attentional or behavioral focus to generate expectations about subsequent actions. In the second set of experiments, the actor fails to witness the apple's transfer either because his back is turned or because his eyes are closed. Under these conditions, there are no significant differences in looking time between test trials. If the tamarins were simply using the actor's attentional or behavioral focus prior to the apple's transfer, then they should have looked longer when he searched in the box currently holding the apple; this is the same prediction as we made for condition B of experiment 1. In both tests, the actor attends to and touches the box where he placed the apple prior to closing his eyes or turning his back. If the tamarins use the actor's attention or behavior to predict subsequent action, then searching in the box where the apple now is represents a mismatch. Further work is therefore needed to clarify the role of eye gaze, as opposed to action cues provided by the head and eyes.
Page 188
The transparent box condition is difficult to interpret. A ceiling effect due to order of presentation is really not likely, given that the same general patterns were obtained when this condition was run first in the session as well as last. Furthermore, subjects did not look for the entire 10second observation period. What is clear, and somewhat surprising, is that subjects looked as long when the actor searched in the empty box as when he looked in the box with the apple. Given that the actor previously handed the tamarin some apple, searching in the transparent box with the apple is certainly of interest. In contrast, we can only guess why the tamarins might be interested in the actor's search into the empty box. Specifically, this represents a violation of what the actor is meant to be looking for, either because of his beliefs or because of his prior attentional and behavioral focus. At present, we cannot resolve this problem. To summarize, I argue that Experiment 1 shows that the tamarins form some expectations about a human actor's goals. This conclusion depends, of course, on the ability to make inferences about an individual's expectations from the lookingtime measure. When the actor is out of the room, they expect one thing, and when he is in the room and tracking the movement of the apple, they expect another. In Experiment 2, however, there were no differences across tests. These conditions (back turned, eyes closed) are intermediate between those presented in the first experiment in that the actor is present, but not tracking the movement of the apple. I tentatively conclude with a comment about the difference between expectation and belief. The content of an expectation is simply the relation between some context or condition (an apple placed in a box), an action or event (experimenter transfers apple to new box in actor's absence), and a predictable outcome (actor searching in box where he left apple). As such, expectations differ importantly from beliefs in that they lack explicit propositions as their content (''the apple is in the pink box"); beliefs do not depend on actions or consequences, although one can certainly set them up in this way (e.g., "Because I left the apple in the pink box, I believe that I will find it there if I search"). Given this distinction, we believe that our experiments provide evidence that tamarins generate correct expectations about an actor's behavior based on coarsegrained cues associated with knowledge, be they attentional or behavioral. These tests are silent, however, about mental states such as beliefs. Expectancy Violation, Looking Time, and Knowledge Returning to the quote taken from his book The Elm and the Expert, Fodor (1994) attempts to detail what makes our minds so clever, dividing his task into an analysis of informational semantics, conceptual content, mental states, and causality. It would be foolish to debate Fodor's evolutionary
Page 189
point about our mental sophistication. How could anyone disagree? Rather, the interesting challenge is to work out the extent to which humans and nonhumans differ with respect to the causal relations between representation and expectation. In this essay, I have discussed data from three empirical domains that I believe bear on this issue, and that add to the growing literature in the area (e.g., Cheney & Seyfarth, 1990b; Povinelli, 1993; Tomasello & Call, 1997; Whiten, 1994). I end with a few conceptual and methodological comments. What I have argued, and defended using different kinds of empirical findings, is that we can learn a great deal about animal representation by looking at the kinds of events that violate their expectations. This is not a new idea. Years ago, David Premack (1986) raised the intriguing possibility that an ant might find it odd to see a ball roll off a table and then shoot straight to the ceiling. Although a bit heretical, one might even argue that the Skinnerians foresaw this idea in describing the behavior of pigeons that developed "superstitious" responses when reinforced under a completely random schedule. Were they looking for something that was consistent with reinforcement, something that could be expected, predicted? Independently of their significance, the connection between expectation and representation has only recently been put to experimental test using the expectancy violation procedure as a critical tool. A Skinnerian might leap up and argue that what I have referred to as representation and expectation is merely a stimulusresponse connection. Consider, as one example, the experiment on object displacement following occlusion. What we need to explain is why, when the object has apparently moved from one chamber to the other, looking time is longer for some objects but not others. On the stimulusresponse account, one might argue that for some objects, movement to a novel location has been observed. Although all objects except the mouse and Froot Loops were novel, perhaps the tamarins simply generalized to other objects. What is most interesting, of course, is the process of generalization, especially in the absence of training. In particular, what features were used to cluster the mouse and frog into one category and all the other objects into another? Simple features such as color, shape, and texture clearly fail to account for the lookingtime data. So too do other more complex properties such as motion or even selfpropelled motion. The clay face was selfpropelled but the tamarins looked longer when it appeared in the novel chamber. Although considerably more work is necessary, the results suggest that an animateinanimate distinction underlies the differences in looking time. This distinction is conceptually abstract and it is unclear on the Skinnerian account how it would be constructed or used to predict object location. A critical assumption underlying the work presented here is that the lookingtime measure tells us something important about what an organism
Page 190
knows, and more specifically, how it uses such knowledge to generate expectations. The problem lies, of course, in working out the nature of such knowledge. In some experiments, longer looks indicate that the organism has detected a violation, whereas in other experiments it indicates that the organism has detected a possible or consistent event. Let me illustrate with a few examples. In several experiments by Kuhl and Meltzoff (1988), gaze direction and looking time are scored while infants are presented with information from two modalities. For example, when 3montholds see two faces on a monitor producing articulatory gestures while a speaker plays one sound, infants look to the face that is articulating the sound played; in essence, they look for an audiovisual match. When an infant sucks on a pacifier with a particular configuration, they look longer at the pacifier that matches this configuration than ones that do not (Meltzoff & Borton, 1979). In an experiment by Dodd (1979), infants looked longer when a talking face was synchronized to the voice track than when it was desynchronized; in fact, infants tended to look away from the asynchronous display. Similarly, infants will look longer when the number of dots on a screen matches the number of tones presented, or when the sound of a bouncing ball hitting a surface is synchronized to an animated sequence of a bouncing ball (Spelke, 1979; Starkey & Cooper, 1980; Starkey, Spelke, & Gelman, 1990). In contrast to most studies of crossmodal perception, studies that focus on a single modality tend to find that infants look longer when a violation (i.e., mismatch) has been detected. At present, it is unclear whether such apparent differences between uni and multimodal studies are real, and if so, what the underlying mechanisms are. The main point, however, is that looking time is being used to infer different kinds of knowledge or understanding. In addition to differences in the interpretation of lookingtime data are differences between studies using looking time as opposed to some other response assay, typically one involving a motor response such as reaching. Thus, a variety of expectancy violation experiments have suggested that infants understand object permanence at a far earlier age than revealed by reaching procedures (Baillargeon, 1995). As Gopnik and Meltzoff (1997) argued, however, it is unclear whether the difference lies in what the child knows or what the experiment is capable of extracting, or some combination of the two. These problems have yet to be resolved. A common question raised for lookingtime studies is why it is necessary to run familiarization or habituation trials in order to demonstrate the detection of a violation. Although the argument for running familiarization or habituation trials is, as mentioned earlier, to remove the effects of novelty, it should be possible to run them without such trials. The tamarin experiments most recently discussed represent one possibility. A second
Page 191
possibility that we have piloted in our lab is to habituate one group of individuals to trials that represent events consistent with physical or psychological principles of the world and habituate another group to trials that represent violations. The prediction here is that subjects should look longer on the first trial of a display showing a physical or psychological violation, and should take longer to habituate than subjects presented with displays that are consistent with such principles. In a test of simple arithmetic, comparing 1 + 1 = 2 versus 1 + 1 = 1, we found support for these predictions (Hauser, 1997b). Due to its simplicity, this approach has several merits, especially in cases where the organism's attention span is short. Furthermore, the habituation procedure is a potentially powerful tool for exploring the relative plasticity of an organism's expectations. In cases where the expectation is based on an innately specified, core principle, one might expect the time course to full habituation to be long. In contrast, where the expectation is based on a recently learned experience (e.g., all red cups hold liquid), one might expect the time course to habituation to be short. Thus, following the lead of several evolutionary psychologists (Cosmides & Tooby, 1994; Pinker, 1997), we should begin to look at the kinds of regularities that both human and nonhuman animals have confronted over their evolutionary past for herein lie some of the best kept secrets for unravelling the domains of knowledge that are innately specified. Given the work reviewed here, and in the chapters by Tomasello (chap. 4, this volume) and Povinelli (chap. 11, this volume), I suggest that humans share with a wide variety of animals some of the core principles underlying object and spatial knowledge. They differ, however, when it comes to social knowledge, knowledge that in humans at least has been refined by a capacity to attribute mental states to others. Finally, I return to a problem raised earlier, of understanding why a violation has occurred as opposed to simply detecting a violation. It seems clear that lookingtime data will not help resolve this distinction. When we watch a magician violating physical principles, we are not only awed, but curious about the gimmick. If we were allowed up on stage, we would presumably want to snoop around, to figure out how the trick was pulled off. To work out whether other organisms are inspired by similar curiosities, other experiments will be necessary. By asking such questions, we are certainly more clever than they, but we need to better understand the nature of such cleverness, and how it evolved. References Allen, C., & Bekoff, M. (1997). Species of mind. Cambridge, MA: MIT Press. Baillargeon, R. (1995). Physical reasoning in infancy. In M. Gazzaniga (Ed.), The cognitive neurosciences (pp. 181204). Cambridge, MA: MIT Press.
Page 192
BaronCohen, S. (1995). Mindblindness. Cambridge, MA: MIT Press. Byrne, R., & Whiten, A. (1990). Tactical deception in primates: The 1990 database. Primate Report, 27, 1101. Carey, S., & Spelke, E. S. (1994). Domainspecific knowledge and conceptual change. In L. Herschfeld & S. Gelman (Eds.), Mapping the mind: Domain specificity in Cognition and culture (pp. 169201). Cambridge, England: Cambridge University Press. Carruthers, P., & Smith, P. K. (1996). Theories of theories of mind. Cambridge, England: Cambridge University Press. Caryl, P. G. (1979). Communication by agonistic displays: What can games theory contribute to ethology? Behaviour, 68, 136169. Cheney, D. L., & Seyfarth, R. M. (1988). Assessment of meaning and the detection of unreliable signals by vervet monkeys. Animal Behaviour, 36, 477486. Cheney, D. L., & Seyfarth, R. M. (1990a). Attending to behaviour versus attending to knowledge: Examining monkeys' attribution of mental states. Animal Behaviour, 40, 742753. Cheney, D. L., & Seyfarth, R. M. (1990b). How monkeys see the world: Inside the mind of another species. Chicago: Chicago University Press. Cheney, D. L., Seyfarth, R. M., & Silk, J. (1995). The responses of female baboons (Papio cynocephalus ursinus) to anomalous social interactions: Evidence for causal reasoning? Journal of Comparative Psychology, 109, 134141. Clements,.W. A., & Perner, J. (1994). Implicit understanding of belief. Cognitive Development, 9, 377395. Cosmides, L., & Tooby, J. (1994). Origins of domain specificity: The evolution of functional organization. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domainspecificity in cognition and culture (pp. 85116). Cambridge, England: Cambridge University Press. Dawkins, R., & Krebs, J. R. (1978). Animal signals: Information or manipulation. In J. R. Krebs & N. B. Davies (Eds.), Behavioural ecology (pp. 282309). Oxford, England: Blackwell Scientific Publications. Dennett, D. (1996). Kinds of minds. New York: Basic Books. Dodd, B. (1979). Lip reading in infants: Attention to speech presented in and outofsynchrony. Cognitive Psychology, 11, 478484. Dretske, F. (1995). Naturalizing the mind. Cambridge, MA: MIT Press. Fodor, J. A. (1994). The elm and the expert. Cambridge, MA: MIT Press. Gelman, R. (1990). First principles organize attention to and learning about relevant data: number and the animateinanimate distinction as examples. Cognitive Science, 14, 79106. Gergely, G., Nádasdy, Z., Csibra, G., & Biró, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165193. Gomez, J. C. (1998). Assessing theory of mind with nonverbal procedures: Problems with training methods and an alternative 'key' procedure. Behavioral and Brain Sciences, 21, 119120. Gopnik, A., & Meltzoff, A. (1997). Words, thoughts, and theories. Cambridge: MIT Press. Hauser, M. D. (1996). The evolution of communication. Cambridge: MIT Press. Hauser, M. D. (1997a). Minding the behavior of deception. In A. Whiten & R. W. Byrne (Eds.), Machiavellian intelligence II (pp. 112143). Cambridge, England: Cambridge University Press. Hauser, M. D. (1997b). Tinkering with minds from the past. In M. Daly (Ed.), Characterizing human psychological adaptations (pp. 95131). New York: Wiley. Hauser, M. D. (1998a). Expectations about object motion and destination: Experiments with a nonhuman primate. Developmental Science, 1, 3138. Hauser, M. D. (1998b). Functional referents and acoustic similarity: Field playback experiments with rhesus monkeys. Animal Behaviour, 55, 16471658.
Page 193
Hauser, M. D., & Carey, S. (1998). Building a cognitive creature from a set of primitives: Evolutionary and developmental insights. In D. Cummins & C. Allen (Eds.), The evolution of mind (pp. 52106). Oxford, England: Oxford University Press. Hauser, M. D., & Marler, P. (1993a). Foodassociated calls in rhesus macaques (Macaca mulatta). I. Socioecological factors influencing call production. Behavioral Ecology, 4, 194205. Hauser, M. D., & Marler, P. (1993b). Foodassociated calls in rhesus macaques (Macaca mulatta). II. Costs and benefits of call production and suppression. Behavioral Ecology, 4, 206212. Hauser, M. D., & Santos, L. R. (1998). Are monkeys mentalists or sophisticated behaviorists? Results from a nonlinguistic falsebelief test. Manuscript in preparation. Heider, F., & Simmel, M. (1944). An experimental study of apparent behavior. American Journal of Psychology, 57, 243259. Heyes, C. M. (1998). Theory of mind in nonhuman primates. Behavioral and Brain Sciences, 21, 101114. Hood, B. (1995). Gravity rules for 24 year olds? Cognitive Development, 10, 577598. Krebs, J. R., & Dawkins, R. (1984). Animal signals: Mindreading and manipulation. In J. R. Krebs & N. B. Davies (Eds.), Behavioural ecology (pp. 380402). Oxford: Blackwell. Kuhl, P. K., & Meltzoff, A. N. (1988). Speech as an intermodal object of perception. In A. Yonas (Ed.), Peceptual development in infancy (pp. 235256). Hillsdale, NJ: Lawrence Erlbaum Associates. Leslie, A. M. (1994). ToMM, ToBy, and Agency: Core architecture and domain specificity. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domainspecificity in cognition and culture (pp. 119148). New York: Cambridge University Press. Meltzoff, A. N., & Borton, R. W. (1979). Intermodal matching by human neonates. Nature, 282, 403404. Michotte, A. (1962). The perception of causality. New York: Basic Books. Millikan, R. (1993). White Queen psychology and other essays for Alice. Cambridge, MA: MIT Press. Mitchell, R. W. (1986). A framework for discussing deception. In R. W. Mitchell & N. S. Thompson (Eds.), Deception: Perspectives on human and nonhuman deceit (pp. 340). Albany, NY: SUNY Press. Munn, C. (1986). Birds that 'cry wolf.' Nature, 319, 143145. Perner, J., & Clements, W. A. (in press). From an implicit to an explicit "Theory of mind". In Y. Rossetti & A. Revonsuo (Eds.), Dissociation BUT interaction between conscious and nonunconscious processing. Amsterdam: John Benjamins. Pinker, S. (1997). How the mind works. New York: Norton. Povinelli, D.J. (1993). Reconstructing the evolution of mind. American Psychologist, 48, 493509. Povinelli, D.J., & Eddy, T.J. (1996). What young chimpanzees know about seeing. Monographs of the Society for Research in Child Development, 247(55, Serial No. 242). Premack, D. (1986). Gavagai! or the future history of the animal language controversy. Cambridge, MA: MIT Press. Premack, D. (1990). The infant's theory of selfpropelled objects. Cognition, 36, 116. Premack, D., & Premack, A.J. (1995). Origins of human social competence. In M. Gazzaniga (Ed.), The cognitive neurosciences (pp. 205218). Cambridge, MA: MIT Press. Premack, D., & Premack, A.J. (1997a). Infants attribute value +/ to the goaldirected actions of selfpropelled objects. Journal of Cognitive Neuroscience, 9, 848856. Premack, D., & Premack, A.J. (1997b). Motor competence as integral to attribution of goal. Cognition, 63, 235242. Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a theory of mind? Behavioral and Brain Sciences, 4, 515526. Santos, L. (1997). Precursors to a theory of mind: Insights from a nonhuman primate. Undergraduate Honors Thesis, Harvard University, Cambridge, MA.
Page 194
Spelke, E. S. (1979). Perceiving bimodally specified events in infancy. Developmental Psychology, 15, 626636. Spelke, E. S. (1985). Preferential looking methods as tools for the study of cognition in infancy. In G. Gottlieb & N. Krasnegor (Eds.), Measurement of audition and vision in the first year of postnatal life (pp. 85168). Norwood: Ablex. Spelke, E. S. (1994). Initial knowledge: Six suggestions. Cognition, 50, 431445. Spelke, E. S., Vishton, P., & von Hofsten, C. (1995). Object perception, objectdirected action, and physical knowledge in infancy. In M. Gazzaniga (Ed.), The cognitive neurosciences (pp. 165179). Cambridge, MA: MIT Press. Starkey, P., & Cooper, R. (1980). Perception of numbers by human infants. Science, 210, 10331035. Starkey, P., Spelke, E. S., & Gelman, R. (1990). Numerical abstraction by human infants. Cognition, 36, 97127. Struhsaker, T. T. (1967). Auditory communication among vervet monkeys (Cercopithecus aethiops). In S. A. Altmann (Ed.), Social communication among primates (pp. 281324). Chicago: Chicago University Press. Tomasello, M., & Call, J. (1997). Primate cognition. Oxford, England: Oxford University Press. Whiten, A. (1994). Grades of mindreading. In C. Lewis & P. Mitchell (Eds.), Children's early understanding of mind: Origins and development (pp. 4770). Hillsdale, NJ: Lawrence Erlbaum Associates. Whiten, A. (in press). Chimpanzee cognition and the question of mental rerepresentation. In D. Sperber (Ed.), Metarepresentation. Oxford, England: Oxford University Press. Whiten, A., & Byrne, R. W. (1988). Tactical deception in primates. Behavioral and Brain Sciences, 11, 233273. Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception. Cognition, 13, 103128.
Page 195
Chapter 11— Social Understanding in Chimpanzees: New Evidence From a Longitudinal Approach Daniel J. Povinelli University of Southwestern Louisiana Behavior, it would now seem, is not necessarily an open window into the mind. If it were, humanity's quest to end its intellectual and spiritual isolation from the rest of the natural world would have ended with Romanes' (1882) publication of Animal Intelligence over a hundred years ago. For there, Romanes took up Darwin's (1871/1982) claim that there was "no fundamental difference" (p. 445) in the mental functioning of humans and other species. Romanes (1882) proceeded to catalogue examples of animals behaving in ways that suggested not only that the animals had minds, but that they were also reasoning about minds in both themselves and in others. In lieu of being able to pin psychological structures of animals on the dissecting table, Romanes (1883) sought what he saw as the nextbest alternative: collect and catalogue their behaviors. These behaviors, Romanes surmised, could serve as behavioral ambassadors of the hidden psychological structures of animals. After all, he reasoned, we know the mental activities that accompany the production of our own behavior, and so, through analogy we ought to be able to infer the mental activities of animals. Similarity in behavior, Romanes concluded, was evidence of similarity in mind. Of course, Darwin and Romanes were really just putting into practice what Hume (17391740/1978) had proclaimed a century earlier: Where humans and animals display similar behavior, so too must they experience similar mental states. " 'Tis from the resemblance of the external actions of animals to those we ourselves perform," Hume observed, "that we judge
Page 196 1
their internal [actions] likewise to resemble ours . . ." (p. 176). In this chapter, I return to the problem that Hume claimed to have solved, namely, how to assess the mental states that accompany similar behaviors in different species. I examine several cases from our research program in which chimpanzees and human children exhibit strikingly similar behavior, and ask whether such behavioral similarity justifies an inference of comparable psychological similarity. Despite the confident urgings of Hume, Darwin, and Romanes, I provide a simple, theoretical framework for suspecting that it does not. Chimpanzee Social Understanding Chimpanzees are remarkably social creatures. Even casual observations of their interactions with each other reveal a complexity and subtlety that startles the uninitiated. For example, consider the actions of one of our female chimpanzees, Brandy, who discovers a banana poking out of some straw in a secluded area of her compound. Just as she is about to seize the banana, in the distance she notices the alpha male of the group, Apollo, approaching her. Instead of reaching for the banana, Brandy turns her back on it. Apollo greets her with a chimpanzee hug, glances around, and then departs. Once he is out of sight, Brandy removes the banana from its hiding place and consumes it, quietly foodbarking in pleasure. Menzel (1974) described the microgenesis of such behaviors many years ago during a landmark set of studies of young chimpanzees, and in the ensuing decades many primatologists have urged us to consider such episodes as evidence that chimpanzees see one another as more than just hairy collections of limbs, torsos, and facial expressions moving through space. Surely, they might insist, Brandy sees Apollo as a subjective being, alert with his own firstperson experiences of attending, wanting, knowing, and believing. Indeed, her reactions would seem to provide prima facie evidence that she not only understands that Apollo possesses most or all of these mental states, but that she can manipulate his mental states through her actions. With Hume's twocenturyold proclamation still echoing, we are encouraged to believe that it could not be otherwise. Similarity in behavior guarantees similarity in mind. I, too, was once convinced by such observations that chimpanzees possess a mentalistic understanding of others very similar to our own. Indeed, the 1
Russell (1948) is perhaps better known for his exposition of the argument by analogy, although he developed it not to justify our belief in the minds of other species, but to justify our belief in the minds of other humans. The logical structure of his argument, however, parallels Hume's. Through introspection we can know that our own mental state of type A caused our own bodily act of type B, and therefore if we see another person engage in a bodily act B, we can reasonably infer the presence of mental state A.
Page 197
description of chimpanzee social maneuvering in de Waal's (1982) Chimpanzee Politics was enough to convince even the skeptics that chimpanzees navigated their way through a complex social milieu by reasoning about the mental lives of those around them (see also de Waal, 1986; Goodall, 1986). Soon, reports surfaced of various species of monkeys acting in deceptive ways, and by 1988 Whiten and Byrne had published a report in the Behavioral and Brain Sciences that amounted to a veritable taxonomy of "tactical deception" in primates. Indeed, primates were not alone in this regard, and numerous other taxa ranging from avians to elephants have been shown to manipulate each other's behavior in some sense or another (see Mitchell & Thompson, 1986). The flood of species that were soon reported to engage in social deception left me, and others, questioning the utility of spontaneous behaviors (anecdotes) as evidence of animals' appreciation of mental states. But the nature of these critiques were often more polemical than theoretical. Some critiques sought to show how such behaviors could emerge through various forms of simple learning, others admonished primatologists for treating learning as something that only occurred in psychological laboratories, and still others appealed to parsimony to resolve the issue (see commentaries following Whiten & Byrne, 1988). Few, however, directly grappled with the more fundamental philosophical problem that Hume claimed to have solved (for an exception, see Thomas, 1988). Indeed, Hume's solution was a general one. He did not claim that certain kinds of particularly dramatic behavior (such as deception) could be taken as evidence for understanding the mental states of animals; rather, the argument by analogy held that the mental states that attend our behavior cause them, and that we can therefore use the presence of a particular behavior to infer the presence of a particular mental state (see also Romanes, 1882, pp. 12; Russell, 1948, pp. 482486). Deception might tempt researchers more than other kinds of social interactions, but as far as Hume's argument was concerned, the hestitancy to extend such attributions into all aspects of animal behavior simply reflected a weakness at the knees. There were at least two possible reactions to the mounting evidence of deception (and other forms of behavioral similarity) between humans and other primates. On the one hand, it was possible to follow Hume into believing that all of these species were reasoning about mental states (excluding, perhaps, such distantly related species as bees and cockroaches at which our intuitions might balk). But this seemed arbitrary and dissatisfying. On the other hand, it was possible to explain away the impressive natural social interactions of animals by invoking such phrases as "innate dispositions," "hardwired behaviors," "social learning," and "parsimony.'' Although I recognized that this was problematic, I nonetheless blithely assumed that one day we would discover a simple onetoone correspon
Page 198
dence between the evolutionary emergence of social understanding and some asofyet unidentified class of social behaviors. In the meantime, experimental methods were needed to pry nature apart at the seams and discover how the minds of chimpanzees (and other animals) really worked. But slowly, under the weight of literally dozens of experiments that we have conducted to diagnose the nature of chimpanzee social understanding, I have begun to change my opinion about the relation between laboratorybased assessments of chimpanzee social understanding and their spontaneous social interactions. This has not been completely an empirical conversion. Rather, I have begun to consider the merits of a different view of the relation between the evolution of complex social behavior and the evolution of social understanding. This, in turn, has led me to rethink the relation between the spontaneous social behavior of animals and their behavioral reactions in carefully designed experiments. Finally, this has led me to construct a framework that reconciles the two, and simultaneously demonstrates why using the spontaneous behavior of animals to infer their understanding of mental states is problematic. In short, this framework provides a principled alternative to the argument by analogy in the case of the representation of mental states. This framework suggests that there may be a far more complex and subtle relation between behavior and cognition than Hume, Darwin, or Romanes ever suspected. So here is the problem that shall occupy me for the remainder of this chapter: A human and a member of some other species engage in a particular behavior to cope with a social situation; the behaviors accomplish the same end result, and the organisms even tailor the behaviors similarly to respond to similar dynamic features of the situation. Given this degree of similarity, are we justified in assuming that the secondorder mental states that often accompany the human execution of the behavior are present in the other species? Does similarity in behavior guarantee comparable similarity in psychology? The problem is a general one, but the case of humans and chimpanzees is especially poignant because here the functional and structural dimensions of similarity converge. After all, not only do these two species possess similar behaviors that serve similar functions, but also, because of their close phylogenetic relationship, they look very similar as they execute them. What Chimpanzees Understand about Seeing The dimensions of behavioral similarity I explore revolve around actions that in humans are often described as "referential"—actions such as pointing and gazing. For a number of years, my colleagues and I have been examining how chimpanzees understand such behaviors. The subjects of most of these
Page 199
studies have been seven chimpanzees that were raised together from birth in a nursery. Admittedly, as far as apes go, our chimpanzees live a fairly unique lifestyle. They play, groom, fight, nap, and otherwise socialize with each other all day long in their indooroutdoor compound, and at the same time they are constantly exposed to caretakers, staff, and students who interact and play with them in a variety of ways. Embedded into this spontaneous social life with other chimpanzees and humans, each of them is also tested two or three times a day in a specially designed testing facility that is connected to their compound. This facility allows us to test each ape in turn for 1020 minutes at a time. While the others play, one of them is transferred into an outside waiting area, which is connected by a shuttle door to an indoor testing room. Typically, this animal waits outside as a trial is set up indoors. When the shuttle door is opened, the animal is free to enter the lab and respond to the task. A Plexiglas panel separates the apes from the humans. This routine of coming out of the group to be tested was established when our apes were quite young, and they have now participated in this routine so many times that it blends seamlessly into the other predictable routines of their day such as eating and sleeping. Although we have investigated their potential understanding of a wide range of mental states, the most detailed work we have conducted concerns their understanding of the attentional aspect of gaze or "seeing." Our interest in this topic emerged from several directions, but perhaps no one has given this issue more careful experimental consideration than Flavell and his colleagues who conducted an extensive series of studies investigating the timing of young children's developing understanding of seeing (e.g., Flavell, Everett, Croft, & Flavell, 1981; Flavell, Shipstead, & Croft, 1978; Lempers, Flavell, & Flavell, 1977). What chimpanzees (and other nonhuman primates) understand about seeing is of central importance to the problem of this chapter, because in some sense or another "attentional structure" is a defining feature of primate social organization. Many years ago, Chance (1967) noted that primate dominance hierarchies revolve around who is paying attention to whom. Social primates seem to constantly monitor the behaviors of others up the dominance hierarchy, apparently computing whom the more dominant animals are approaching, threatening, or even at whom they are looking. Thus, both monitoring the gaze of others and looking into the eyes of others are a part of the social life of many primates, although the nature and consequences of these exchanges range from extremely hostile (as in many species of Old World Monkeys such as macaques; see Perrett et al., 1990), to extremely affiliative (as in chimpanzees; see de Waal, 1989). We recently provided the first empirical demonstration of gazefollowing in nonhuman primates (see Fig. 11.1). Because of the importance that has been placed on the emergence of gazefollowing in human infants (see
Page 200
Fig. 11.1. Gazefollowing in a 6yearold chimpanzee in response to movement of eyes + head.
BaronCohen, 1994; Butterworth & Cochran, 1980; Butterworth & Jarrett, 1991; Corkum & Moore, 1994), we conducted a series of studies to examine first, whether our apes follow gaze at all, and second, the level of sophistication of whatever gazefollowing system was present (see Povinelli, Bierschwale, & Cech, in press; Povinelli & Eddy, 1996a, 1996c [Experiment 12], 1997). This research has demonstrated a strong similarity in the gazefollowing system of humans and chimpanzees. First, chimpanzees display the gazefollowing response whether it is instantiated by movements of the head and eyes in concert, or simply the eyes alone. Second, chimpanzees will use another's gaze to visually search into space outside their immediate visual field in response to eye plus head/upper torso movement, eye plus head movement, or just eye movement alone. Furthermore, chimpanzees do not even need to witness the shift in another's gaze direction
Page 201
in order to follow it into space outside their immediate visual field. Rather, the gazefollowing response can be triggered when chimpanzees encounter another's head and eye direction that is statically oriented above and behind them. Third, chimpanzees extract specific information about the direction of gaze from others (at least when that gaze occurs in concert with head and upper torso movement). Finally, chimpanzees may also possess at least a tacit understanding of how another's gaze is interrupted by solid, opaque surfaces. In general, these findings reveal that chimpanzees follow gaze with a sophistication comparable to that exhibited by 18month old human infants (see Butterworth & Cochran, 1980; Butterworth & Jarrett, 1991; Corkum & Moore, 1994). More recently, other researchers have demonstrated at least a general capacity for gazefollowing in other nonhman primates (Emery, Lorincz, Perret, Oram, & Baker, 1997; Tomasello, Call, & Hare, in press). Although the sophistication of gazefollowing in chimpanzees and other nonhuman primates may satisfy the part of us interested in the dynamics of social complexity, the philosopher and psychologist in us remains curious, wanting to know what these phenomena reveal about these species' understanding of attention as an internal mental state. As we have seen, Hume would have considered the problem answered as soon as we described the behavior: To the extent that humans reason about attention in such situations, so do other organisms who exhibit the behavior. But in truth there are at least two very different possibilities. First, chimpanzees who follow gaze may do so because they appreciate the underlying attentional states. On the other hand, it is possible that chimpanzees are merely looking where others are looking, without entertaining the idea that the gaze of the other organism is connected to internal visual or attentional experiences (see Povinelli & Eddy, 1996a). Although I run the risk of being misunderstood, there is a need to have succinct labels for these alternative accounts and so I hereafter refer to them as the low and highlevel models, respectively. To summarize, the highlevel model stipulates that chimpanzees form concepts about internal mental states (such as attention) and use these concepts to assist them in interpreting the behaviors that unfold around them. In contrast, the lowlevel model stipulates that chimpanzees reason about behavioral propensities, not mental states. Do Chimpanzees Know That Others See? A Longitudinal Project We have addressed the problem of what chimpanzees understand about seeing from a number of different perspectives, but in the interests of space, I focus on one particular procedure that we have used to diagnose our animals' understanding of seeing over a period spanning their late
Page 202
juvenile, adolescent, and young adult years. In what follows, I highlight some of the major findings of this project (see Povinelli, 1996; Povinelli & Eddy, 1996c; Reaux, Theall, & Povinelli, in press). Year 1: Genuine Versus "AsIf" Understanding We developed the procedure I shall be discussing in order to determine if chimpanzees could appreciate the difference between someone who could see them versus someone else who could not. This procedure evolved from our observations of their spontaneous play behavior. We had repeatedly observed our apes engaging in behaviors in which they used buckets, burlap sacks, pieces of cardboard, or even their own hands to obscure their vision as they moved around their compound. The animals engaged in this behavior so frequently, and seemed to enjoy it so much, that it was tempting to assume that they knew precisely what they were doing: preventing themselves from seeing. This raised the rather straightforward question of how the chimpanzees would react when confronted with familiar caretakers and students who engaged in similar behaviors. In order to test the animals, we capitalized on their natural begging gesture, which they frequently use to "request" food or other desirable objects from us. The gesture consists of an outstretched arm with palm up—a "gimme" gesture. Humans naturally respond to this gesture by handing the subjects what they want (e.g., fruit, cookies, or some other desirable object). We formalized this routine by establishing the following training procedure. First, with an ape waiting outside, we set up a trial indoors by having a familiar person stand or sit just out of reach in front of a hole on the right or left side of the Plexiglas partition. Next, the shuttle door was opened, allowing the subject to enter the test unit and gesture to the person for a treat (Fig. 11.2). If the ape gestured through the correct hole (the one directly in front of the experimenter), the experimenter praised the ape and handed him or her a treat. This process was repeated until the subjects were virtually flawless, thus setting the stage for asking the apes about their understanding of the visual experiences of the people with whom they were interacting. We tested the apes by carefully choreographing several experimental conditions that instantiated the contrast between seeing and not seeing in the most obvious manner possible, while still modeling these conditions after the behaviors in which we had seen the apes engage. In the initial study, we constructed four conditions: blindfolds, buckets, handsoverthe eyes, and back/front (see Fig. 11.3a). These conditions were embedded as probe trials into sessions consisting of the easy, one experimenter trials described previously. Our central question was: Whom would the animals gesture to when confronted with one person who could see them, and
Page 203
Fig. 11.2. Chimpanzee entering test unit and using speciestypical beg ging gesture to request food from a familiar experimenter.
Page 204
Fig. 11.3. (a) Conditions used to test chimpanzees for their understanding of seeing/not seeing (blindfolds, buckets, handsovereyes, back/front). (b) Results of first four trials of blindfolds, buckets, handsovereyes, and back/front conditions. The dotted line indicates level of perfor mance expected by chance responding. ** p < .01.
Page 205
another who could not? The highlevel model predicted that the subjects would gesture to the person who could see them from the very first trial forward; the low level model predicted that the subjects would initially respond at random, and only gradually learn to choose the correct person. Figure 11.3b presents the results of these initial studies. As can be seen, the subjects responded according to the lowlevel model in three of the four conditions (blindfolds, buckets, handsovertheeyes). The animals entered the test unit, measurably paused, and then gestured to one of the experimenters. However, much to our surprise, they were just as likely to gesture to the person who could not possibly see them, as to the one who could! In direct contrast, the animals performed excellently from trial 1 forward on the back/front condition. Here, the animals had no trouble gesturing to the person facing forward. Two possible explanations for this discrepancy presented themselves. One was that the back/front condition was simply the most obvious case of seeing and not seeing—the one easiest for the apes to recognize. However, a more mundane possibility presented itself as well. The apes might have been gesturing to the person facing forward because this was precisely what we had taught them to do in the training phase: Enter the test unit, approach a person who happens to be facing forward, gesture in front of him or her, and get a reward. We attempted to distinguish between these possibilities by confronting the apes with another, equally natural posture that they experience with us and each other daily: someone looking over the shoulder toward them (see Fig. 11.4a). This allowed a precise test of the high and lowlevel explanations of the back/front results from the first experiment. In this new condition the general frontal aspect of both experimenters was not visible, but one of the two could clearly see the animal. We administered this condition, along with the back/front condition as a control, using our standard probe trial techniques. Again, the animals entered the testing unit, paused in reaction to the novel postures, but then proceeded to gesture to the person looking away from them as often as to the person looking over the shoulder in their direction (see Fig. 11.4b). We conducted an extended series of studies which followed up on these surprising results. We created several new conditions for these studies including one involving someone holding a screen in front of the face versus someone else holding one over the shoulder (see Fig. 11.5a). Despite an extended amount of initial experience with the screens in the context of play, the animals initially performed randomly. Slowly, however, as we continued to administer these screen trials, their performances began to improve, until finally they were performing at levels exceeding chance (see Fig. 11.5b). Our two models explained these results in very different ways. The highlevel model could explain these results by arguing that the subjects had finally learned to select the experimenter who could see them.
Page 206
Fig. 11.4. (a) Lookingovertheshoulder condition. (b) Results of first four trials of lookingovertheshoulder condition. The dotted line indicates level of performance expected by chance responding.
Page 207
Fig. 11.5. (a) Screens condition. (b) Results of screens condition with repeated experience. The dotted line indicates level of performance expected by chance responding.
Page 208
In contrast, the lowlevel model argued that the subjects had simply learned a rule such as, ''Gesture in front of the person whose face is visible." In order to test these competing interpretations we readministered all of the conditions to the subjects. Both models converged in that they predicted that the apes would perform well in the buckets, handsovertheeyes, and back/front conditions, but they differed in their expectations about the blindfolds condition. The high level model expected that the apes would do well on this condition because it was another particularly salient example of seeing/not seeing. But if the lowlevel model were correct, and the subjects were merely selecting the experimenter whose face was visible, they ought to continue to respond randomly on the blindfolds condition; after all, in this condition an equal amount of the two faces were visible (see Fig. 11.3). And, in full support of the lowlevel model, this is exactly what our apes did. Finally, we conducted several additional tests contrasting predictions of the low and highlevel models (such as one involving a distracted experimenter versus an attending one, someone with eyes open versus eyes closed, and several conditions designed to examine the relative importance of the eyes versus the face). As before, the lowlevel model continued to generate the most accurate predictions concerning the animals' responses. However, even in these new conditions the animals showed evidence of learning. Indeed, by the end of the initial 14 experiments we conducted, the subjects were able to learn new conditions within 34 trials involving differential feedback. Indeed, several of the final experiments provided confirmation for our idea that the apes were relying on a series of hierarchical rules related to the front, face, and eyes of the experimenters. First, they sought to approach and gesture to someone facing forward. Clearly, the back/front condition easily allowed them to satisfy this rule. On the other hand, if this were not possible (e.g., if both persons were facing forward), then the apes moved on to the next most important rule which was to gesture to the person whose face was visible. In the case of buckets and screens, this rule was easily satisfied. Of least importance was the eyes rule, and indeed, at this point there was only limited evidence that they were even using it. They did perform well on the blindfolds condition (which could be solved by the eyes, but not the face rule), but they showed much weaker evidence of understanding the eyes open/closed condition. Nonetheless, by the end of these studies our seven young chimpanzees were behaving exactly as if they understood something about seeing as a mental event. They would approach two familiar caretakers, look carefully at each one, and then gesture in front of the person who could see them. Indeed, by the end of these experiments, our apes were performing in much the same manner as the 2, 3, and 4yearold children that we tested on these same procedures had performed on their very first trials (see Povinelli & Eddy, 1996c, Experiment 15).
Page 209
We have now arrived at the heart of the problem I promised to address in this chapter: Does the fact that two species engage in similar behavior imply similar accompanying mental states? And does it matter how the behavior arises? For example, is there a psychological distinction between a 2 or 3yearold child who arrives at our laboratory, participates in our seeing/not seeing test, and performs perfectly from trial 1 forward, and our chimpanzees, who, after months of differential feedback, finally learned to do likewise? It is important not to trivialize what our chimpanzees had learned. Although our investigation of their understanding of seeing was best predicted by the lowlevel model, by the end of these tests the apes were using the direction of the face and the presence or absence of the eyes as the bases for their choices. Although the lowlevel model explained their use of these features in terms of a psychological system that reasons about physical postures, here the low and highlevel models converge. After all, even though the highlevel model envisions the operation of a psychological system that interprets these postures in terms of underlying mental states, even this system must make the relevant postural discriminations in the first place. Another way of thinking about this is to realize that whatever the underlying differences in interpretation, our chimpanzees and 2 to 3yearold children were attending to the same physical stimuli when they were making their decisions. Perhaps the problem is best stated from the point of view of the children. Prior to visiting our testing center, these children have had numerous semistructured experiences of seeing/not seeing (in the context of playing with parents and peers). Thus, long before participating in our tests, the children have been confronted with games, and even "real" social experiences, involving the distinctions between seeing and not seeing. Although their reactions to these situations were not yoked to receiving or not receiving a sticker, surely the range of their responses were linked to a range of consequences from their parents and peers. The question I raise is simple. How do these kinds of experiences received by children compare to the kinds of experiences received by our chimpanzees in testing situations? And even more directly, what can such comparisons tell us about the psychological structures that cause and/or accompany the final behavioral outcomes? At this juncture, two separate arguments present themselves. On the one hand, it is possible to question whether the 2 to 3yearold child's performance on our task really reflects an understanding of seeing/attention in the first place. After all, our tests with the children were not nearly as extensive as those with the apes, primarily because even the youngest children performed well from trial 1 forward. However, those initial tests consisted of the conditions that could be solved by the face rule. Perhaps a lowlevel model of their behavior would better predict their reactions on
Page 210
more complicated tests. Although possible, there are independent confirmations of the 2 to 3yearold's understanding of seeing/attention (see Baldwin, 1991, 1993; Gopnik, Meltzoff, & Esterly, 1995; Lempers, Flavell, & Flavell, 1977). So, perhaps too hastily, I move to the second argument, which is of more central importance to this chapter. Given the amount of experience that children have with explicitly created instances of seeing/not seeing by their parents and siblings, how do we know that it is not precisely this experience that allows the child to create the idea of visual attention in the first place? If so, then perhaps our apes' final trials on our tests are more comparable to the children's very first trials. In other words, perhaps chimpanzees simply need sufficient experience to allow them to construct a concept of seeingasattention. Thus, although the lowlevel model best predicted our apes' behavior at each critical testing juncture, this might have been the case only because our tests were chasing the apes' concurrent construction of the idea of attention. Extreme modularity theorists will object immediately, claiming that the insularized nature of social understanding excludes this possibility. Yes, they admit, children have such seeing/not seeing experiences in play, but these experiences are incidental to the ontogenetic unfolding of their understanding of mental states in others. In this view, their knowledge of the mental state of attention matures biologically; it is not constructed through experience. Thus, only the apes' initial trials are diagnostic; what they learn through trial and error is not relevant to the question of cognitive development. Apes either naturally and spontaneously develop an understanding of seeing, or they do not; changes in task performance that occur as the result of feedback should be assigned to an "asif" category of understanding. But such objections derive most of their force through empirically underdetermined theoretical positions. Although there are good reasons for considering stronger and weaker versions of modularity in the development of cognitive structures, there is nothing about our knowledge of cognitive development that forces the modularity view. Indeed, several theorists have argued that experience does play a crucial role in cognitive development. For example, Premack (1988) argued that training chimpanzees to use a symbol for same/different judgments about objects, altered the natural state of the species' cognitive structures in such a manner that they could now engage in abstract analogical reasoning. More recently, Tomasello (1995) has resurrected the argument that exposure to human culture dramatically affects the cognitive development of apes—in this case, the apes' understanding of joint attention (see Tomasello, chap. 4, this volume). If opinions matter, I certainly favor accounts of cognitive development that emphasize the complex epigenetic interactions that occur during development. But in the final analysis, the analogy between the construction of bodily structures (including the brain) on the one hand, and the construction of concepts
Page 211
related to mental states, on the other hand, remains unclear. In fact, the ambiguity of this analogy manifests itself in the context of interpreting the results of our seeing/not seeing experiments. For example, some have interpreted our results as evidence that apes do not understand seeing, whereas others have concluded that our apes' final performance demonstrates that they do. With respect to the latter claim, our apes' ability to learn to gesture to the person who could see them can be taken as evidence that they understand the attentional aspect of seeing in one of two ways: (a) they finally figured out what we were (awkwardly) asking them, or (b) they finally constructed an explicit concept of seeing (which they normally do not need in order to get along in chimpanzee society). In either case, our results can be interpreted to show that reasoning about the mental dimension of seeing is not beyond the capacity of chimpanzees. In summary, then, there are at least three distinct ways in which to characterize the nature of the social understanding connected with our chimpanzees' final, successful performances on the seeing/not seeing tests that we gave them. 1. A first possibility is that before participating in our tests, our apes did not possess a concept of attention. However, through the differential feedback they received, and the through our refinement of the tests, they came to construct such a concept, and indeed, learned one of its sensory bases (e.g., seeing). 2. A second possibility is that our apes entered the tests with a general, amodal conception of attention (perhaps interpreting attention as being governed by proximity, as opposed to sensory channels such as seeing, hearing, touching, etc.). However, again through the feedback that our tests provided, they constructed the notion of visual attention. 3. A final possibility is that our apes neither entered nor exited from our tests with an understanding of the mental state of attention. Rather, they constructed an "asif" understanding of seeingasattention. On this view, our feedback procedures sculpted their behaviors into a form that matched our own. Although there may be theoretical reasons for favoring one account over another, we found very little in our empirical results to exclude any of these possibilities (Povinelli & Eddy, 1996c, p. 134). Year 2: Failure of Retention Thirteen months after we completed the initial series of tests described previously, at least one means of distinguishing between these alternatives serendipitously presented itself. In the context of preparing our apes for a
Page 212
different set of experiments concerning their understanding of joint attention, we returned to our seeing/not seeing protocols, and tested them on the eyes open/closed procedure. To our surprise, rather than finding this easy, the animals were just as likely to gesture to the person who had their eyes open as to the person who had their eyes closed. Indeed, even after 48 trials of this condition, the subjects were still not responding above chance (see Fig. 11.6). At first, we assumed that this was because eyes open/closed was the most subtle condition of all of those we had previously used, and therefore the animals may have never developed a robust understanding of it. Intrigued, we decided to test the apes on the condition with which they had received the most previous experience: screens. Again to our surprise, however, their performance only gradually crept up across the 12 trials we administered to levels just above chance (57% correct; see Fig. 11.6). We were thus forced to consider the possibility that despite the fact that our apes had been 89% correct on their final series of the screens trials a year earlier, they had apparently not consolidated this understanding into a form that would endure a year of participating in other tests. Let me emphasize the significance of this failure of retention by pointing out that our apes were not simply lounging in the sun during the year that intervened between these two longitudinal time points. On the con
Fig. 11.6. Results from tests conducted during Year 2 for eyes open/closed, screens, and back/front conditions.
Page 213
trary, they had participated daily in at least a dozen other experiments, all of which were designed to probe their potential understanding of attention (or other mental states; see Povinelli & O'Neill, in press; Povinelli, Bierschwale, & Cech, in press; Povinelli, Davidson, & Theall, 1997; Povinelli & Eddy, 1996a, 1996b, 1997; Povinelli, Perilloux, Reaux, & Bierschwale, 1998). Indeed, in many of these studies, the role of visual attention played a prominent role. Although these tests provided no better evidence that our apes possessed a genuine understanding of attention (or any other mental state), our fortuitous findings on the longitudinal seeing/not seeing task raised a much broader, and much more interesting, question: How did the experiences on these various tests interact with each other? One might naturally assume that much like young children attending school, such testing experiences would build cumulatively, mutually reinforcing each other in a manner that would assist our apes in homing in on, and better understanding, the questions we were posing to them (using a variety of procedures). Indeed, we carefully planned the nature and sequence of our tests with this assumption in mind. To some extent, of course, this must be true. But our animals' failure to rapidly understand even the easiest seeing/not seeing tasks at the second longitudinal time point caused us to realize that we needed to more directly confront the exact manner in which their testing experiences interact with each other. Certainly skills accumulate. We have ample evidence that abilities the apes learned at one time point were retained years later: Once they learn how to do something, they remember for a very long period of time. But then why did they exhibit such poor retention on the seeing/not seeing tests? There are several possible explanations. One is that despite a full year of experiences on tests that should have helped to clarify the central construct that we were asking them about (i.e., the mental state of attention), our apes failed to integrate these new experiences with their older ones. More intriguing still, these new experiences might actually have interfered with what they had learned a year earlier. If our apes had never deeply understood why they were rewarded after gesturing to the person with the bucket on the shoulder as opposed to the person with the bucket over the head, for example, then such rules might never have been wellconsolidated—especially given that we did not overtrain them on any of the conditions; the maximum number of trials they received in any of the conditions was 28 (in the screens condition). Given this rather weak understanding to begin with, rules or relations learned during intervening tests may have displaced or interfered with these older structures. In more stark terms, if our apes had no concept of attention, then all of our tests must have seemed like a bizarre collection of arbitrary reactions to certain social stimuli with nothing more concrete than our reinforcement procedures uniting them.
Page 214
Year 3: Prying Apart Chimpanzees' Understanding of Seeing We were so struck with our apes' absence of retention at 7 years of age, that we returned to these tests when they were 8 to 9 years of age (see Reaux et al., in press). At this point, they had reached the end of adolescence—indeed, within a year, the first baby would be born to the oldest female in our group. It seemed reasonable to suppose that if chimpanzees develop a notion of seeing it ought to have emerged by this point. Our strategy for these final longitudinal tests was twofold. First, we exposed the animals to the main conditions we had administered two years previously (screens, buckets, blindfolds, eyes open/closed, distracted/attending, and back/front). We did this to determine if they would understand these conditions immediately (perhaps indicating a qualitative change in their understanding), or whether they would be forced to relearn them. Second, we intended to design some novel transfer tests that would allow us to make some final inferences about whether they had developed an appreciation of the attentional aspect of seeing. First, the results of their first four trials of each of the old conditions (screens, buckets, blindfolds, eyes open/closed, distracted/attending) are depicted in Fig. 11.7. Perhaps the most striking aspect of these results is that, in the majority of these conditions, the apes displayed little evidence that they preferred to request food from the person who could see them. However, the apes did perform at levels exceeding chance in the buckets condition (and, as usual, were perfect on the back/front trials). There was no easy way to characterize these results. The results of the buckets condition might suggest that the apes were able to use the face rule, but their poor performance on the screens condition (in which the face rule could work just as well) did not fit with this idea. Next, we administered eight more trials of each of these conditions. As can be seen from the filled bars in Fig. 11.7, this experience helped—at least for most of the conditions. With this additional experience, the animals' performance improved to levels exceeding chance in 5 of the 6 conditions. The exception was the eyes open/closed condition. (Why the animals had greater difficulty on eyes open/closed is unclear. Perhaps it was simply a more subtle discrimination. On the other hand, it is possible that the apes thought the eyes might open at any moment). At any rate, on the whole these results seemed to fit the procedural rule model that had captured the apes' performances over the two previous years. The apes seemed to have constructed a hierarchical rule structure in which the front, face, and eyes of the experimenters (in descending order of importance) served as the bases for where the apes directed their begging gestures. Recall that this model envisioned that the apes were trying to satisfy certain rule structures. First, they sought to approach and gesture
Page 215
Fig. 11.7. Results from initial tests during Year 3 for screens, buckets, blindfolds, eyes open/closed, distracted/attending, and back/front. The dotted line indicates level of performance expected by chance responding. *p < .05, **p < .01.
to someone facing forward. Clearly, the back/front condition easily allowed them to satisfy this rule. On the other hand, if this were not possible (e.g., if both persons were facing forward), then the apes moved on to the next most important rule which was to gesture to the person whose face was visible. In the case of buckets and screens, this rule was easily satisfied. Of least importance was the eyes rule, and indeed, at this point there was only limited evidence that they were even using it. They did perform well on the blindfolds condition (which could be solved by the eyes, but not the face, rule). But with the exception of a female ape named Megan, the animals did not appear to understand the eyes open/closed condition yet. Furthermore, the subjects had relearned the distracted/attending condition, which could not be satisfied with any of these rules (because in this condition both of the experimenters' eyes and faces were visible). Collectively, these results suggested that the apes were in the process of learning a set of hierarchical rules, but because these rules did not work all of the time (e.g., in the distracted/attending condition), they were also
Page 216
learning conditionspecific rules, which strictly relied on a discrimination between the two postures within those conditions. After reflecting on such conditionspecific learning, we attempted to determine how the subjects were reasoning about those conditions on which they were already performing well. Our first approach was to mix together the correct (seeing) and incorrect (not seeing) option from each of several of the different conditions (see Fig. 11.8a). We reasoned that if the apes had learned a set of conditionspecific relational rules, then their performance would be expected to decline if a correct option from one condition was mixed with an incorrect option from another. On the other hand, if the apes had extracted a contextindependent understanding that certain configurations were correct (e.g., "gesture to the person holding a screen over the shoulder"), they could be expected to perform well on these mixed conditions. Finally, we considered the conditions that the majority of the subjects did not yet understand, such as eyes open/closed. We reasoned that by mixing the incorrect option from the eyes condition (eyes closed) with the correct option from one of the conditions on which they were performing well (blindfolds), we might uncover whether the apes understood these correct options as abstractly correct, or whether they were understood as correct only within the context of other options known to be incorrect. The results of these tests are depicted in Fig. 11.8b. The most striking aspect is that the mixture of correct and incorrect options from conditions on which they were performing well posed no problem for the animals—in these cases they performed at levels exceeding chance (see results for +screens/buckets and +screens/ distracted in Fig. 11.8b). In direct contrast, when we mixed the correct option from a condition on which they were performing well, with the incorrect option from the one on which they were not, their performance was random (see results of +blindfolds/eyes and +buckets/eyes in Fig. 11.8b). Together, these results provided further evidence that the apes were learning a rule about gesturing to a person whose face was visible, and that this rule could be flexibly deployed when we mixed the different options from such conditions together. However, their inability to move beyond this rule on the mixed conditions that involved the incorrect, eyesclosed option, suggested that the apes were treating the conditions as problems to be solved by comparing physical postures, not by reasoning about who could see them. Megan, however, did not fit this pattern. She performed reasonably well on both the final series of eyes open/closed trials (6/8 correct) as well as the mixed conditions involving eyes closed (6/8 correct), a performance significantly above chance (12/16, binomial test, p = .038). Megan's performance intrigued us because she, unlike the others, had clearly learned the eyes open/closed condition during the first time point in this longitudinal project (see Povinelli & Eddy, 1996c, Experiment 13,
Page 217
Fig. 11.8. (a) Mixed conditions used in Year 3. (b) Results of mixed conditions in Year 3. The dotted line indicates level of performance expected by chance responding.
Page 218
p. 100)—although she did not learn it at the second time point, and showed a learning curve during the final time point. Such a subtle discrimination was impressive, and we knew that some researchers might interpret this as evidence that she, at least, had learned something about seeing per se. Her performance thus provided a focal point for our final test between the low and highlevel models. Recall that the lowlevel model stipulated that the apes were learning a set of procedural rules in which the front, face, and eyes served as the bases for their choices in descending order of importance. If this were true, then even for Megan the frontal aspect of a person would be more important than whether his or her eyes were open. To this end, we constructed the new condition shown in Fig. 11.9a, in which the correct option from lookingovertheshoulder (a face rule condition) was combined with the incorrect option from eyes open/closed. This presented the animals with a choice between someone facing forward (but who could not see them because the eyes were closed) versus someone facing away from them, but looking over the shoulder with eyes wide open looking in their direction. We also tested the apes on three other conditions: back/front, eyes open/closed, and lookingoverthe shoulder (a condition the apes had not experienced in 2 1/2 years). The lowlevel model predicted that the subjects would succeed on the lookingovertheshoulder condition because they could use the face rule. However, the lowlevel model also predicted that on the new mixed condition the subjects would prefer the incorrect option! The reason for this counterintuitive prediction is because of the postulated dominance of the front rule: In this condition the subjects would apply the front rule, and because it could be satisfied, would proceed to gesture to the person facing forward, even though the eyes were closed, and not to the person looking over the shoulder with clearly visible eyes and face. The results of this final test are depicted in Fig. 11.9b, and provide striking confirmation for the lowlevel model. First, as expected, the subjects tended to prefer the correct person on the lookingovertheshoulder trials. Second, and most important, the subjects performed significantly below chance in the mixed condition, meaning that they preferred to gesture to the person who was facing forward but could not see them—exactly as the lowlevel model had predicted. Although this was striking enough, there are additional aspects of the data set that are equally revealing. Megan, as well as Brandy and Kara, were almost perfect on the lookingoverthe shoulder and eyes open/closed conditions (8/8, 8/8, and 7/8, respectively, in the two conditions combined). Thus, these three animals exhibited a strong understanding of the conditions from which the novel, mixed condition had been composed. Yet when confronted with these conditions mixed together, Megan, Brandy, and Kara exhibited a strong preference for the incorrect option, selecting the person with eyes closed on 4/4, 4/4, and
Page 219
Fig. 11.9. (a) Final condition used in Year 3 consisting of eyes closed () and lookingovertheshoulder (+) options. (b) Results from eyes closed () and lookingovertheshoulder (+) condition. The dotted line indicates level of performance expected by chance responding.
Page 220
3/4 trials, respectively. However one chooses to interpret these results, they certainly do not support the idea that Megan, or any of the other animals, selected the eyes open option because they understood that this person could ''see" them. The Reinterpretation Hypothesis Chimpanzees may or may not possess a mentalistic appreciation of certain aspects of behavior. To be sure, their spontaneous behavioral patterns mirror our own so closely that we are tempted to assume that their general understanding of others is similar to our own. But the results that I have reviewed here counsel caution in following Hume, Darwin, and Romanes into assuming that behavioral similarity guarantees psychological similarity. In the remainder of this chapter, I outline an alternative view of the relation between the evolution of spontaneous social behavior and the evolution of social understanding that we have labeled the "reinterpretation hypothesis." (In the interests of space, the indictment of the argument by analogy is kept broad, but a detailed, formal critique can be found in Povinelli & Giambrone, in press.) In order to appreciate the alternative that I am about to outline, it is necessary, curiously, to conceptually decouple the evolution of behavioral complexity from the evolution of cognitive abilities related to theory of mind. Recent thinking in this area has been in the opposite direction, with many theorists supposing that theory of mind evolved in order to cope with online social problems—the socalled social intelligence hypothesis (for various treatments of the social intelligence hypothesis, see BaronCohen, 1995; Byrne, 1995; de Waal, 1982; Gallup, 1982; Humphrey, 1976; Jolly, 1966; Whiten, 1996; Whiten & Byrne, 1988; see also contributions to Byrne & Whiten, 1988). But I explore an alternative possibility, namely, that although intense sociality (characteristic of many species of primates) drove the evolution of increasingly complicated social behaviors, these behaviors were governed by psychological systems essentially no different from the earliest mammals. To be sure, various lineages of primates may have evolved cognitive specializations in various domains to cope with the unique challenges posed by their way of life. But I suppose that social complexity—mediated through and manifested by gazefollowing, deception, selective retaliation, reconciliation, and appeasement—all evolved long before abilities related to theory of mind. Thus, I wish to draw a distinction between the emergence of behavioral complexity generated through fairly lowlevel psychological mechanisms, and an understanding or interpretation of such behaviors in terms of mental states such as attention, desire, knowledge, and belief.
Page 221
Next, let us consider the evolution of theory of mind. It is possible that only one species, our own, evolved a cognitive specialization in theory of mind. Perhaps it evolved as a separate, domaingeneral faculty; on the other hand, it may have evolved as a more specific system to aide in already complicated social computations. Finally, and perhaps most likely, it may have evolved in connection with the evolution of a specialized language capacity. But, no matter how it arose, this new psychology may not have replaced the psychological structures that predated it. A fundamentally new specialization in theory of mind may have been created without destroying the rich array of social behaviors that has emerged during the course of primate evolution. Psychological states (ultimately reducible to brain states) may, in a complicated and interdependent manner, carry the ancient alongside the new. Just as developmental constraints that were established in the Cambrian are reflected in the bodily structures of modern lineages, so, too, might we suppose that innovations in psychological structures accommodated the old. If true, this new psychological system may have allowed (indeed, forced) humans to reinterpret ancient social behaviors (behaviors that evolved long before we did) in explicitly mentalistic terms. Gazefollowing is an excellent case in point. This is a phenomenon shared by humans, chimpanzees, and many (perhaps most) species of monkeys. Humans are certainly capable of interpreting the shift in gaze of others in an explicitly mentalistic manner. That is, a shift in gaze can be equated with a shift in attentional or referential focus. Indeed, some researchers have interpreted gazefollowing by 9 to 12monthold human infants as reflecting precisely this kind of understanding (see BaronCohen, 1994). Others have demanded more rigorous demonstrations, but even using these methods they have argued that by 18 months or so there is evidence for such understanding (see review by Baldwin & Moses, 1994). Although I have reservations about both positions, surely there is some period during the preschool years in which a mentalistic understanding of gaze is consolidated. That is, at some point, gaze is understood as a behavioral ambassador for the mental state of attention. This raises two problems for the argument by analogy, one developmental, the other evolutionary. From the developmental perspective, we may have already identified a case where similarity in behavior does not guarantee similarity in attending psychological states: gazefollowing in 9to 12monthold infants. Although it is true that even in adults there may also be many cases in which the behavior of gazefollowing is triggered before the representation of the other person's attentional state, there may also be other cases in which the adult's gazefollowing behavior is directly caused by the representation of the other's internal state (e.g., "What is she looking at?"). In contrast, it is at least possible that the 9 to
Page 222
12monthold's gazefollowing behavior may never be triggered by a representation of attention—perhaps because they cannot yet represent the behavior in mentalistic terms. Indeed, from the developmental perspective, this argument may really just be a general statement of KarmiloffSmith's (1992) notion of "representational redescription." She argues that the dissociations often found in human development between children's ability to produce and their inability to understand it, are evidence for a process whereby procedural information is progressively redescribed in increasingly explicit ways. The second case is the evolutionary problem. Instead of comparing an infant human to an adult human, let us compare an adult chimpanzee to an adult human. Although I have only presented a small portion of our work in this chapter, collectively it suggests that despite their sophisticated gazefollowing abilities, chimpanzees have little or no understanding that gaze is linked to internal mental states. Thus, although a chimpanzee who follows your gaze tugs on a Humean inclination to assume that he or she is trying to figure out what you are "looking at," there is considerable reason to suppose that he or she does not understand the concept of "looking at" in the first place. Furthermore, the longitudinal project I have described in this chapter reveals that although we can get our chimpanzees (like 2 to 3yearold children) to selectively gesture toward someone who can see them as opposed to someone who cannot, this similarity between their behavior belies profound underlying differences in the attending psychological interpretation of the act. This hypothesis has at least one clear advantage over the social intelligence hypothesis: It neatly reconciles the evolution of social complexity with the data summarized in this chapter. The superficially confusing mosaic of similarity and difference no longer need be seen in conflict. In the view that I am offering here, ancient social behaviors are shared in common by humans and other primates such as chimpanzees. Thus, we expect and see an astounding degree of spontaneous behavioral similarity. But humans do more. Humans interpret this behavior. It may seem as if I am advocating a form dualism in which the representation of mental states accompany, but play no causative role in, behavior. On the contrary, I assume that humans evolved the capacity to represent the mental states of self and other because of its useful, causal connection to behavior. But as the case of gazefollowing should make clear, two species may share a very similar behavior, controlled by similar lowlevel mechanisms, but interpret that behavior in very different ways. Does this mean that the human trait of representing the attentional aspect of gaze is superfluous? No, because although it may not be directly linked to the act of following gaze, the representation of attention (and other mental states) may play a profound role at a higher level of behavioral organization. Once humans evolved the capacity to represent other organ
Page 223
isms not just as behaving beings, but as beings who attend, want, and think, it became necessary to use already existing behaviors to respond to the world of the mind, as well as the world of behavior. This distinction may be especially important in those psychological arenas in which we seem to differ most from other species such as culture, pedagogy, and ethics. Determining the correct causal connections between individual behavioral elements and the mental states that ultimately control their expression may simply be a project beyond the reach of introspection alone (for a more thorough description of the causal role I ascribe to higher order intentional states, see Povinelli & Giambrone, in press). Thus, the error in the argument by analogy appears twofold. First, the argument mistakenly assumes that an accurate introspective assessment of second (or higher) order intentional states that accompany our own behaviors is likely to yield an accurate inference as to their cause. Second, the error is compounded by the assumption that because mental states accompany our behavior, similar mental states must accompany similar behavior in other species. Although humans form internal representations of mental states such as desires, knowledge, and beliefs, and these representations are now connected in causal ways to our behaviors (albeit in ways too complicated for introspection to reveal), our basic behavioral repertoire evolved long before those higher level representations were possible. Acknowledgments This research was supported by NSF Young Investigator Award SBR8458111 to D. J. Povinelli. This research would not have been possible without the talents of numerous staff and students who have assisted in the training and testing of the chimpanzees over the past five years, in particular Anthony Rideaux, James Reaux, Donna Bierschwale, Laura Theall, and Timothy Eddy. Photographs are by Donna T. Bierschwale and Corey G. Porché. References Baldwin, D. A. (1991). Infants' contribution to the achievement of joint reference. Child Development, 63, 875890. Baldwin, D. A. (1993). Early referential understanding: Infants' ability to recognize referential acts for what they are. Developmental Psychology, 29, 832843. Baldwin, D. A., & Moses, L.J. (1994). Early understanding of referential intent and attentional focus: Evidence from language and emotion. In C. Lewis & P. Mitchell (Eds.), Children's early understanding of mind (pp. 133156). Hillsdale, NJ: Lawrence Erlbaum Associates. BaronCohen, S. (1994). How to build a baby that can read minds: Cognitive mechanisms in mindreading. Current Psychology of Cognition, 13, 513552.
Page 224
BaronCohen, S. (1995). Mindblindness: An essay on autism and theory of mind. Cambridge, MA: MIT Press. Butterworth, G., & Cochran, E. (1980). Towards a mechanism of joint visual attention in human infancy. International Journal of Behavioral Development, 3, 253272. Butterworth, G., & Jarrett, N. (1991). What minds have in common is space: Spatial mechanisms serving joint visual attention in infancy. British Journal of Developmental Psychology, 9, 5572. Byrne, R. W. (1995). The thinking ape. Oxford, England: Oxford University Press. Byrne, R. W., & Whiten, A. (Eds.). (1998). Machiavellian intelligence. Oxford, England: Clarendon Press. Chance, M. R. A. (1967). Attention structure as the basis of primate rank orders. Man, 2, 503518. Corkum, V., & Moore, C. (1994). Development of joint visual attention in infants. In C. Moore & P. Dunham (Eds.), Joint attention: Its origins and role in development (pp. 6183). Hillsdale, NJ: Lawrence Erlbaum Associates. Darwin, C. (1982). The descent of man. New York: Modern Library. (Original work published 1871) de Waal, F. (1982). Chimpanzee politics: Power and sex among Apes. New York: Harper and Row. de Waal, F. (1986). Deception in the natural communication of chimpanzees. In R. W. Mitchell & N. S. Thompson (Eds.), Deception: Perspectives on human and nonhuman deceit (pp. 221244). Albany, NY: SUNY Press. de Waal, F. (1989). Peacemaking among primates. Cambridge, MA: Harvard University Press. Emery, N.J., Lorincz, E. N., Perret, D. I., Oram, M. W., & Baker, C. I. (1997). Gaze following and joint attention in rhesus monkeys (Macaca mulatta). Journal of Comparative Psychology, 111, 286293. Flavell, .. H., Everett, B. A., Croft, K., & Flavell, E. R. (1981). Young children's knowledge about visual perception: Further evidence for the level 1level 2 distinction. Developmental Psychology, 17, 99103. Flavell, J. H., Shipstead, S. G., & Croft, K (1978). What young children think you see when their eyes are closed. Cognition, 8, 369387. Fodor, J. (1992). A theory of the child's theory of mind. Cognition, 44, 283296. Gallup, G. G., Jr. (1982). Selfawareness and the emergence of mind in primates. American Journal of Primatology, 2, 237248. Goodall, J. (1986). The chimpanzees of Gombe: Patterns of behavior. Cambridge, MA: Belknap, Harvard University Press. Gopnik, A., & Meltzoff, A. (1996). Words, thought, and theories. Cambridge, MA: MIT Press. Hume, D. (1978). A treatise of human nature. 2 vols. (L. A. SelbyBigge, Ed.). Oxford: Clarendon. (Original work published 17391740) Humphrey, N. K. (1976). The social function of intellect. In P. P. G. Bateson & R. A. Hinde (Eds.), Growing points in ethology (pp. 303317). Cambridge, England: Cambridge University Press. Jolly, A. (1966). Lemur social intelligence and primate intelligence. Science, 153, 501506. KarmiloffSmith, A. (1992). Beyond modularity: A developmental perspective on cognitive science. Cambridge, MA: MIT Press. Lempers, J. D., Flavell, E. R., & Flavell, J. H. (1977). The development in very young children of tacit knowledge concerning visual perception. Genetic Psychology Monographs, 95, 353. Menzel, E. W., Jr. (1974). A group of young chimpanzees in a oneacre field. In A. Schrier & F. Stollnitz (Eds.), Behavior of nonhuman primates: Modern research trends (pp. 83153). New York: Academic Press. Mitchell, R. W., & Thompson N. S. (Eds.). (1986). Deception: Perspectives on human and nonhuman deceit. Albany, NY: SUNY Press.
Page 225
Perrett, D., Harries, M., Mistlin, A., Hietanen, J., Benson, P., Bevan, R., Thomas, S., Oram, M., Ortega, J., & Briery, K. (1990). Social signals analyzed at the single cell level: Someone is looking at me, something touched me, something moved! International Journal of Comparative Psychology, 4, 2555. Povinelli, D.J. (1996). Growing up ape. Monographs of the Society for Research in Child Development, 61(2, Serial No. 247). Povinelli, D. J., Bierschwale, D. T., & Cech, C. G. (in press). Comprehension of seeing as a referential act in young children, but not juvenile chimpanzees. British Journal of Developmental Psychology. Povinelli, D. J., Davidson, C. A., & Theall, L. A. (1997). Chimpanzees deploy attentiongetting behaviors independent of the attentional states of others. Manuscript submitted for publication. Povinelli, D.J., & Eddy, T.J. (1996a). Chimpanzees: Joint visual attention. Psychological Science, 7, 129135. Povinelli, D. J., & Eddy, T. J. (1996b). Factors influencing young chimpanzees' recognition of attention. Journal of Comparative Psychology, 110, 336345. Povinelli, D.J., & Eddy, T.J. (1996c). What young chimpanzees know about seeing. Monographs of the Society for Research in Child Development, 61(2, Serial No. 247). Povinelli, D. J., & Eddy, T. J. (1997). Specificity of gazefollowing in young chimpanzees. British Journal of Developmental Psychology, 15, 213222. Povinelli, D. J., & Giambrone, S. (in press). Inferring other minds: Flaws in the argument by analogy. Philosophical Topics. Povinelli, D.J., & O'Neill, D. K. (in press). Do chimpanzees use gestures to instruct each other during cooperative situations? In S. BaronCohen, H. TagerFlusberg, & D. J. Cohen (Eds.), Understanding other minds: Perspectives from autism–2nd edition. Oxford, England: Oxford University Press. Povinelli, D.J., Perilloux, H. K., Reaux, J. E., & Bierschwale, D. T. (1998). Young and juvenile chimpanzees' reactions to intentional versus accidental and inadvertent actions. Behavioral Processes, 42, 205218. Premack, D. (1988). Minds with and without language. In L. Weiskrantz (Ed.), Thought without Language (pp. 4665). Oxford, England: Clarendon Press. Reaux, J. E., Theall, L. A., & Povinelli, D. J. (in press). A longitudinal investigation of chimpanzees' understanding of visual perception. Child Development. Romanes, G. J. (1882). Animal intelligence. London: Keagan Paul. Romanes, G. J. (1883). Mental evolution in animals. New York: Appleton. Russell, B. (1948). Human knowledge: Its scope and limits. London: Unwin Hyman. Thomas, R. K. (1988). Misdescription and misuse of anecdotes and mental state concepts. Behavioral and Brain Sciences, 11, 265266. Tomasello, M. (1995). The power of culture: Evidence from apes. Human Development, 38, 4652. Tomasello, M., Call, J., & Hare, B. (in press). Five primate species follow the visual gaze of conspecifics. Animal Behavior. Tomasello, M., Kruger, A. C., & Ratner, H. H. (1993). Cultural learning. Behavioral and Brain Sciences, 16, 495552. Whiten, A., & Byrne, R. W. (1988). Tactical deception in primates. Behavioral and Brain Sciences, 11, 233244. Whiten, A. (1996). When does smart behaviourreading become mindreading? In P. Carruthers & P. Smith (Eds.), Theories of theories of mind (pp. 277292). Cambridge, England: Cambridge University Press.
Page 227
PART III THE SOCIOCULTURAL CONTEXT OF INTENTIONALITY
Page 229
Chapter 12— Making Sense of the Social World: Mindreading, Emotion, and Relationships Judy Dunn Institute of Psychiatry, London, England In the decade since the 1986 meeting in Toronto (Astington, Harris, & Olson, 1988), interest in children's understanding of mind blossomed in an extraordinary way: It dominates developmental psychology to an extent that must surprise even the organizers of that Toronto occasion. Interest in children's mindreading brought together theorists and ideas from very different arenas—from philosophy, psychology, evolutionary theory, neurobiology, psychopathology—echoing Baldwin's (1895) insistence that these disciplinary boundaries must be done away with. This chapter is framed around two issues: what we now know, after a decade of this energetic research into mindreading, and what we need to know about children's understanding of mind, emotion, and the social world. What We Now Know The chief focus of research has been on the delineation of normative age changes in children's grasp of what the mind does. We now have a picture of what children understand of their own and others' mental states in the preschool years; of their growing ability to distinguish thoughts and things and to reflect on and play with hypothetical events; and of their propensity to talk about what they and other people want, feel, and see. Then, as they reach 4 years, their burgeoning interest in talking about thinking, knowing, and remembering is documented (Bartsch & Wellman, 1995). We have
Page 230
learned much about the nature of children's understanding of mental representation, and their grasp of the links between people's action and their representation of the world—which may be mistaken. This research is summarized by Astington (1993) in her wonderfully lucid book The Child's Discovery of the Mind, and I do not discuss it here. Rather, my focus is on what we now know of how this understanding becomes apparent in children's social lives, and what social processes may be implicated in the development of this understanding, issues that have received much less attention. I examine the relation of these extraordinary developments in cognition to the social and emotional context of children's lives, and the pragmatics of their communication with others. If we are concerned about any of the following—the nature of children's understanding of mind and its links to human action, development and change in that understanding, the processes or mechanisms by which it changes, or the use of that understanding in children's lives—then we have to take seriously the relations of emotion and social context to the understanding that we are investigating. Otherwise we run the risk of misrepresenting the first two, and failing to make progress in understanding the second two. So what have we learned about the links between understanding and socioemotional context? Mindreading Begins in the Family First, a commonplace. The daily lives of young children are full of emotional drama. The routine events of family life can be events of major affective significance to children (as well as to adults). They are not situations of emotional neutrality but settings in which children are emotionally ready to attend to and learn about the other persons in their world (Dunn, 1988). Even in the earliest exchanges between parent and child in the first year, emotions play a central role. Bruner and others highlighted the key place that mothers' attribution of intention to very young babies may play in the development of communication (Bruner, 1983). When babies cry in distress, they may have no intention to communicate, yet their parents treat such evidence of distress as intentional communication. It is within this situation that children begin to understand the meaning of their crying for others: The argument is that children gradually acquire the ability to communicate intentionally with the conventions of the culture they share with their parents, through the interpretations offered by adults in such exchanges (see also Reznick, chap. 13, this volume). A central point is that such interactions are rarely emotionally neutral. A key theme in this chapter is that we should take serious account of the place and role of emotions in the earliest stages of communication, and in the growth of intentionality and understanding of intentions. In these early exchanges, parents set both sides of the conversation with explicit
Page 231
comments on what the baby wants or feels: "Oh, you are feeling hungry, I know!" or "Oh all right, we know what you want!" Thus parents frame the beginnings of intentional communication for the baby; their attributions of intentions, feelings, and desires to the baby's grimaces and wailing and flailing gestures are attempts to make the world more intelligible to the baby, and the baby more intelligible to them. Indeed, parental interpretation of the baby is centered on the baby's expression of emotion—his or her distress, frustration, anger, and a little later, happiness and pleasure. In these incidents, the baby's expression of emotion is crucial in parents' attribution of intentions, and I suggest that emotions are crucial in how the intentional stance is applied much more generally. If we are to understand babies' growing appreciation of intentions, we cannot ignore the significance of emotions in that appreciation. Early signs of mindreading (evident in the second year) are revealed in family interactions in which emotions often play a central part. Examples of teasing, which reflects some grasp of what will annoy or upset another, jokes that indicate some anticipation of what another will find funny, and comforting, which reflects a grasp of what will decrease someone else's distress (Dunn, 1988; Reddy, 1991), are all evidence for growing understanding. The situations in which children reveal these early powers of understanding all have a particular emotional valence—frustration, amusement, desire. Rarely are these situations affectively neutral. In the third and fourth years of children's lives, four kinds of interactive events reveal particularly vividly children's mindreading abilities—and emotions are, here too, especially salient. These four categories of interaction are important because they also suggest processes or mechanisms that may be implicated in the development of children's mindreading abilities. They are conversations about inner states, joint pretend play, early narratives, and early deception. Conversations About Inner States Children show a very early and rapidly growing curiosity about and interest in inner states. Over the third year, our studies showed, their questions about other people's feelings and wants, and about why they act the way they do, increase markedly in frequency (Dunn, 1988). Frequently their questions about others are focused on hurt or upset: "Are you all right?" "Why are you crying?" and "What's that frighten you, Mum?" Lively curiosity was also shown in our studies about the observer's likes and dislikes. Questions we were asked included: "Do you like monsters?" "Do you like dogs?'' "Want a little bit of biscuit?" "Do you get cross?" and "You like your baby?" Children's questions about other people during the third year included questions about storybook characters as well as familiar friends and family members. Polly, in the example that follows, is puzzling over the actions of a rabbit in the story her mother is reading to her, and offers a gloss on the rabbit's reaction to missing a bus:
Page 232
Polly (36 months) and her mother are looking at the book. Polly's mother quotes the rabbit in the story as saying, "Wait for me!": Polly: Why saying, "Wait for me"? Mother: Catching the bus. Polly: And it does annoy rabbit he don't got the bus.
Three different longitudinal studies have now shown links between engagement in conversations about inner states and later differences in the sophistication and maturity of children's understanding of mind and emotion (Dunn, Brown, & Beardsall, 1991; Dunn, Brown, Slomkowski, Tesla, & Youngblade, 1991; Hughes & Dunn, 1998). It is interesting to note that experimental studies using theory of mind scenarios show that children's success on these tasks is enhanced if children have the opportunity to engage in conversations about the task and the behavior of the protagonists (Appleton & Reddy, 1996). What precipitates conversations about inner states, in natural settings? Two themes stand out from our naturalistic studies. The first theme concerns emotion. It is clear that children's affective expressions are important in the genesis of talk about feelings during the second and third years. In our Pennsylvanian study (Dunn, Brown, Slomkowski, Tesla, & Youngblade, 1991), the mothers were more than twice as likely to talk about feelings with their children when the children were expressing distress or anger than when the children were happy or expressing neutral feelings (Dunn & Brown, 1994). Also, the children were more likely to engage in causal discussion of feelings when they were mildly angry or upset; the frequency of such causal conversations was correlated over time with later emotion understanding in test situations (Dunn & Brown, 1993). The second key theme that stood out from these analyses of conversations about inner states was the significance of the particular pragmatic context, not just the content, of the talk. For example, when mothers' causal talk was in the context of shared play or joking, children were most successful in the later standardized assessments of social understanding (Dunn & Brown, 1993); the correlations with test performance of earlier causal talk in the context of control and conflict were, in contrast, negative. It is clearly important to examine what the interlocutors are trying to do in the conversation—the "contexts of practice" to which Bruner (1983, 1990) has drawn our attention in the study of language acquisition. The significance of family talk about inner states does not lie just in an affectively neutral "meeting of minds." The recent research on children's interactions with other children highlights the potential of quite different pragmatic contexts from those of most parentchild conversations. Children's social worlds include others, as well as those expert, didactic adults, whose role, within a Vygotskian framework, has been seen as crucial for childrenasapprentices. From early
Page 233
childhood onward in most cultures, including our own, children spend more time interacting with siblings and peers than they do with adults. A close examination of these childchild interactions, which present quite different challenges and rewards from those of communicating with parents or sympathetic adults, reveals evidence for their potential as influences on social understanding. Three lines of evidence highlight this point. First, children with siblings do better on theory of mind tasks than those without siblings (Perner, Ruffman, & Leekam, 1994). Second, the number of siblings and other kin with whom children interact daily is positively correlated with theory of mind success (Lewis, Freeman, Kyriakidou, MaridakiKossotaki, & Berridge, 1996). Third, a key contributor to success on theory of mind tasks is the experience of earlier positive cooperative relations with a sibling (Dunn, Brown, Slomkowski, Tesla, & Youngblade, 1991). What communicative experiences between children might be implicated in the growth of early understanding of mind and emotion? A key finding from our research that included observations of the same children with their parents, their siblings, and their friends is that the children's conversations about inner states were far more frequent with their friends and siblings than with their mothers (Brown, DonelanMcCall, & Dunn, 1996). Whereas their references to mental states in conversation with their mothers were chiefly references to their own thoughts and beliefs, with their siblings and friends they were more likely to talk about shared thoughts and ideas, and these references to shared thoughts were correlated with their own performance on theory of mind tasks. These references happened chiefly in positive cooperative play: Their frequency was positively related to the observers' rating of the positive quality of the friendship, to the duration of the friendship, and to the frequency with which the friends played together. That is, the quality of the friendship was closely related to the extent of their discourse about inner states. And the discussion involving mental state terms served important pragmatic functions, facilitating collaboration by clarifying intentions, sustaining the joint narrative—the mental state talk was in the service of the collaborative enterprise, the goal of keeping the exciting play going. The interactional context in which negotiation of beliefs and suppositions are particularly relevant is, of course, joint pretend. Joint Pretend Four independent studies have now shown that individual differences in children's experiences of sharing a pretend world with another child are associated with their success on theory of mind and emotion understanding assessments (Astington & Jenkins, 1995; Dunn & Cutting, 1998; Hughes & Dunn, 1997; Youngblade & Dunn, 1995). Plainly, this play in which roles are taken up and a story line jointly planned should be
Page 234
greatly facilitated by the ability to take account of another child's thoughts and intentions; in turn, efforts to collaborate on sharing and shaping a world of pretend identities appear likely to foster this ability. The general point of importance is that these data show us that children begin to entertain multiple hypothetical realities and "decouple" reality from fantasy not as solitary cognitive enterprises, but through negotiating social interactions in which these cognitive states are shared (a very Baldwinian idea). An important cautionary note is that we cannot make causal inferences from such correlational data: The patterns over time in our studies could well reflect a common causal factor, such as continuing characteristics of children that are reflected both in their performance on mindreading tasks and in their propensity to engage in joint pretend play. But there is clearly a plausible case to be made for the potential of joint pretend with another child as a context in which mind reading can be fostered. It is interesting that Baldwin (1895), observing the extended pretend play of his two daughters aged 2 and 4, commented, "The prolonged situations, maintained sometimes whole days or possibly weeks, give strength to the imagination. . . . The sense of essential reality, and its distinction from the unreal . . . is helped by this sort of symbolic representation" (p. 346). Early Narratives The third kind of communicative event that reveals and possibly fosters children's early understanding of the links between mind, emotion, and human action is their engagement in narratives. Bruner and Feldman (1993) argued powerfully for the significance of narrative as a process through which the development of understanding of mind and emotion may be influenced. Patterns of narrative, they propose, scaffold the kind of metacognition about intentions that lies at the core of theories of mind. In our studies of the second, third, and fourth years, we found that development of children's unsolicited narratives about the past parallels that of their comments on psychological issues more generally. But these early narratives were not emotionally neutral examples of cognitive sophistication. My colleague Jane Brown's elegant study of children's early talk about the past highlighted two key issues concerning these early narratives (Brown, 1995; Dunn, 1988). The first issue was that the children mustered their most sophisticated linguistic skills—referring to inner states, sequencing events temporally and causally—when they reported events involving fear, anger, or distress. In the following example, a 47monthold runs back from the garden to the house to tell her sibling about a sequence of events, and in her story she links these events to her feelings and subsequent action: Child to sibling (as she runs into house from the garden): I came running back 'cause I saw two snakes and I was scared and I runned back!
Page 235
It was the emotional dramas in their lives that prompted the children to embark on telling coherent stories about the past—and especially about their negative experiences. As Brown (1995) noted, "When children talked about negative events, and in particular, when they did so in efforts to get another to act on their behalf, they mustered their most sophisticated language skills." Hudson and her colleagues (Hudson, Gebelt, Haviland, & Bentivegna, 1992) have parallel findings, showing that children were more likely to use a plotted story structure and focus on causes and consequences when telling a story involving anger or fear than when telling one involving happiness. Happy stories tended to be simpler, focusing on the emotional tone of the moment in time. Hudson and colleagues argued that these differences may well reflect differences in how we conceive of and remember emotional experience. Negative emotional experiences frequently involve blocked goals and thus a complicating action; so, they lend themselves to a causally or temporally sequenced structure. Happy experiences, in contrast, are often simply conveyed, as when one of the children in our studies exclaimed to her older brother, "I peed by myself, Ryan!" (Brown, 1995). The second issue to which Brown (1995) drew attention concerned pragmatics. She showed that talk about negative past events predominated when children were attempting to influence another's behavior. In contrast, talk about happy or neutral events occurred most often when children were not trying to get someone to do something. This evidence for the instrumental context of narratives is important. Families and friends engage in frequent negotiations about whose rights were violated in a transgression, or who was responsible when an accident happened. Their discussion of the past often concerns the recent past, which was experienced jointly but from different and often competing perspectives. A child's ability to construct her own compelling and believable account of what happened may be called on to a greater degree in the context of a dispute than in other settings. This "communication pressure," to use Mannle and Tomasello's (1987) phrase, and the urgency to have his or her own needs met are probably contributors to the development of a child's communicative competence. Our data also showed that children have multiple opportunities to learn how best to tell what happened from listening to the exchanges between their older siblings and parents; older siblings frequently win arguments because of their more sophisticated narrative accounts (Brown, 1995). The younger children witnessing these arguments are given clear opportunities to learn from more competent others—whose interests were often at odds with the younger children's own. Again we come back to the limitations of a Vygotskian tradition that focuses only on relatively didactic interactions. It is important not to lose sight of the highly charged emotional contexts in which children's cognitive
Page 236
skills develop—and the relationship goals that should surely be included in our investigations. As Goodnow (1990) pointed out, the picture of children's worlds that is painted in much research within a Vygotskian perspective is of a "benign and relatively neutral world" that ignores the affective aspects of cognitive development. Early Deception The fourth kind of communicative event of evident relevance to children's growing understanding of mind is deception. Incidents of apparent attempts to deceive parents, to shift blame for misdeeds onto a brother or sister, or to give false excuses for wrongdoing are so striking, so common, and seen so early in children's development, if we observe children in their family world (Dunn, 1988). The question of whether young children of 3, 4, or 5 years of age intentionally deceive others is right at the center of accounts of the development of understanding other minds. And it is a very contentious issue: Experimental studies vary very much in the accounts of the development of deception that they generate (Chandler, Fritz, & Hala, 1989; Sodian, 1991; Strichartz & Burton, 1990). However, Newton (1994) carried out a series of studies of naturally occurring incidents in which children apparently attempted to deceive other family members. These studies showed first that 3yearolds engage in types of deception similar to those of 4yearolds, although not as frequently. That is, the 3yearolds were already attempting to manipulate what other family members thought or expected. Second, even children who failed the theory of mind tests were observed to attempt to deceive others. And third, children's attempts to deceive were most frequently made "in situations of conflict when the child is in an emotionally charged state of opposition to parental control" (Newton, 1994). Interesting to note, children's motivation to deceive appeared, according to Newton, to be far more subtle than simply the avoidance of punishment. Examples of false boasting, and bravado in the face of painful punishment demonstrated that even 3 and 4yearolds engaged in impression management. These findings are a challenge to the accounts based on experimental settings, and reinforce the argument that if at the end of the day we want to explain reallife behavior, we simply cannot ignore the reality of children's social behavior in situations that matter to them, or their interactive goals in their close relationships. What We Need to Know We have learned much over the last decade about the key significance of the emotional context, and of what children and those interacting with them are trying to do when they communicate, for children's growing
Page 237
social understanding. This last section focuses on three sets of questions that stand out as issues for future research agenda. Causes One of the most intractable of questions concerns the causes of developmental changes in understanding. The emphasis here is on the potential power of social experiences as influences on children's mindreading. Other accounts, of course, emphasize very different possibilities: the role of biological maturation and innate mechanisms, as in Leslie's (1987) argument for a module within the brain that constrains the development of understanding of mind; the power of children's introspection in constructing their understanding (e.g., Harris, 1991); and—the most widely accepted—the view that children form changing theories about why people behave as they do. How should we weigh these different accounts? Should we consider the possibility that different children reach these developmental achievements by different routes? Or that combinations of all of these processes are important? Or that for different aspects of social understanding, different processes may be implicated? The issues of differentiation between various aspects of social understanding, and of individual differences in children's mindreading abilities are now considered. First, note one implication of the general developmental account that research on children's talk about the mind has suggested (e.g., Bartsch & Wellman, 1995). According to this account, very young children explain people's actions initially in terms of feelings and desires, and through their social experiences—particularly those that involve affective issues—they come to incorporate the notion of belief in their understanding of people's actions. One implication of this account is that an understanding of cognitive states arises from an earlier understanding of emotional (noncognitive) states. If this is so, then as Bartsch and Estes (1996) noted, it suggests that "a comprehensive account of metacognition will have to be anchored in a much broader understanding of development and will require a better understanding of the relationship between cognitive and noncognitive psychological phenomena" (p. 299). We need to be thinking, then, about developmental issues of a considerably wider span than that suggested by a conventional cognitive approach. Differentiation The second set of questions concerns differentiation. In focusing on children's discovery of the mind, are we investigating some general aspect of cognitive development—some domaingeneral understanding? Or should the various features of social understanding—understanding emotions, be
Page 238
liefs, mental representation, for instance—be differentiated? How are children's abilities to "read minds" related to their powers of memory, to language, to executive function, to their understanding of emotion or prosocial communication? Here we face a key issue—a matter of central dispute—bringing in the evidence for and against modularity of brain function. It was argued convincingly by Moore (1996) that mindreading is domain general in terms of adaptive function, but that differentiation of aspects of mindreading may well take place developmentally. The parallel is drawn with KarmiloffSmith's (1992) argument that modularity in adult information processing can arise through domaingeneral cognitive mechanisms working in particular environmental contexts. This brings us back to emotion. Does the significance of emotion and social context vary for the development of different aspects of understanding? It is clear that the emotional and social context is centrally important for the use children make of their understanding of other minds; our analyses of children's understandinginaction (their engagement in joint pretend role play, their discourse about mental states, and their management of conflict) in the context of their different relationships with mother, sibling, and friend showed that there were no significant correlations across these relationships. The use children made of their understanding of their partner's inner states was crucially related to the emotional context of the interaction (Dunn, 1996). Beyond this point about the significance of emotional context for use of understanding, there are now two exciting new—as yet tentative—findings on developmental patterns linking emotion and understanding. First, there are findings that imply that the sequelae of early sophistication in emotion understanding and in mind reading may well differ (Dunn, 1995). Second, there are important studies that indicate that differences in early attachment relationships may be related to some, but not necessarily all, aspects of understanding mental states (e.g., Meins, Fernyhough, Russell, & ClarkeCarter, 1998). More generally, a host of developmental questions arise when we focus on links between emotion and social understanding. Does emotion become less important in the way children use their powers of understanding others, as they grow up? How do relations between emotion and cognition change with development? As children's powers of metacognition develop, do they become less at the mercy of their own emotions? At present we hardly know. The preoccupation of researchers till now with children under 5 years old means that we have not begun to explore the changes in the skills and understanding of older children. There is a marvelous passage in The Mill on the Floss in which George Eliot, describing her young heroine's desperation after a humiliating incident, refers to the "perspectiveless conception of life" (p. 122) of young children, which gives their emotional
Page 239
experiences a crucial intensity. It is the changes in the nature of that conception and its relation to emotional experience that we urgently need to understand. Differences: Individual and Cultural The recent increase in interest in individual differences in children's understanding of mind is proving tremendously fruitful for examining links between social experiences and mindreading (Astington, 1993). And a whole range of issues has been raised concerning such differences. For instance, do patterns of development and links between cognitive and social development differ for children within the normal range, and those at the extremes? There are for instance hints that among children designated as having attention or behavioral problems, patterns of mindreading and executive function may differ from those of children within the normal range. These differences may contribute to the problems these children have in their close relationships. Thus, Hughes' (Hughes, Dunn, & White, 1998) research on children classed as hardtomanage by their preschool teachers suggested both interesting differences in the nature of the children's understanding of mind, and intriguing connections between their antisocial behavior and their cognitive abilities, as reflected in theory of mind tasks, executive planning, and language. Thus, when given false belief tasks that involved emotions (a nice surprise and a nasty surprise) significantly more of the hardtomanage children passed the test when a nasty surprise was involved than when a nice surprise was the focus of the story. For the control children, no such task effect was found. Second, those whose antisocial acts involved teasing were marginally better at the deception tasks than the other children, whereas those whose behavior involved aggression and snatching were particularly poor at the executive function tasks that involved planning. Finally, a note on cultural differences. The focus on family and friend interactions and discourse should not lead us to ignore the significance of cultural messages about why people behave the way they do, for children's growing understanding. We now know that such cultural messages reach children very early indeed, through these very conversations about people's actions and feelings, through prescriptive messages, and through shared pretend play. In our own comparisons of the prescriptions reaching 3yearolds in Centre County Pennsylvania and in Cambridgeshire, England, we found differences in a host of aspects of these messages (Dunn & Brown, 1991). First, the topic of such prescriptions varied (in the United Kingdom, the focus was often on politeness and acceptable mealtime behavior, whereas in the United States, it was frequently on appearance). Second, the form of the prescription differed (in the United States, pro
Page 240
hibitions were chiefly in the form of direct prohibitions of individual action; in the United Kingdom, there were also messages about normative behavior, and indirect suggestions for what was generally acceptable beyond the family). Third, evaluative moral messages more often were attached to prescriptions in the United Kingdom, whereas references to feelings were more commonly made in the U.S. families. Cultural differences were evident in the children's fantasies, and in their narratives, too. The British children's narratives were unlike those of American children in that they more often involved other people—raising the possibility that these English children were growing up with the expectation that their lives and exploits would be the subject of public scrutiny. In summary, even at 3 or 4 years old, the impact of the wider world beyond the family was clear—but these differences were mediated through the family discourse and relationships. Conclusion The discoveries of the last 10 years about the nature of children's understanding of the mind have been extremely exciting. However, it is now time, surely, to pursue the developmental questions that so preoccupied Baldwin, and to recognize that we cannot make progress in answering these questions without paying attention to children's use of their understanding in the real world, and to the extraordinary forces on children to become effective participants in their social worlds. A focus on the daily dramas of children's lives highlights first, the sophistication of children's understanding of familiar others, and second, the significance of the emotional quality of these interactions and the intentions of the interlocutors. To describe the development of children's discovery of the mind, we need to pay attention to the connections between children's social and intellectual lives that are revealed in those daily dramas, a point succinctly made by Baldwin (1895) himself: "The best school in the humanities for every man is in his own house" (p. 348). References Appleton, M., & Reddy, V. (1996). Teaching 3yearolds to pass falsebelief tests: A conversational approach. Social Development, 5, 275291. Astington, J. W. (1993). The child's discovery of the mind Cambridge, MA: Harvard University Press. Astington, J. W. , Harris, P. L., & Olson, D. R. (Eds.). (1988). Developing theories of mind Cambridge, England: Cambridge University Press.
Page 241
Astington, J. W., & Jenkins, J. M. (1995). Theory of mind development and social understanding. Cognition and Emotion, 9, 151166. Baldwin, J. M. (1895). Mental development in the child and the race. New York: Macmillan. Bartsch, K, & Estes, D. (1996). Individual differences in children's developing theory of mind and implications for metacognition. Learning and Individual differences, 8, 281304. Bartsch, K., & Wellman, H. M. (1995). Children talk about the mind. Oxford, England: Oxford University Press. Brown, J. R. (1995). What happened?: Emotional experience and children's talk about the past. Unpublished manuscript. Brown, J. R., DonelanMcCall, N., & Dunn, J. (1996). Why talk about mental states?: The significance of children's conversations with friends, mothers and siblings. Child Development, 67(3), 836849. Bruner, J. S. (1983). Child's talk Oxford, England: Oxford University Press. Bruner, J. S. (1990). Acts of meaning. Cambridge, MA: Harvard University Press. Bruner, J., & Feldman, C. (1993). Theories of mind and the problems of autism. In S. BaronCohen, H. TagerFlusberg, & D. Cohen (Eds.), Understanding other minds: Perspectives from autism (pp. 267291). Oxford, England: Oxford University Press. Chandler, M., Fritz, A. S., & Hala, S. (1989). Smallscale deceit: Deception as a marker of two, three and fouryearolds' theories of mind. Child Development, 60, 12631277. Dunn, J. (1988). The beginnings of social understanding Cambridge, MA: Harvard University Press. Dunn, J. (1995). Children as psychologists: The later correlates of individual differences in understanding of emotions and other minds. Cognition and Emotion, 9, 187201. Dunn, J. (1996). Children's relationships: Bridging the divide between cognitive and social development. Journal of Child Psychology and Psychiatry, 37, 507 518. Dunn, J., & Brown, J. (1991). Becoming American or English: Talking about the social world in England and the U.S. In M. Bornstein (Ed.), Crosscultural approaches to parenting (pp. 155172). Hillsdale, NJ: Lawrence Erlbaum Associates. Dunn, J., & Brown, J. (1993). Early conversations about causality: Content, pragmatics, and developmental change. British Journal of Developmental Psychology, 11, 107123. Dunn, J., & Brown, J. (1994). Affect expression in the family, children's understanding of emotions, and their interactions with others. MerrillPalmer Quarterly, 40, 120137. Dunn, J., Brown, J., & Beardsall, L. (1991). Family talk about feeling states and children's later understanding of others' emotions. Developmental Psychology, 27, 448455. Dunn, J., Brown, J., Slomkowski, C., Tesla, C., & Youngblade, L. (1991). Young children's understanding of other people's feelings and beliefs: Individual differences and their antecedents. Child Development, 62, 13521366. Dunn, J., & Cutting, A. (in press). Understanding others, and friendship interactions in young children. Social Development. Eliot, G. (1979). The mill on the floss. Harmondsworth, England: Penguin Books. Goodnow, J. J. (1990). The socialization of cognition: What's involved? In J. W. Stigler, R. A. Shweder, & G. Herdt (Eds.), Cultural psychology (pp. 259286). Cambridge, England: Cambridge University Press. Harris, P. L. (1991). The work of the imagination. In A. Whiten (Ed.), Natural theories of mind (pp. 283304). Oxford, England: Blackwell Publications. Hudson, J. A., Gebelt, J., Haviland, J., & Bentivegna, C. (1992). Emotion and narrative structure in young children's personal accounts. Journal of Narrative and Life History, 2, 129150. Hughes, C., & Dunn, J. (1997). ''Pretend you didn't know": Preschoolers' talk about mental states in pretend play. Cognitive Development, 12, 381403. Hughes, C., & Dunn, J. (1998). Understanding mind and emotion: Longitudinal associations with mentalstate talk between young friends. Developmental Psychology, 34(5), 10261037.
Page 242
Hughes, C., Dunn, J., & White, A. (1998). Trick or treat?: Uneven understanding of mind and emotion and executive function among "hard to manage" preschoolers. Journal of Child Psychology and Psychiatry, 39(7), 981994. KarmiloffSmith, A. (1992). Beyond modularity: A developmental perspective on cognitive science. Cambridge, MA: MIT Press. Leslie, A. M. (1987). Pretense and representation: the origins of 'theory of mind'. Psychological Review, 94, 412426. Lewis, C., Freeman, N. H., Kyriakidou, C., MaridakiKassotaki, K., & Berridge, D. M. (1996). Social influences on false belief access: Specific sibling influences or general apprenticeship? Child Development, 67, 29302947. Mannle, S., & Tomasello, M. (1987). Fathers, siblings and the bridge hypothesis. In K. E. Nelson & A. van Kleeck (Eds.), Children's language (Vol. 6, pp. 23 41). Hillsdale, NI: Lawrence Erlbaum Associates. Meins, E., Fernyhough, C., Russell, J., & ClarkeCarter, D. (1998). Security of attachment as a predictor of symbolic and mentalising abilities: A longitudinal study. Social Development, 7, 124. Moore, C. (1996). Evolution and the modularity of mindreading. Cognitive Development, 11, 605621. Newton, P. E. (1994). Preschool prevarication: An investigation of the cognitive prerequisites for deception. Unpublished doctoral dissertation, University of Portsmouth, Portsmouth, England. Perner, J., Ruffman, T., & Leekam, S. R. (1994). Theory of mind is contagious: You catch it from your sibs. Child Development, 65, 12281238. Reddy, V. (1991). Playing with others' expectations: Teasing and mucking about in the first year. In A. Whiten (Ed.), Natural theories of mind (pp. 143158). Oxford, England: Blackwell. Sodian, B. (1991). The development of deception in children. British Journal of Developmental Psychology, 9, 173188. Strichartz, A. F., & Burton, R. V. (1990). Lies and the truth: A study of the development of the concept. Child Development, 61, 211220. Youngblade, L. M., & Dunn, J. (1995). Individual differences in young children's pretend play with mother and sibling: Links to relationships and understanding of other people's feelings. Child Development, 66, 14721429.
Page 243
Chapter 13— Influences on Maternal Attribution of Infant Intentionality J. Steven Reznick University of North Carolina at Chapel Hill I am intrigued by the sorts of infant behaviors that we label as intentional. A folk psychological definition of intentionality claims that intentional behaviors are done deliberately, on purpose, or with awareness. Confusion emerges because philosophers often use the term intentionality in a more general sense. In the scholastic tradition, brought to psychology in the 19th century by Franz Brentano (1874/1973) and to modern ears by John Searle (1983) and Daniel Dennett (1987), intentionality is a term used to describe mental states that are directed at, or are about, objects and states of affairs in the world. A belief is an example of intentionality because it is a hypothesis about something (e.g., a belief that it will snow); a fear is an example of intentionality to the extent that the fear is of something or that something will occur (e.g., a fear of snakes). So, the behaviors that we normally think of as intentional in the folk psychological sense of being deliberate (e.g., pulling a string to retrieve a toy) are also intentional in the philosophical sense in that they imply an action undertaken with a goal in mind. More important, the philosophical definition of intentionality includes a much larger set of mental states that are relevant. That is, an infant who has hopes, fears, or thoughts could be considered intentional. Research on Englishspeaking adults conducted by Malle and Knobe (1997) indicates that speakers generally use the term intention to describe the mental state of a person who "intends to do something" (i.e., has a specific set of beliefs and desires about a goal state), and that intentional is more often used to describe an action that is performed by an agent who has
Page 244
not only intention, but also the skill needed to perform an appropriate goaloriented action. The specification of skill is important: I might have an intention to throw a dart into the bull's eye (i.e., a belief that it is a good thing to do and a desire to make it happen), but if I do not have the requisite skill, my accurate toss is not likely to be considered intentional. There is considerable interest among philosophers in the broad questions of what it means to say that a behavior is intentional (e.g., Anscombe, 1957) and of whether humans are the only organisms that can behave intentionally (e.g., Griffin, 1992; Radner & Radner, 1989), but little interest in the developmental question of when and how intentionality emerges in human infants. Two paths diverge at this point in my presentation, and it is important to recognize their differences. The realist path leads to the question of whether or not infants can behave intentionally. There is a rich history of debate in developmental psychology concerning the infant behaviors that suggest intentionality in the folk psychological sense. Candidates include coordinated reaching and grasping (Bruner, 1973; Lockman, Ashmead, & Bushnell, 1984; Thelen et al., 1993), meansends behaviors (Mosier & Rogoff, 1994; Willatts, 1984), and communication via gestures or words (Bates, Benigni, Bretherton, Camaioni, & Volterra, 1979; Bruner, 1975; Dore, Franklin, Miller, & Ramer, 1976; Harding & Golinkoff, 1979). Intentionality can also be inferred on the basis of interactive behaviors with a social partner (Bakeman & Adamson, 1984; Trevarthen, 1979); acts of shared reference or joint attention in which infants follow an adult's direction of gaze or pointing (Butterworth & Cochran, 1980; Moore & Dunham, 1995; Murphy & Messer, 1977; Scaife & Bruner, 1975); or in acts of social referencing in which infants alter their goaldirected activity in response to emotional signals from an adult (Campos & Stenberg, 1981; Emde, 1984; Feinman, 1982; Walden & Ogan, 1988). These interactive behaviors could even suggest some rudimentary skill at ascribing intentional states to others, the usual topic in theory of mind research conducted with preschool children. Given this range of behaviors that could be labeled intentional, there is little agreement regarding the exact birth date of intentionality. However, across various theoretical approaches, there is a broad consensus that intentional action blossoms during the second 6 months of life, along with transitions in cognition, manipulative ability, mobility, and the capacity to experience and share emotions. I confess to being a realist about intentionality; I am comfortable with the claim that some systems have internal properties that allow them to be intentional. In the present context, that translates into a belief that research on human infants will lead to persuasive claims about a psychological state that could be called "having intentionality" and that becomes an increasingly salient aspect of the infant late
Page 245
in the first year. On the other hand, as Dennett (1987) has noted, there are many nonhuman systems that appear to have intentionality: It can be said that sunflowers turn in order to face the sun, thermostats seek a particular temperature, or cars hate to start on cold mornings, with each statement implying some degree of intentionality. Dennett (1987) suggested that we describe these systems as intentional because this designation helps us explain, understand, and predict them. This notion of an intentional stance is relevant in the present context because even if infants cannot lay a substantive claim to intentionality on the basis of internal properties, parents and other people certainly regard infants as intentional: That is, they take an intentional stance toward children. Thus, the second path, and the one on which we travel here, is the ascription/attribution path, in which we focus on the willingness of adults to ascribe or attribute intentionality to infants. From a folk psychological perspective, this probably means that adults assume that infants have intentions (i.e., beliefs and desires) and the skills necessary to perform actions pursuant to these beliefs and desires. In this chapter, I use the phrase "parent perception of infant intentionality" (PPII) to refer to the parent's belief that an infant can do things on purpose or deliberately. It is important to note that my use of the term perception does not necessarily entail a commitment to a realist stance in which parents perceive the intentionality that is inherent in the infant's behavior. Most adults are willing to ascribe or attribute some degree of intentionality to various infant behaviors. The result of this ascription/attribution is some degree of perception of infant intentionality. My interest in PPII is not just idle curiosity: The willingness of parents to ascribe intentionality could have significant effects on infant development. Psychologists often contrast the relative influence of nature and nurture, but it would be ludicrous to deny the importance of the parent's role in the infant's psychological development. Kaye (1982) describes the parental role as a series of psychological frames. Of most relevance here is Kaye's "instrumental frame" in which the adult carries out what appears to be the infant's intentions. Kaye suggests that this context of intentional interpretation facilitates the development of the infant's skills (e.g., meansends activities), a view also shared by Bruner (1983). Tomasello (1992, 1996) extends this reasoning to suggest that when infants are treated as intentional, they come to understand that others are intentional agents, which in turn facilitates their acquisition of language (see also Dunn, chap. 12, this volume). I present some of our recent findings, which are organized into three themes that illustrate ongoing progress in understanding PPII and that motivate our research.
Page 246
Measuring Parent Perception of Infant Intentionality I first describe some of the ways in which we measure PPII. In our research to date, we have used three main techniques that can be considered direct in the sense that an explicit narrative context is provided that evokes the parent's response. These techniques can be compared with other approaches, not described here, in which PPII is inferred on the basis of the parent's behavior in a less obvious context. Infant Intentionality Tape The Infant Intentionality Tape (IIT) is a videobased instrument that could be considered either a questionnaire or a projective test. Suzanne Zeedyk (1994), in her dissertation, filmed infants in normal daytoday situations and then selected brief film clips in which infants were behaving more or less intentionally in the sense of performing the sorts of behaviors that parents tend to consider intentional (e.g., persisting in efforts to obtain a goal, participating in interactive games). The most recent version of the tape (Zeedyk & Reznick, 1996) has 28 segments of 5 to 12monthold infants of various ethnicity. Some film clips have infants alone and others show a parent as well; some infants are playing with toys and others are unencumbered. Participants view the film clips and rate each of them on an 8point scale that indicates how confident they are that the infant's behavior in the film clip is deliberate or on purpose. Clips in which the infant is not behaving intentionally are rated between 1 and 4, and clips in which the infant is behaving intentionally are rated between 5 and 8, with gradations to reflect the parent's confidence that the behavior was or was not deliberate or on purpose. We assume that the average rating across the IIT clips reflects the participant's general tendency to infer infant intentionality. Infant Intentionality Questionnaire Our second measure of PPII is even more direct. The Infant Intentionality Questionnaire (IIQ, Version 4.2; Reznick & Feldman, 1998), contains 32 specific questions about intentionality (e.g., Does your infant choose appropriate ways to reach his or her goals? Does your infant recognize when you are disappointed? Can your infant feel sadness?), with the target always "your infant." The broad themes of these questions concern the infant's awareness of needs, desires, and goals, and the infant's ability to cause the environment to respond. In the most recent version, we have added several questions that probe for aspects of intentionality that are negative. One
Page 247
such question is "Does your baby do things on purpose to make you angry?" Given the content of the questions, it seems reasonable to expect that the IIQ taps the parent's assessment of his or her own infant's intentionality (in contrast to the general view about infant intentionality tapped by the IIT). Predicting Infant Intentional Behavior Our third measure of PPII is in some sense the most direct. As part of our ongoing longitudinal study, we needed a context in which we could observe the extent to which individual 9monthold infants display the sorts of behaviors that parents regard as intentional. A considerable amount of pilot testing led to a set of games and tasks in which individual differences in intentionality might be revealed. For example, in one task the infant sees an attractive toy that is on the other side of some horizontal bars. An infant who ignores the toy would be regarded as less intentional. An infant who retrieves the toy would be regarded as more intentional. In another episode, the experimenter attempts to engage the infant in a game of peekaboo. A less intentional infant would not participate. A more intentional infant would not only participate but also actively initiate subsequent rounds of peekaboo. When the battery of intentionality tasks was ready to use and coding procedures had been established, we wrote a questionnaire that describes each episode and asks the parent to predict how his or her infant will respond. The multiplechoice answers available for each question correspond to our coding scheme for intentional behaviors. For example, "Your baby is wearing a long bib. The experimenter raises the bib in front of her own face, lowers it, and says, 'Peekaboo.' The experimenter then raises the bib again. Your baby will: 1. Not be interested at all. 2. Pull the bib down to encourage the game. 3. Pull the bib down and then continue the game himself/herself by pulling the bib up." In our ongoing study, mothers complete the 20item questionnaire Predicting Infant Intentional Behavior (PIIBVersion 1.2; Reznick, 1998) before their infant participates in the battery of tasks. In comparison to the other instruments, the PIIB is focused on a specific infant's behavior but the inference of intentionality is based on our interpretation of specific behaviors: The PIIB describes behavior and never mentions the term intentional. The Present Data The data that will be presented in this chapter are from a sample of 130 mothers. Previous research suggests some minor differences between perception of infant intentionality for mothers and fathers, but data from
Page 248
fathers was not available for the present analysis. Fortyseven mothers were recruited in our laboratory at Yale through letters mailed to infants with birth certificates on file in the New Haven Hall of Records. This recruitment method provides a middleclass and uppermiddleclass sample that is almost entirely EuropeanAmerican, suburban, and well educated. These mothers had infants within the 6 to 12month age range. An additional 83 mothers were recruited while they attended wellbaby pediatric visits at the Hill Health Center in New Haven. The Hill Health Center serves a disadvantaged urban neighborhood and has a clientele that identify themselves as primarily African American or Hispanic. The Hill Health sample includes 41 AfricanAmerican mothers, 34 Hispanic mothers (primarily of Puerto Rican origin), and 8 EuropeanAmerican mothers. These mothers were tested when their infants were between 9 and 10 months of age.1 The mothers recruited at Hill Health Center reflect a range of educational attainment but are almost all relatively poor: 92% of them qualify for WIC funds, 67% receive food stamps, and 62% are on Aid to Families with Dependent Children, the eligibility of which programs are based on income. All of the Hill Health sample and most of the Yale sample are participating in an ongoing longitudinal study of PPII. Additional mothers were tested by Nicole Eldridge (1997) for her undergraduate Senior Project at Yale. Influences on Parental Perception of Infant Intentionality Do Parents Attribute Intentionality Because of the Presence of Certain Infant Behaviors and Characteristics? Infant Behaviors Suzanne Zeedyk (Zeedyk, 1992, 1994) used various interviews and questionnaires to determine the criteria that parents claim would lead them to infer that an infant is acting deliberately or on purpose. Most of the criteria involve actions. For example, parents say that they consider infants to be intentional when the infant anticipates the outcome of an action, selects actions appropriate to a goal, persists in actions to attain a goal, and stops action when a goal is attained. Other criteria imply some sort of emotional or social presence. For example, parents claim that they notice when infants maintain eye contact or modulate their emotional intensity. It is important to determine whether or not parents do indeed bring these criteria to bear when making a judgment about an infant's inten 1
The terms African American, Hispanic, and EuropeanAmerican are not satisfying for various reasons but are used here because they designate a group of interest and/or because they are used by the mothers themselves in response to a question about their ethnicity.
Page 249
tionality. At one level, we are questioning the validity of the IIT. That is, are parent ratings of the video clips determined (or at least, influenced) by the infant's behavior? At a more generative level, we are somewhere on the road between intentionality as a projection and intentionality as constitutive: If we can identify the behaviors that imply intentionality, we may have some guidance in our search for the underlying mechanisms that allow infants to behave intentionally. The data suggest a simple answer but in the context of a more complex answer. The simple answer is yes, many parents do bring the same specific criteria to bear when judging an infant's intentionality. Table 13.1 lists the 28 IIT clips and summarizes parental response to each clip for the 130 mothers. The clips in Table 13.1 are ordered according to the average intentionality score assigned to the clip by the mothers. Average scores ranged between 6.79 and 2.55 on the 8point intentionality score. The distribution of scores suggests that parents tended to honor the distinction between intentional clips (ratings of 58) and notintentional clips (ratings of 14): There was a notable gap between clip 12 (M = 5.23) and clip 19 (M = 4.29). On the basis of this discontinuity, we divided the clips into those that were considered on average to be intentional and those that were considered on average to be not intentional. We used two statistical approaches to confirm this dichotomy. First, we calculated the percentage of mothers who responded with a score of 5 or higher to each clip. As indicated in Table 13.1, 70% or more of the mothers responded with 5 or higher on the clips designated as intentional and no more that 52% of mothers gave a score of 5 or higher on the notintentional clips. Second, we computed an analysis of variance (ANOVA) in which each mother's rating of each clip was the unit of analysis. This analysis indicated a main effect of clip, F (27, 3449) = 39.40, p < .01. A Student Neuman Keuls post hoc test indicated several clusters of clips, but the primary locus of the effect was a difference between the clips that we have designated as intentional and not intentional. Inspection of the scripts for the items with high intenionality ratings confirms the criteria reported by Zeedyk (1994). High intentionality scores were assigned to clips in which the infant performs some goaloriented action (e.g., manipulates a "busy box," shakes a rattle, inserts a cube in a slot, slides a shovel across a table). High intentionality scores were also assigned when the infant participates in a social game (e.g., putting a foot in the mouth as mother says no, hand games, touching father's face). From this informal categorization of clips, it would appear that a third cue for intentionality is perceptual vigilance. For example, most mothers assigned high intentionality scores to a clip in which an infant is crawling and then stops to look at something, or a clip in which the infant is simply scanning the room. This implies some parental sensitivity to the broader
Page 250 TABLE 13.1 Average Intentionality Rating, Description, Distribution of Ratings, and Script for Each Clip Rating
Sex
Skin
Age
³ 5
1,27,8
Script
6.79
24 Boy
Light
11
93%
7%75%
Manipulates various parts of "busy box"
6.60
6 Girl
Light
12
91%
9%71%
Puts foot in mouth as mother says no, smiles
6.51
22 ?
Light
11
92%
8%64%
Responds reciprocally to mother's hand games
6.43
28 Girl
Light
8
86%
9%65%
Imitates mother's spin of "busy box" wheel
6.38
5 Girl
Light
9
92%
11%64%
Shakes rattle up and down
6.31
10 ?
Dark
10
82%
10%63%
Inserts cube into correct slot, mother claps
6.23
17 Girl
Light
7
86%
10%60%
Crawls, stops to look at something
Light
11
87%
13%55%
Bangs hands on highchair tray repeatedly
Dark
10
86%
13%52%
Slides shovel on high chair, watches attentively
6.12 5.98
3 Boy 4
5.75
23 Girl
Light
9
81%
12%44%
Bumps head against back of high chair repeatedly
5.40
11 Girl
Light
7
76%
19%45%
Chews on box vacantly, scans the room intently
5.26
14 Boy
Light
11
76%
18%36%
Manipulates train
5.24
16 Boy
Light
11
70%
16%37%
Restless in high chair, stretches, babbles
5.23
12 Boy
Light
6
80%
24%41%
Touches father's face, smiles
4.29
19 Girl
Light
8
47%
33%26%
Flaps arms excitedly, stick in hand hits xylophone
4.28
15 Girl
Dark
6
52%
26%22%
Touches toy, grabs it, puts it in mouth
4.23
21 Girl
Dark
10
48%
28%20%
Pulls doll, makes eye contact with mother
4.21
9 Girl
Light
8
46%
26%20%
Pats floor, points (?), flaps arms
4.20
7 Boy
Light
10
52%
34%22%
Picks up ball, drops or throws it
4.19
1 Boy
Light
11
49%
32%21%
Mother speaks, child drops or throws toy
4.01
27 ?
Dark
10
36%
28%16%
Mother claps, infant scratches head
3.74
13 Girl
Dark
10
43%
37%12%
Responds to mother's smile, lifts arms
3.67
26 ?
Light
7
30%
36%15%
Stares, smiles, falls on side
Dark
6
29%
43%13%
Stretches legs, rubs feet together
3.44
8 Girl
3.34
25 ?
Dark
8
33%
47%9%
Rolls ball, pushes it off highchair tray
3.18
18 Boy
Light
8
30%
49%12%
Mother claps child's hands, child not attentive
Light
11
29%
54%2%
Mother moves child's hands, child not attentive
Light
5
16%
63%6%
Watches father touch toy, coughs, squirms
3.10 2.55
Clip
2 ? 20 Boy
Page 251
definition of intentionality in which one infers that the infant has something in mind. Finally, intentionality was awarded when the infant repeated a controlled action. For example, many mothers assigned intentionality to a clip in which a boy bangs his hands on the highchair tray several times or one in which a girl bumps her head against the back of the high chair repeatedly. This implies parental sensitivity to the infant's ability to control his or her ongoing activity. Infant Age The intentionality score was correlated with the age of the infant in the clip, r (26) = .39, p < .05. This finding is difficult to interpret because we had attempted to counterbalance age with intentionality when we selected the clips, but counterbalancing was difficult because age limits the range of intentional behaviors that an infant emits and the probability that an intentional event would occur during our filming. From this perspective, it is notable that some young infants (i.e., 6 or 7montholds) were rated as highly intentional in some clips. Also, one of the oldest infants participated in one of the least intentional clips. Infant Sex and Skin Color Similar difficulty in interpretation occurs when we look at the sex or skin color of the infant in the clip. Infant facial features and clothing can be used to categorize most infants in the clips as boys or girls but some are ambiguous. We confirmed these classifications using several independent judges. Also, some infants are obvious members of a particular ethnic group but many are ambiguous, particularly when a parent is not present. For the present analysis, we categorized the infants based on skin color per se, which is a more compelling variable than ethnicity. The EuropeanAmerican and Hispanic infants in the IIT were all categorized as having light skin and most of the AfricanAmerican infants had dark skin. These classifications were confirmed using independent judges. We had attempted to counterbalance level of intentional behavior across sex and skin color, but we were apparently unsuccessful. Of the 11 clips rated as the most intentional, 6 were clips of girls, 3 were ambiguous, and 2 were clips of boys. One way to quantify this effect is to calculate separate scores for each type of clip (i.e., IIT ratings across clips of girls, boys, or ambiguous infants). An ANOVA on these scores confirmed a main effect of clip, F (2, 242) = 12.09, p < .01, with average scores of 5.08 (SD= 1.52), 4.75 (SD= 1.41), and 4.71 (SD= 1.28), respectively: Girls were rated as more intentional than boys or ambiguous infants. The counterbalancing across race was not successful either: Despite our best efforts to have comparable distributions of skin color within each level of intentionality, the clips that the mothers rated as highly intentional contained a disproportionate representation of lightskinned infants. As indicated in Table 13.1, 12 of the 14 high intentional
Page 252
clips contained a lightskinned infant. An ANOVA on mean scores calculated across clips containing infants with light or dark skin indicated a significant difference, F (1, 121) = 75.98, p < .01, with average scores of 5.07 (SD= 1.38) and 4.42 (SD= 1.41), respectively. One cause of the disproportionate representation of girls and lightskinned infants in the highly intentional clips could be an unbalanced selection of clips (i.e., the clips that we chose with girls or with lightskinned infants are more likely to contain intentional behaviors). A second cause could be that mothers have a bias to attribute more intentionality to girls and to lightskinned infants. There are some reasons to expect maternal bias. First, regarding heightened expectations of intentionality in girls, Feldman and Reznick (1996) reported the same effect in a previous version of the IIT using an entirely different set of clips. Moreover, various studies indicate that parents often have predictable expectations about sex differences in infants. For example, Rubin, Provenzano, and Luria (1974) and Reid (1994) found gender stereotypes in the parents of newborns (e.g., that sons will be strong and that daughters will be fragile). Also, Grant (1994) found sexspecific differences in how parents interact with their infants (viz., they stimulate, initiate, and goad boys and are responsive and interactive with girls). The present results point to a possible connection between PPII and the sex of the infant that is being rated. Further research is needed to elucidate this relation and determine whether PPII causes parents to respond differently to boy and girl infants or is itself the result of some more fundamental process that affects both PPII and gender stereotypes. There is also reason to expect different attributions of intentionality as a function of skin color. Williams and Morland (1976) note that idiomatic speech reveals ample usage of the word black to signify negative evaluation (e.g., a black mark, to blacklist, to blackball, blackmail, black sheep) and white to signify positive evaluation (e.g., white hope, white knight, a white lie). Kenneth Gergen (1967) reported numerous examples of white as positive and black as negative in the religious practice and symbolism in nonWestern groups and similar distinctions emerge in JudeoChristian imagery, literature, and mass media (see Williams & Morland, 1976, for an extended discussion of this topic). This bias is also consistent with the dominant conceptual framework in psychology in which White, middleclass prototypes are accepted as the standard for normal development, and anything else is regarded as inferior (García Coll et al., 1996). And, while it is not inconceivable that our efforts to counterbalance race and intentionality might have failed, it is hard to accept that we would have failed so definitively. Indeed, inspection of the clips with dark skinned infants suggests that some of these clips received lower average ratings than did clips showing comparable behaviors in lightskinned infants. These data raise the possibility that some adults are biased toward attributing
Page 253
more intentionality to infants who are girls and infants who are lightskinned. Definitive proof of this effect would require creative methodology in which the level of intentional behavior could be matched explicitly across clips that vary in sex or skin color. Do Parents Agree on Their Attributions of Intentionality? Although there was consensus on the intentionality ratings for many clips, this statement must be qualified by three observations. First, the consensus was notably stronger for clips that seem intentional than for clips that seem not intentional. Table 13.1 lists the percentage of mothers who assigned a score of 7 or 8 (i.e., the highest ratings of intentionality) or a score of 1 or 2 (i.e., the lowest ratings of intentionality). Nine of the 14 intentional clips recruited 50% or more ratings of 7 or 8. In contrast, only 2 of the 14 notintentional clips evoked a score of 1 or 2 from more than 50% of the mothers. Thus, in some absolute sense, many mothers tended to perceive some infant behaviors as intentional, but mothers were less willing to agree on the infant behaviors that were not intentional. In an earlier study (Feldman & Reznick, 1996), we reported this same tendency to agree on intentionality and not agree on the lack of intentionality on the basis of IIQ data. This asymmetry could reflect a disjunctive decision rule in which a cluster of positive cues provide compelling evidence for intentionality whereas isolated cues are not widely viewed as diagnostic. A second qualification that applies to general statements about consensus is that the more prominent aspect of the IIT is its tendency to evoke variability in PPII. The distribution of answers was highly skewed for some clips, but every clip evoked every response at least once. Some clips were relatively balanced, evoking an almost rectangular distribution, and some evoked strong disagreement. For example, in Clip 19, a girl holds a stick in her hand and flaps her arms excitedly. The flapping movement causes the stick to hit a xylophone, which makes a loud sound, and this sound seems to encourage the girl to flap her arms with increasing gusto. Thirty three percent of mothers rated this clip at 1 or 2, interpreting the girl's actions as accidental. Twentysix percent of mothers rated this clip at 7 or 8, apparently assuming that the girl was performing the action on purpose. And, even when there was considerable consensus that a clip was intentional or not intentional, some mothers rated the clip in the opposite direction. For example, Table 13.1 indicates that 7% of mothers used a very low score (i.e., 1 or 2) to describe the clip that evoked the highest average intentionality score, and 6% of mothers used a very high intentionality score (i.e., 7 or 8) to describe the clip that evoked the lowest average intentionality score. Given that we intended that the IIT be a measure of individual difference, variability is expected and suggests that
Page 254
many clips provided participants with an opportunity to project intentionality rather than just detect it, as would be expected if intentionality is ascribed by the viewer. On the other hand, the lack of consensus regarding the most extreme clips raises the possibility that some mothers were responding randomly either due to lack of motivation or because they failed to understand the task. To explore this possibility, we defined mothers as anomalous if they responded with a 1 or 2 on any of the four items rated as most highly intentional or with a 7 or 8 on any of the four items rated as least intentional. Only 3 of 130 mothers were anomalous from both perspectives, which suggests a low rate of haphazard response. We compared each group of anomalous mothers (i.e., the mothers who responded anomalously on high intentional items and the mothers who responded anomalously on low intentional items) with the notanomalous mothers on their overall IIT score and on their scores for the items rated as highly intentional or not highly intentional. Mothers who were anomalously low on one or more of the 4 most intentional items gave significantly lower ratings than other mothers on their overall IIT score, their score for the 14 highly intentional IIT items and their score for the 14 least intentional IIT items, and the converse was true for mothers who were anomalously high on one or more of the 4 least intentional items. This suggests that most mothers tended to be consistent in their rating of infants as either likely or unlikely to behave intentionally. Third, despite various differences in PPII across specific clips, the data support the claim that an IIT score combined across clips is meaningful. Cronbach's coefficient alpha (Cronbach, 1951) was .93, suggesting strong consistency across clips. A principal component analysis indicated a single component (eigenvalue = 10.85) that accounted for 39% of the variance. This consistency is important. It indicates that although mothers are sensitive to the criteria that differentiate among the clips, the tendency to be strict or lenient in the ascription of intentionality is a consistent general bias that influences parent perception across clips. A subset of the mothers who completed the IIT also completed the IIQ. Version 4.2 of the IIQ contains questions about infant intentional behaviors but portrays intentional behaviors as either positive or negative. Instances of negative intentionality can be described verbally but are difficult to capture visually, so the IIT would seem to be primarily about positive intentionality. Despite the presence of these two types of items within the IIQ, the Cronbach coefficient alpha for the 32 IIQ items was .88. When items were separated into positive and negative, the alpha for the 24 positive items was .88 and for the 8 negative items was .80. A principal components analysis of the IIQ items confirmed the unique character of the negative items. Five components can be identified that account for 50% of the
Page 255
variance in the IIQ. Component 2 loaded exclusively on the 8 negative intentionality questions and included no other questions. The other components captured various aspects of positive intentionality. Do Aspects of the Parent Influence Perception of Infant Intentionality? We looked at factors in the mother that could be expected to influence her perception of infant intentionality. This analysis is based on the IIT and also on the IIQ. Education Our previous research (Feldman & Reznick, 1996) indicated a negative correlation between PPII and education because several very highly educated mothers (doctors and lawyers) were hesitant to attribute intentionality to infants. We assumed that these mothers sought more compelling evidence to support a claim of intentional behavior and thus attributed relatively little intentionality. The present sample, which had a notably larger range of educational attainment with few extremely high scores, indicated a significant positive correlation between IIT and maternal education, r (100) = .23, p < .05. Also, mothers with more education attributed less negative intentionality in the IIQ, r (90) = .30, p < .01. This relation seems more reasonable than our previous finding: Across a normal range of education, mothers who have had more intellectual training are more sensitive to the possibility of intentionality in an infant's behavior and are unwilling to believe that infants can do things on purpose to be mean or spiteful. This effect also emerged in a measure that reflects the effect of education. Peabody Picture Vocabulary Test (Dunn & Dunn, 1959) scores, which were available for a limited subset of mothers, were positively correlated with IIT scores, r (64) = .31, p < .01, and negatively correlated with negative intentionality in the IIQ, r (63) = .36, p < .01. This suggests that mothers who are more educated and/or literate (in English) are more inclined to see infants as intentional (or to detect subtle signals of intentionality) and are less likely to view this intentionality as negative (i.e., they attributed less negative intentionality to infants). Finally, as reported in previous work, IIT scores were positively correlated with the mother's age, r (100) = .23, p < .05. Negative intentionality scores were negatively correlated with the mother's age, r (87) = .24, p < .05. Ethnicity The mother's ethnicity also affects her tendency to perceive intentionality. Almost all mothers in the present sample identified themselves as either Caucasian (which we have termed EuropeanAmerican), AfricanAmerican, or Hispanic. Our sampling strategy tends to yield a EuropeanAmerican group that is more highly educated than the African
Page 256
American or Hispanic mothers, but there is considerable range within each group. To form interpretable categories, we divided mothers into those who reported 12 years of education or less (which includes having a Graduate Equivalency Degree) versus those who had attended some college or professional training.2 An Education Level × Ethnic Group ANOVA on IIT scores indicated a significant main effect for ethnicity, F (2, 121) = 6.83, p .01. As indicated in Table 13.2, European American mothers and AfricanAmerican mothers were comparable, and both groups had higher IIT scores than Hispanic mothers. The average score for more educated mothers was somewhat higher than the score for less educated mothers, but the effect was not significant. More important, education did not interact with ethnicity. An alternative view of this effect is to use individual clips as the level of analysis rather than the IIT score averaged across clips. This approach allows us to determine if the influence of maternal characteristics is specific to certain types of clips. Education did not interact with scores across individual clips but ethnicity did interact with clip scores, F (54, 3449) = 1.92, p < .01. To explore this interaction, we used separate ANOVAs to test for ethnicity effects in each clip. The pattern of effects was remarkably consistent: Eleven of the 14 high intentional clips had a significant ethnicity effect (p < .01) due to Hispanic mothers with lower scores than European American or AfricanAmerican mothers. There were no effects of the mother's ethnicity among the less intentional clips. This analysis indicates that the low IIT scores for Hispanic mothers reflects an unwillingness to ascribe intention to acts recognized as intentional by mothers in the other groups rather than a general filter that lowers intentionality in all circumstances. For an alternative view of this effect, we computed separate scores for the 14 high intentional clips and the 14 low intentional clips and analyzed these two scores in a repeatedmeasures ANOVA. The Ethnicity x Intentionality interaction was statistically significant, F (2, 121) = 9.03, p .01 and can be decomposed into a significant ethnicity effect for the mean score across the highly intentional clips, F (2, 121) = 11.22, p .01, and no effect of ethnicity across the less intentional clips. Consistent with the findings presented thus far, Hispanic mothers had the lowest scores. The maternal ethnicity effect across individual clips described previously could reflect ethnic differences in attributions about highly intentional infants, but given that girls and lightskinned infants were more likely to be viewed as intentional, the effect could be more complicated. We com 2
It is difficult to equate educational attainment in different ethnic groups. We are currently gathering additional data to refine these classifications. The present analyses should be regarded as preliminary.
Page 257 TABLE 13.2 PPII Scores for Each Ethnic Group
IIT
M
5.17
SD
1.34
N
41
M
3.35
IIQpositive
IIQnegative
PIIB
SD
AfricanAmerican
.61
N
34
M
2.01
SD
.87
N
34
M
1.94
SD
.14
N
32
=
=
>
>
EuropeanAmerican 5.25 .89 52 3.31 .49 30 1.28 .31 30 1.86 .12 29
>
>
<
=
Hispanic 3.99 1.58 34 2.70 .79 28 1.92 .78 28 1.85 .16 23
puted two additional ANOVAs, one with separate scores for girls and boys and one with separate scores for lightskinned and darkskinned infants. These ANOVAs revealed the main effects of ethnicity as just described, and the main effects of infant sex and infant skin color, but there was no interaction with ethnicity: European American mothers, AfricanAmerican mothers, and Hispanic mothers rated the clips of girls as more intentional than the clips of boys and the clips of lightskinned infants as more intentional than the clips of darkskinned infants. The ethnicity effect revealed in the IIT was consistent with effects for the IIQ, but this analysis must be regarded as tentative at present because of relatively small group sizes for the highly educated Hispanic mothers (N = 7) and the less educated EuropeanAmerican mothers (N = 9). Because of the small sample size for these two groups, we did not use education level as a term in the analysis of IIQ. This causes some degree of confounding between ethnic group and education, with EuropeanAmerican mothers being more educated and Hispanic mothers being less educated. A repeatedmeasures ANOVA on the IIQ scores indicated a main effect of ethnic group, F (2, 89) = 4.77, p .01, and question type (i.e., positive vs. negative), F (1, 89) = 353.13, p < .01, but these two effects interacted, F (2, 89) = 22.89, p < .01. The ethnicity effect for the positive IIQ questions, F (2, 89) = 9.49, p .01, emerged because AfricanAmerican mothers and EuropeanAmerican mothers had higher positive IIQ scores than Hispanic mothers (as indicated by a StudentNewmanKeuls test; see Table 13.2). The ethnicity effect for negative questions was also significant, F (2, 89) = 9.96, p < .01, but the pattern of means was different. AfricanAmerican and Hispanic mothers rated infants as more negatively intentional than did EuropeanAmerican mothers.
Page 258
Finally, we looked at scores on the PIIB in which mothers attempted to predict how their own infant would perform in situations that evoke intentionality. There was a main effect of ethnicity, F (2, 81) = 3.76, p < .05. As indicated in Table 13.2, AfricanAmerican mothers rated their infants as more intentional than did either EuropeanAmerican mothers or Hispanic mothers. The effects of ethnicity should be interpreted from a crosscultural perspective but with considerable caution. As noted already, our recruitment procedures caused a confounding of ethnicity and education such that the Hispanic mothers in our sample tend to be less educated than either the AfricanAmerican mothers or the EuropeanAmerican mothers. Second, all of the Hispanic mothers could communicate in English but for many it was their second language. Thus, they may have had less understanding of the questions used in each of the measurement procedures. Finally, our assessments were conducted in a laboratory either at Yale or at the Hill Health Center. This context was unnatural for all mothers, but may have evoked different levels of anxiety, compliance, socially desirable responding, and candor from the different groups of women. These caveats notwithstanding, Hispanic mothers considered their infants to be less intentional than did other mothers across the two primary measures of positive intentionality. One interpretation of this effect is that the Hispanic mothers are less knowledgeable regarding infant development. For example, Pachter and Dworkin (1997) reported that Puerto Rican mothers tend to expect children to attain milestones such as feeding from a spoon, smiling at a face, or recognizing their mother at a later age than do mothers for other ethnic groups. However, the fact that milestones are misstated systematically (i.e., they are predicted to be relatively late) rather than randomly suggests a broader interpretation. Hispanic culture places considerable emphasis on group affiliation and feelings of identification, loyalty, and solidarity among family members (Harrison, Wilson, Pine, Chan, & Buriel, 1990). Thus, many Hispanic mothers may derive potency, pleasure, and prestige from their nurturing behavior, and the infant's dependency enhances the mother's sense of self by providing her with a better opportunity to fulfill her role as a nurturer. In the present context, Hispanic mothers might tend to understate the infant's ability to do things deliberately, particularly in comparison to other mothers who see their mission as to encourage the infant's autonomy or foster the infant's intellectual development. This hypothesis is speculative but is consistent with other findings. For example, Field and Widmayer (1981) examined motherinfant interactions during feeding and facetoface play and found that Puerto Rican mothers used more ''baby talk" (i.e., exaggerated intonations and brief phrases with very little intentional content), displayed more contingent responsivity, and played more social games.
Page 259
AfricanAmerican mothers rated their infants as more intentional on the measures of negative intentionality and in the context of predicting specific behaviors. Many of the cautions described previously are also relevant here, but this effect suggests that the AfricanAmerican mothers in our sample are more willing to view infants as capable of being annoying or difficult on purpose. This may be related to the finding that many AfricanAmerican mothers are concerned with the possibility of "spoiling" an infant by providing too much attention (Field & Widmayer, 1981; Solomon, Martin, & Cottington, 1993). Moreover, because many of the African American and Hispanic mothers in the present sample were young, previous findings about adolescent mothers may also be relevant. For example, Carter, Osofsky, and Hann (1991) noted that adolescent mothers often have unrealistic expectations about their infants (e.g., encouraging very young infants to hold their own bottles and feed themselves, being insensitive to the child's growing need for autonomy, exercising excessive control over the child's locomotion and exploration). And, Karraker and Evans (1996) found that many adolescent mothers have poor knowledge of infant milestones and are not accurate in predicting their infant's behavior. From the present perspective, the willingness to see infants as behaving negatively on purpose could be related to a more punitive style of childrearing in which infants must be punished for willful transgressions in order to avoid "spoiling" them and to keep them under control. For example, Dix, Ruble, and Zambarano (1989) found that mothers of 4 to 12monthold infants often base their discipline decisions on inferences about the infant's competence and responsibility for negative acts. If this decision is based on a faulty inference (e.g., that an infant is overly intentional), then discipline could be entirely inappropriate. Mental Health Finally, our previous research indicated a tendency for women with more symptoms of depression or anxiety to report less infant intentionality, but the effects have been small and scattered across various measures. The present data are consistent with previous findings: Beck Depression Inventory scores were negatively correlated with IIT scores, r (105) = .25, p < .01, and positively correlated with negative intentionality scores in the IIQ, r (92) = .24, p < .05. Beck Anxiety Inventory scores were not related to IIT scores but were correlated with negative intentionality scores in the IIQ, r (92) = .22, p < .05. It seems likely that a depressed or anxious mother might be insensitive to her infant's emerging intentionality either because of a lack of interest in the infant's development or because an intentional infant is more demanding than an infant who is passive or docile. Or, the effect could be reversed: Depressed or anxious mothers might provide little stimulation or contingent response and thus undermine their child's intentionality. Careful clinical work is needed to unravel this phenomenon.
Page 260
General and Specific Perception of Infant Intentionality The second broad question addressed here is whether parents separate their perception of their own infant's intentionality from a more general belief about the intentionality of infants. Our data suggest that many parents of young infants do not make this distinction. In a study of 40 motherinfant dyads tested at 4 and 8 months, the correlation between the IIT and the IIQ was a hefty .54 and .57 (Feldman & Reznick, 1996). In the most recent version of the IIQ, we attempted to focus the parent on his or her own infant by removing questions that were worded too generally. Still, for the present sample, the IIT and the IIQ were positively correlated, r (81) = .26, p < .05. The positive relation was particularly strong for the IIQ questions that tap positive aspects of intentionality, r (81) = .35, p .01. Many mothers also completed the PIIB, which is a radically specific view of their infant's intentionality in the sense that respondents are asked to predict how their infant will respond in a particular situation. As noted earlier, we chose the situations and response alternatives based on our own interpretation of how an intentional infant would act in specific situations (e.g., a relatively intentional 9monthold would attempt to roll a ball back to an experimenter or would try repeatedly to retrieve a toy through a slot that was a bit too small). Scores on the PIIB did not correlate with the IIT scores, but did correlate positively with the IIQ, r (80) = .34, p .01. This pattern of correlations is consistent with the view that our current instruments are tapping PPII at different but related levels: The IIT captures the mother's general view, the PIIB captures the mother's specific expectations about her own infant, and the IIQ has aspects of both. It is interesting to note that the PIIB score correlated positively with both aspects of the IIQ: r (80) = .31, p < .01 with the positive IIQ items, and r (80) = .21, p < .05 with the negative IIQ items. This suggests that the PIIB may be picking up both positive and negative aspects of intentionality. When the sample size in the present research is larger, these claims can be investigated using structural equation modeling. The notable finding here and in previous work is that despite some degree of differentiation, PPII in a general sense (i.e., ratings of anonymous babies) and PPII in a specific sense (i.e., questions about the parent's own child) have a high degree of overlap. This overlap between general and specific views vaults us into an intriguing zone of speculation. One interpretation of the overlap is that some parents are less likely than others to attend to aspects of their infant's behavior that could support an ascription of intentionality (i.e., they do not notice their infant's tendency to manifest behaviors that suggest intentionality) so they draw on their general opinion about infant capabilities when answering specific questions about their
Page 261
infant. A second interpretation is that each parent uses somewhat different criteria for deciding if an infant's behavior is intentional, and they use these same criteria for both their own and other's children. Finally, it could be that parents are giving an accurate report of their infant's level of intentionality, but the parent's general beliefs about infant intentionality have led them to treat their infant in a way that influences the infant's development of intentionality (e.g., the parent has been particularly responsive or has provided appropriately challenging play materials), thus causing an alignment of the infant's intentional behaviors and the parent's general views about infant intentionality. Parent Perception of Intentionality As Measurement or Intervention? The third and final question addressed here is whether the assessment of PPII should be regarded as a measurement or an intervention. When we ask parents to rate infant intentionality using the IIT, we must first provide an explanation of the construct. In an early study (Feldman & Reznick, 1996), we noticed that many parents were intrigued by our presentation of the idea that infants may differ in their ability to do things deliberately. Moreover, in a followup study in which we assessed PPII after providing a less comprehensive explanation, the effects were disappointing, suggesting that we had not measured the construct as well. With the IIQ, the questions themselves paint a vivid picture of what it would mean to claim that an infant is intentional. An intriguing question looms: Do our measurement techniques assess the perception of intentionality or do they create the perception of intentionality (i.e., teach parents how to recognize or ascribe intentional behaviors)? To answer this question, Schwartz (1997) administered an early version of the IIQ to one longitudinal sample of mothers when their infants were 8, 10, and 12 months, a second longitudinal group when their infants were 10 and 12 months, and a comparison group tested only at 12 months. The results indicated that IIQ scores increased with age for the 81012month longitudinal comparison, F(2, 68) = 81.91, p < .01, and for the 1012month longitudinal comparison, F(1, 19) = 31.07, p < .01, but not for the crosssectional comparison. In other words, mothers who completed the IIQ for the first time when their infant was 8, 10, or 12 months old reported the same level of intentionality. At 10 and 12 months, IIQ scores were affected by the number of previous assessments. Indeed, the three cohorts were significantly different at 12 months, F (2, 74) = 4.67, p < .05, with IIQ scores proportional to the number of previous assessments. So, from a longitudinal perspective, the number of previous administrations of the
Page 262
IIQ is a better predictor of IIQ score than is the age of the infant, again suggesting that this is a matter of willingness to ascribe intentionality rather than the perception of something that is really there. These results can be contrasted with a different parentreport measure that was administered to the same mothers at the same time as the IIQ: The MacArthur Communicative Development Inventory (Fenson et al., 1993) was used to measure the infant's ability in word production, word comprehension, and gestural naming. For the language assessment, scores increased over time for the 81012month longitudinal comparison, F (2, 68) = 87.26, p < .01, and for the 1012month longitudinal comparison, F (1, 17) = 24.20, p < .01. But, in contrast to the IIQ scores, the language scores were significantly different in a crosssectional analysis, F (2, 67) = 38.25, p < .01. Also, language scores for the 2 groups of mothers assessed when their infant was 10 months and the 3 groups of mothers assessed when their infant was 12 months did not differ as a function of the number of previous assessments. This result indicates that providing a parent with an opportunity and perhaps an inducement to ascribe intentionality in a particular circumstance may affect the parent's subsequent perception. From a methodological perspective, the measurement of PPII is reactive, raising the parent's consciousness about the possibility that infants could be regarded as intentional. Thus, we must be cautious in our efforts to measure PPII, particularly in a longitudinal context, and search for methods that would assess PPII indirectly (e.g., by analyzing the language that parents use in a seemingly irrelevant context such as narrating a videotape). From an interventionist perspective, there are certainly circumstances in which it would be advantageous to alter PPII. As noted earlier, Kaye (1982), Bruner (1973, 1975), and Tomasello (1992, 1996) argued persuasively that by treating the infant as intentional, parents facilitate various aspects of the infant's cognitive development. There are also reports that suggest that PPII can be problematic, particularly at the extremes. Parents who abuse children have more unrealistic expectations about child development (Azar, Robinson, Hekimian, & Twentyman, 1984; Twentyman & Plotkin, 1982), attribute malevolent intention to children (Bauer & Twentyman, 1985), are more likely to use powerassertive and negatively oriented strategies for controlling their child (Oldershaw, Walters, & Hall, 1986), and are less likely to respond contingently to the infant and to demonstrate enjoyment of and responsiveness to the infant's goaldirected behavior (Crittenden & Bonvillian, 1984; LyonsRuth, Connell, Zoll, & Stahl, 1987). Also, clinicians who work with parents who abuse or neglect a child report that from the parent's perspective, the rationale for their behavior is often stated in terms of intentionality: Some parents may hold an inappropriately overstated view of infant intentionality in which the infant must be pun
Page 263
ished for intentional misdemeanors, taught who is boss, or brought into line with some ethical precepts (e.g., Helfer, McKinney, & Kempe, 1976; Pollock & Steele, 1972). On the other extreme, some parents may hold an inappropriately understated view of infant intentionality in which it is easy to ignore the infant's interactional bids or need for stimulation and mastery experiences. From this perspective, the fact that parents attribute more intentionality when these sorts of behaviors are made salient offers a useful clue as to how an effective intervention might be structured. Conclusion The present data confirm previous claims that parents use goaloriented action and participation in social games as clues that suggest infant intentionality. We also found that perceptual vigilance and controlled action are viewed as evidence suggesting intentionality. Further research is needed to identify the specific psychological mechanisms that emerge during the first year and allow infants to perform these sorts of behaviors. It is not surprising that intentionality is more likely to be ascribed to older infants, but the apparent effects of sex and skin color are intriguing. We could propose various mechanisms that might explain these effects either as accurate descriptions of inherent differences in infants, projections of stereotypic attitudes held by parents, or as artifacts of the present measurement strategy. However, we prefer to adopt the more prudent strategy of acknowledging the need for replication and experimental manipulation as the necessary next steps. From the perspective of PPII as an explanatory gloss, the present analyses confirm and extend previous efforts to identify factors that influence the parent's willingness or ability to ascribe intentionality to infants. Age and education play a role, suggesting that infant intentionality is something we learn to appreciate. Indeed, even providing a second or third opportunity to answer questions about intentionality appears to have some influence on this practice. The present sample includes mothers only, but it seems likely that fathers would show the same pattern of effects. The effects of culture, as revealed in the analysis of ethnicity, will no doubt become more complicated when we have a large enough sample size to isolate the influences of education and age. However, the present results point to a tendency for Hispanic mothers to view infants as less intentional and for AfricanAmerican mothers to be more likely to ascribe negative intentions. It will be interesting to see if these differences are related to differences in the infant's behavior and to subsequent psychological development. The same is true for effects related to symptoms of depression and anxiety.
Page 264
Finally, we note that parents may be able to separate the intentionality of their own infant from their view of infants in general, but there is a good degree of overlap. This relation is important because it reflects the bidirectional nature of the present phenomenon. Parents are likely to make attributions about the intentionality of their infant's behavior. These attributions are based on both the infant's ongoing behavior capacities and the parent's criteria for the attribution or ascription of intentionality. Moreover, the parent's response to the infant's behavior is influenced by the parent's perception of the intentionality of that behavior, and the parent's response may encourage or discourage the infant's emerging intentionality. Thus, returning to the theme of this volume, there is much to be gained from embracing the claim that intentions develop in a social world. Whether or not infants are capable of intentional behavior, there can be no doubt that parents tend to regard infants as intentional, and it seems likely that they temper their behavior accordingly. When we focus on intentionality as a parent perception, we discover useful clues about the specific infant behaviors that are taken as indicating intentionality and the ways that parental ascription of intentionality is likely to encourage or thwart the infant's development. Acknowledgments I am grateful to the W. T. Grant Foundation for their support of this research. Helpful editorial comments were provided by David Olson, Cynthia Garcia Coll, and Robin Harwood. The present research grew out of a lively discussion group at Yale in the early 1990s called IntentionaliTea. Suzanne Zeedyk conducted her dissertation on the topic and published an excellent survey of developmental accounts of intentionality (Zeedyk, 1996). More recently, my research involved collaborations with Ruth Feldman, Bena Brandwein Schwartz, Nicole Eldridge, Stephanie Aubry, Christine Fingado, and Nicole Straight during their time at Yale University. Each of these individuals has contributed to the present research. References Anscombe, G. E. M. (1957). Intention. London: Blackwell. Azar, S. T., Robinson, D. R., Hekimian, E., & Twentyman, C. T. (1984). Unrealistic expectations and problemsolving ability in maltreating and comparison mothers. Journal of Consulting and Clinical Psychology, 52, 687691. Bakeman, R., & Adamson, L. (1984). Coordinating attention to people and objects in motherinfant and peerinfant interactions. Child Development, 55, 1278 1289.
Page 265
Bates, E., Benigni, L., Bretherton, I., Camaioni, L., & Volterra, V. (1979). The emergence of symbols. New York: Academic Press. Bauer, W. D., & Twentyman, C. T. (1985). Abusing, neglectful, and comparison mothers' responses to childrelated and nonchildrelated stressors. Journal of Consulting and Clinical Psychology, 53, 335343. Brentano, F. C. (1973). Psychology from an empirical standpoint. (A. C. Rancurello, D. B. Terrell & L. L. McAlister, Trans.). New York: Humanities Press. (Original work published in 1874) Bruner, J. S. (1973). Organization of early skilled action. Child Development, 44, 111. Bruner, J. S. (1975). The ontogenesis of speech acts. Journal of Child Language, 2, 119. Bruner, J. S. (1983). Child's talk: Learning to use language. New York: Norton. Butterworth, G., & Cochran, E. (1980). Towards a mechanism of joint visual attention in human infancy. International Journal of Behavioral Development, 3, 253272. Campos, J. J., & Stenberg, C. R. (1981). Perception, appraisal and emotions: The onset of social referencing. In M. E. Lamb & L. R. Sherrod (Eds.), Infant social cognition (pp. 333373). Hillsdale, NJ: Lawrence Erlbaum Associates. Carter, S. L., Osofsky, J. D., & Hann, D. M. (1991). Speaking for the baby: A therapeutic intervention with adolescent mothers and their infants. Infant Mental Health Journal, 12, 291301. Crittenden, P. M., & Bonvillian, J. D. (1984). The relationship between maternal risk status and maternal sensitivity. American Journal of Orthopsychiatry, 54, 250262. Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16, 297234. Dennett, D. (1987). The intentional stance. Cambridge, MA: MIT Press. Dix, T., Ruble, D. N., & Zambarano, R. J. (1989). Mothers' implicit theories of discipline: Child effects, parent effects, and the attribution process. Child Development, 60, 13731391. Dore, J., Franklin, M. B., Miller, R. T., & Ramer, A. L. H. (1976). Transitional phenomena in early language acquisition. Journal of Child Language, 3, 1328. Dunn, L. M., & Dunn, L. M. (1959). Peabody Picture Vocabulary Test. (Available from American Guidance Service, P.O. Box 99, Circle Pines, MN 550141796) Eldridge, N. E. (1997). Parental attributions of infant intentionality. Unpublished manuscript, Yale University. Emde, R. N. (1984). The affective self: Continuities and transformations from infancy. in J. D. Call (Ed.), Frontiers of infant psychiatry. New York: Basic. Feinman, S. (1982). Social referencing in infancy. MerrillPalmer Quarterly, 28, 445470. Feldman, R., & Reznick, J. S. (1996). Maternal perception of infant intentionality at 4 and 8 months. Infant Behavior and Development, 19, 483496. Fenson, L., Dale, P., Reznick, J. S., Thal, D., Bates, E., Hartung, J. P., Pethick, S., & Reilly, J. S. (1993). The MacArthur Communicative Development Inventories: Users guide and technical manual San Diego, CA: Singular Press. Field, T. M., & Widmayer, S. M. (1981). Motherinfant interactions among lower SES black, Cuban, Puerto Rican, and South American immigrants. In T. M. Field, A. M. Sostek, P. Vietze, & P. H. Leiderman (Eds.), Culture and early interactions (pp. 4160). Hillsdale, NJ: Lawrence Erlbaum Associates. Garcia Coll, C., Lamberty, G., Jenkins, R., McAdoo, H. P., Crnic, K., Wasik, B. H., & Vázquez Garcia, H. (1996). An integrative model for the study of developmental competencies in minority children. Child Development, 67, 18911914. Gergen, K.J. (1967). The significance of skin color in human relations. Daedalus, 96, 390407. Grant, V. J. (1994). Sex of infant differences in motherinfant interaction: A reinterpretation of past findings. Developmental Review, 14, 126. Griffin, D. R. (1992). Animal minds. Chicago: University of Chicago Press.
Page 266
Harding, C. G., & Golinkoff, R. M. (1979). The origins of intentional vocalizations in prelinguistic infants. Child Development, 50, 3340. Harrison, A. O., Wilson, M. N., Pine, C.J., Chan, S. Q., & Buriel, R. (1990). Family ecologies of ethnic minority children. Child Development, 61, 347362. Helfer, R. E., McKinney, J. P., & Kempe, R. (1976). Arresting or freezing the developmental process. In R. E. Helfer & C. H. Kempe (Eds.), Child abuse and neglect (pp. 5573). Cambridge, MA: Ballinger. Karraker, K. H., & Evans, S. L. (1996). Adolescent mothers' knowledge of child development and expectations for their own infants. Journal of Youth & Adolesence, 25, 651666. Kaye, K. (1982). The mental and social life of babies. Chicago: University of Chicago Press. Lockman, J. J., Ashmead, D. H., & Bushnell, E. W. (1984). The development of anticipatory hand orientation during infancy. Journal of Experimental Child Psychology, 37, 176186. LyonsRuth, K., Connell, D., Zoll, D., & Stahl, J. (1987). Infants at social risk: Relations among infant maltreatment, maternal behavior, and infant attachment behavior. Developmental Psychology, 23, 223232. Malle, B. F., & Knobe, J. (1997). The folk concept of intentionality. Journal of Experimental Social Psychology, 33, 101121. Moore, C., & Dunham, P. (1995). Joint attention: Its origins and role in development. Hillsdale, NJ: Lawrence Erlbaum Associates. Mosier, C. E., & Rogoff, B. (1994). Infants' instrumental use of their mothers to achieve their goals. Child Development, 65, 7079. Murphy, D. J., & Messer, D. J. (1977). Mothers, infants and pointing: A study of a gesture. In R. H. Schaffer (Ed.), Studies in motherinfant interaction (pp. 323 354). New York: Academic Press. Oldershaw, L., Walters, G. C., & Hall, D. K. (1986). Control strategies and noncompliance in abusive motherchild dyads: An observational study. Child Development, 57, 722732. Pachter, L. M., & Dworkin, D. H. (1997). Maternal expectations about normal child development in 4 cultural groups. Archives of Pediatric and Adolescent Medicine, 151, 11441150. Pollock, C., & Steele, B. (1972). A therapeutic approach to the parents. In C. H. Kempe & R. E. Helfer (Eds.), Helping the battered child and his family (pp. 3 21). Philadelphia: Lippincott. Radner, D., & Radner, M. (1989). Animal consciousness. Buffalo, NY: Prometheus Books. Reid, G. M. (1994). Maternal sexstereotyping of newborns. Psychological Reports, 73, 14431450. Reznick, J. S. (1998). Predicting Infant Intentional Behavior, Version 1.2. (Available from J. Steven Reznick, CB# 3270, University of North Carolina, Chapel Hill, NC 275143270) Reznick, J. S., & Feldman, R. (1998). Infant Intentionality Questionnaire, Version 4.2. (Available from J. Steven Reznick, CB# 3270, University of North Carolina, Chapel Hill, NC 275143270) Rubin, J., Provenzano, R., & Luria, Z. (1974). The eye of the beholder: Parents' views on sex of newborns. American Journal of Orthopsychiatry, 44, 512519. Scaife, M., & Bruner, J. (1975). The capacity for joint visual attention in the infant. Nature, 253, 265266. Schwartz, B. B. (1997). Memory for locations: Developmental function and individual differences. Unpublished doctoral dissertation, Yale University. Searle, J. R. (1983). Intentionality. Cambridge, England: Cambridge University Press. Solomon, R., Martin, K., & Cottington, E. (1993). Spoiling an infant: Further support for the construct. Topics in Early Childhood Special Education, 13, 175 183. Thelen, E., Corbetta, D., Kamm, K, Spencer, J. P., Schneider, K., & Zernicke, R. F. (1993). The transition to reaching: Mapping intention and intrinsic dynamics. Child Development, 64, 10581098. Tomasello, M. (1992). The social bases of language acquisition. Social Development, 1, 6887.
Page 267
Tomasello, M. (1996). The cultural roots of language. In B. M. Velichkovsky & D. M. Rumbaugh (Eds.), Communicating meaning: The evolution and development of language (pp. 275307). Mahwah, NJ: Lawrence Erlbaum Associates. Trevarthen, C. (1979). Communication and cooperation in early infancy: A description of primary intersubjectivity. In M. Bullowa (Ed.), Before speech.: The beginnings of interpersonal communication (pp. 321347). Cambridge, England: Cambridge University Press. Twentyman, C. T., & Plotkin, R. C. (1982). Unrealistic expectations of parents who maltreat their children: An educational deficit that pertains to child development. Journal of Clinical Psychology, 38, 497503. Walden, T. A., & Ogan, T. A. (1988). The development of social referencing. Child Development, 59, 12301240. Willatts, P. (1984). Stages in the development of intentional search by young infants. Develomental Psychology, 20, 389396. Williams, J. E., & Morland,J. K. (1976). Race, color, and the young child Chapel Hill: University of North Carolina Press. Zeedyk, M. S. (1992). Parental perceptions of infant intentionality. Unpublished manuscript, Yale University. Zeedyk, M. S. (1994). Maternal interpretations of infant intentionality. Unpublished doctoral dissertation, Yale University. Zeedyk, M. S. (1996). Developmental accounts of intentionality: Toward integration. Developmental Review, 16, 416461. Zeedyk, M. S., & Reznick, J. S. (1996). Infant Intentionality Tape, Version 2.1 [Film]. (Available from J. Steven Reznick, CB# 3270, University of North Carolina, Chapel Hill, NC 275143270)
Page 269
Chapter 14— Intention and Emotion in Child Psychopathology: Building Cooperative Plans Jenny Jenkins Rachel Greenbaum University of Toronto Cognitive theories of intention contribute to our understanding of how one person can represent the beliefs, intentions, and goals of another person. Typically the interest in cognitive science is to examine the structural development of this representational system. This chapter switches the focus away from the structure of these representations to the reasons why they develop and their semantic content. We argue that cognitive structures for representing our own and others' intentions and goals developed to enable complex cooperation across multiple agents. Such a cognitive system allows for the construction and execution of joint plans (Oatley, 1992). Much of cognitive science has been concerned with understanding the individual mind. And yet to understand the major achievements of culture, science, and so on, we need to understand how multiple agents can work together on a joint goal, and how new goals emerge that are not the product of any one individual, but represent the emergent properties of individuals' minds in interaction (Hutchins, 1995). We use the term intention to mean the representation of a goal. The individual wants something and forms a plan to execute the goal. Such a plan may involve a sequence of actions. As the execution of the plan is underway, it will also involve adjustments to the plan, repairs, modifications of the original goal. The psychology of planning and action is made more complex and more exciting, however, because it occurs in a social context. Intentions are not typically just intentions of an individual in isolation. They often concern what we want from others, what we think they want from us, what we believe about ourselves and them.
Page 270
In this chapter we look at individual differences in children's abilities to accurately represent their own and others' intentions. Some children show major inaccuracies in how they think about the intentions of others. They also build up an intentional structure toward others that makes it difficult for them to execute a cooperative plan. We review data from the field of developmental psychopathology linking environmental adversities to the development of cognitive structures in which children show biases in their own intentions toward others and their understanding of others' intentions toward themselves. These subjects are not the typical subjects for the psychology of intention. However, we argue that by examining the intentional structure of children who show a problematic developmental trajectory, we gain an understanding of the primary function of our ability to represent the intentions of self and other: the creation of new pieces of knowledge (or objects in the world) that can only be created by the cooperation of individuals. The GoalCorrected Partnership How does a father who has just prepared dinner manage to get his two children, who are happily engaged in performing a play, to the table when he wants them there? How are two children, who do not really like one another but are asked to work on a project together at school, able to come up with a joint plan and get the project done? Although such events pose challenges to humans, we do manage these cooperative endeavors. We plan with one another. We think about what other people want and interleave our own goals accordingly. We build cooperative structures that allow for people to live together amicably and productively. This ''jointness" of planning and action requires something special. We have to be able to represent what the other person wants and why they want it. We have to be able to think about our own goals. Giving consideration to our own individual goals and the goals of the other, we have to agree on a joint goal. Sometimes we also need to make an explicit plan for how the goal will be achieved. Parents and their young children come to be able to construct these joint plans in the service of their goals. Bowlby (1971) called this the goalcorrected partnership. Bowlby is not usually mentioned in cognitive discussions of intention. We argue, however, that his theory of the goalcorrected partnership is very useful for understanding the development of intention. Such partnerships are based on schemas, that is to say cognitive/emotional structures, in which the goals or intentions of both the self and the other are represented. Such schemas allow people to construct both a shared semantics, and representations of internal states of the other. This theory implies
Page 271
that intentionality—the mental qualities of "aboutness," that is to say, of semantics and of being able to represent aspects of the world—arises not just because the mind somehow has the ability to represent things outside itself, but because such representations figure in the sharing of intentions and plans, in the first instance with caregivers. During infancy we see the beginnings of goalcorrected partnerships. Parents match their babies' vocalizations and actions with their own vocalizations and actions, sensitively timing them to create a finely tuned dance with their baby. The baby gets a feeling for what it is like to act cooperatively, to execute an action and to have someone else take it up and respond to it. On the basis of such synchrony, cooperative action feels familiar and rewarding (Stern, 1985). Language brings with it a much more complex tool for engaging in cooperative action. With language, children acquire both the means for representing and referring to objects in the outside world, and for negotiating around the internal states of other people. A 2yearold can find out about someone else's feelings and thoughts, and talk about his or her own feelings and thoughts (Dunn, 1988). A place between people is constructed in which the meanings and representations both of the tangible and the more intangible aspects of the world can be negotiated. Consider how a child as young as 2 years can negotiate about conflicting goals in such a way as to clarify his or her own and others' internal states and repair awkward moments. One of us (Jenkins) was babysitting a 2yearold one evening while staying in his home. He went to sleep and woke before his parents had returned. He started crying and I went to him. When he realized that I was not one of his parents, he started kicking and crying and shouting, "Go away, Go away. I don't want you. I want my daddy." Thenext morning he spontaneously started to talk about the incident: Ewan: I cried. Jenny: Yes, you were unhappy, weren't you? Ewan: I want my daddy. You not my daddy. Jenny: No, and you didn't want me there, did you? Ewan: No I sad. I cried. Jenny: Yeah. That's okay. I know you missed your daddy.
Then he gave me a train that he had been playing with so that I would play trains with him. He was doing a repair on our conflict. He was explaining himself and making sure that I was not mad at him. Such are the building blocks of being able to operate cooperatively. One person wants something different from another. They push for what that is. And then
Page 272
they wonder about the other person. Have I gone too far? Can we still plan together, work together effectively? With internal state language, it is possible to find out. I can ask you to let me know how the execution of my goals has affected you. Children's skills in representing the internal states of others become increasingly advanced in the preschool years. By about age 4, children can predict action on the basis of someone's belief even when it conflicts with their own (Astington, Harris, & Olson, 1988). You believe one thing, and I believe another. I know that you will act on the basis of your belief. Although children have a basic understanding of the nature of belief at this age, it is still a couple of years before they analyze complex social situations in terms of the intentions and beliefs of all actors (Case, 1991). By the early to middle school years, children have the cognitive skills to think about their own goals in relationships, to think about how other people's goals relate to their own, and to carry out negotiation and repair around these internal states. Bowlby (1971) used the term goalcorrected partnership to refer to the parentchild relationship in the toddler period. We think that this use of the term is too narrow. Throughout life we are engaged in the difficulties of finding out what other people want or need, negotiating in relationships so that we understand someone else's goals and can find a way of attuning ourselves to these. Indeed this type of social problem solving accounts for the majority of conversation that we have as humans. As Dunbar (1996) showed, men and women spend most of their time during conversations talking about what goes on between people, trying to figure out goals and intentions. What did Jackie mean when she spoke sharply to Peter? Why has the director been so preoccupied? These are the topics that form the bases of conversation. There are several key components necessary to form a goalcorrected partnership. One element is a goal orientation towards the other in which the will to cooperate with someone else is paramount. Affiliation and cooperation are valued over other goals such as dominance or retribution. Being able to develop an accurate model of theother is also a key component of being able to form a goalcorrected partnership. The mental states that are ascribed to the other have to be accurate. If I think that your intention as you walk toward me is to hurt me, when in fact it is to help me, my model of your intentions toward me is a long way from what you actually intend. A third component, and one that is uniquely available to us as humans, is the capacity to negotiate, alter, and repair these models of self and other through language. Even if I misinterpret your intentions, I can talk to you about my model and correct inaccuracies. In order to do this, I need a language for internal states. I need to be able to find words for what I am
Page 273
thinking and feeling and also to talk about what you are thinking and feeling (Salovey & Mayer, 1990). Goals and Emotions The goal orientation toward the other and the self's perception of the other's goals can be understood as cognitive/emotional representations of the self and other (Baldwin, 1992; Stern, 1994). Our proposal is that there is a limited number of goalbased cognitive/emotional representations of the self and other that correspond to the basic emotions. The emotion system developed as a system that monitors the status of our goals during our transactions with the environment (Oatley & Johnson Laird, 1987). Emotions function to facilitate this transaction, signaling to the self and other the nature of the individual's goals. Happiness is an emotion of plans proceeding well. It signals cooperation and affiliation toward the other (Lazarus, 1991). Anger is an emotion that signals an adversarial stance. Goals have been thwarted, but others have done this intentionally to block the self. Anger signals a fight against this blocking, with an attempt to dominate or force the other to alleviate the goal block (Averill, 1982; Lazarus, 1991). Emotions are responses to immediate environmental stimuli, but they crucially depend on the goal orientation that we bring to the situation (Lazarus, 1991). When someone pulls out in front of me in a car and blocks my exit, I can appraise this as a temporary inconvenience without consequences to my primary goals. Or I can appraise this as an insult needing immediate retribution. In the first case, I feel no emotion in response to the event. In the second case, I become enraged and want to retaliate against the other driver. There are longterm goal structures that we carry with us into encounters, determining whether we pay attention to something, how we interpret it, what we feel, our actions in response to it, and how we remember the encounter later. The emotional qualities of such encounters are determined by how we perceive the goals of others and what our goals are toward other people. For most children and adults, goalbased representations of the self and other, and the cognitive tools for negotiating internal states, work well enough so that joint planning and cooperation is frequently possible. For a small number of children, however, such cooperative endeavors are difficult to achieve. Children with externalizing disorders have longterm difficulties in their relationships (Caspi, Elder, & Bem, 1987; Huesmann, Eron, Lefkowitz, & Walder, 1984). One way of thinking about their disturbance is that the capacity to enter into a cooperative relationship has become impaired. Instead of viewing others as cooperative and helpful, they view these others as ill intentioned. Instead of feeling that they want
Page 274
to construct joint plans, such children feel that they want to get the better of other people, pay them back for insults against them. Instead of believing that goals can be negotiated, they try to force the other to facilitate their own goals. Our first aim in this chapter is to examine what this bias in goal structure looks like for children who show externalizing behaviors. Externalizing Psychopathology Children who score high on externalizing psychopathology are children who are angry and defiant much of the time. They engage in activities such as lying, stealing, bullying, and being oppositional and aggressive. They show a general disregard for the rights of others. Recently researchers have been interested in the relation between emotions and psychopathology. We agree with Lazarus (1991), Malatesta and Wilson (1988), Tomkins (1979) and others in thinking that in psychopathology one emotion becomes predominant in the emotional experience of the individual. In one study of the relation between emotions and psychopathology, we found that externalizing children show high levels of anger, and low levels of sadness and happiness (Jenkins, 1999b). We interpret this pattern of emotional organization as indicative of an underlying goal structure. These children value the adversarial stance: dominance in relationship, showing who is in control, retaliating against perceived slights. Their low levels of happiness suggests that affiliation and cooperation are less important to them. There are longterm continuities in levels of angry affect. Caspi and colleagues (Caspi et al., 1987) followed up angry children into their adult life. Not only were they angry adults, they had created environments for themselves that were more negative and likely to elicit anger. We think that these longterm continuities in affective behavior represent stable goal structures. What is important to a child with regard to their relationships remains important to them as they develop into adults. If in childhood they have appraised other people's intentions toward them as hostile, in adulthood such an appraisal structure will also be more likely. If in childhood it was important for them to retaliate against slights to make it clear that they were dominant, in adulthood this is also likely to be an important goal influencing their interactions with others. Many people have talked about models or schemas of the other that affect continuities in behavior (see Baldwin, 1992, for review). The unique aspect of the work presented here is that these models are situated within the framework of a cognitive theory of emotion (Lazarus, 1991; Oatley & JohnsonLaird, 1987), in which it is postulated that there is a finite number of basic emotions that represent core goal orientations for humans. For an analysis of externalizing behavior, we are principally concerned with the balance between anger and happiness and thus the goal dimensions
Page 275
of (a) cooperation, affiliation as indicated by happiness, and (b) power, control, dominance as indicated by anger. In this chapter we present data from a longitudinal study of children's goals, emotions, and psychopathology. We test the hypothesis that the goals of the self in relation to the other are biased. Our hypothesis is that it is more important to be dominant or powerful in the relationship than to be affiliative. We think of these children as being children who become angry much more readily than other children because they have a bias in their goalbased cognitive/emotional representations of the self and other. Here we present evidence only on the goals of the self toward other people. We do not present evidence on children's biases in how they perceive the intentions of others, as there is already a substantial literature in this area. Children with externalizing disorders have been found to have a hostile attribution bias. They are much more likely to interpret neutral events as if someone has been intentionally hostile towards them (Dodge & Coie, 1987). Because externalizing children think that the other is out to wound them, they have to fight back and retaliate. We also predicted that children with externalizing psychopathology were less skilled at the language of mind. Internal state language deficits have been found to be more common in children who have experienced aversive and stressful environments. In a study in which abused children were compared with nonabused children matched on socioeconomic status, Beeghly and Cicchetti (1994) found that abused children showed significantly less use of internal state language than the nonabused group. Similarly, Jenkins and Greenbaum (1996) found that children who lived in homes with high levels of parental conflict used less internal state language. We also found that children with psychopathology used fewer emotion words than children without psychopathology. In this study we wanted to follow up these associations. We hypothesized that children with higher levels of externalizing behavior would use less internal state language overall, such that shared goals would be harder to achieve and that repairs between people, once problems emerged, would be harder to execute. Finally, we thought that, as compared with normal children, children with externalizing psychopathology would talk relatively more about their own internal states than about those of others. If we are correct in our prior hypotheses, it is a very challenging task for such children to extend themselves beyond the perspective of the self. Methods and Sample These data were collected from children who were attending two classes in a school serving a middleclass population in Toronto. They represent a preliminary analysis of a longitudinal data set. Children were 8 to 9 years
Page 276
old at time 1, and 9 to 10 years old at time 2. By time 2, three children had left the school and were not included in the followup sample. Stage 1 In order to determine how children talk about others' intentions, we videotaped groups of four boys, or four girls, planning a play together. Children were then interviewed individually to talk about their own goals during the play and the goals of other children. Thirty children took part in this stage of the study. At three different points during the tape, when a child showed a positive emotion, a negative emotion, and a neutral expression, children were asked what was [child's name] thinking about the other kids at that point? What was [child's name] wanting the other kids to do at that point? From children's answers to these questions we drew up a list of goals: being dominant and in control of the interactions, hurting others, wanting to listen to what others are saying, being close to others. Stage 2 We then asked children to rate all other children in the class as well as themselves on the goals that were generated. Two teachers per class (one teacher and one trainee teacher) also rated the children in their class on these goals. Two classes took part which resulted in ratings for 43 children. Goal Ratings Dominance Children were asked to rate goal statements on a 5point scale. Dominance items were "How important is it to [child's name] to get other people to take up her/his ideas?" and "How important is it to [child's name] to make others feel sad?" These items were strongly correlated within teacher report and consequently these were combined to form a dominance scale. As they were also correlated for peer report they were combined to form a peer report of dominance scale. These items were not, however, correlated in children's selfreport. There was almost no variance for children's responses to "How important is it to make others feel sad," and this was dropped from further analysis. Consequently for selfreport the only item to make up the dominance measure was ''How important is it to get other people to take up your ideas?" Each child was rated by approximately 20 children. We calculated the mean score for each child on each goal scale, and these scores became the peers' ratings of children's goals. The child's class teacher also rated
Page 277
each child on the goal statements. The child rated him or herself on the same scale and this is referred to as the child's rating of their own goals. Affiliation The only affiliation item used at time 1 was "How important is it to [child's name] to find out things from other people, or to learn things from other people by listening?" This was a behavior that many children talked about during the interviews to describe cooperative behavior. It represented a child's curiosity about another, the desire to know what they thought. We added another affiliation question halfway through the time 1 interviews. Twenty children were asked to rate "How important is it for [child's name] to get to know others really well and get close to them?" Because we did not have scores for every child on this item, we did not use it to form a composite scale. We did, however, find that these two items were strongly related to one another (peer report: r (20) = .68, p < .001; selfreport: r (20) = .74, p < .001). Ratings of Externalizing Psychopathology Teachers completed ratings of children's emotions and behavior in the classroom using the Teacher Report Form (Achenbach, 1991b). This instrument yields an externalizing factor that involves items about the child stealing, lying, cheating, being disruptive at school, being aggressive, cruel to animals, and so on. Tscores ranged from 39 to 63. Four children in this sample were in the borderline clinical range of externalizing problems. A different teacher completed psychopathology ratings from the one who completed goal ratings in order to ensure that any significant associations between goals and psychopathology were not the result of within informant report. Ratings of Anger Frequency Teachers, peers and children themselves were asked to rate how frequently the child became angry on a 5point scale, with higher ratings indicating a higher frequency of anger expression. Stage 3 Use of Language in Talking About Intentions The individual interviews with children when talking about their own and others' goals were transcribed and coded. Thirty children were involved in this part of the study. We coded three different categories of internal state talk: cognition terms (know, think, remember), desire terms (want, need) and emotion terms (happy, sad, crazy). We also counted the total number of words used by the child. Onetailed significance tests are quoted throughout the results section.
Page 278
Results Stability of Relational Goals Over a 1Year Period Dominance goals were found to be highly stable over a 1year period, as can be seen by the correlations between time 1 and time 2 goals presented in Table 14.1. Correlations were higher for teachers and peer reports. It should be noted that children changed teachers between time 1 and time 2 so that the correlations given for teacher data represent different teachers' reports 1 year apart. Correlations for children's affiliation goals were less high, but still significant for the children's selfreport. The levels of stability particularly for the dominance goals and for children's anger frequency are much higher than the stability obtained for externalizing psychopathology. Externalizing behavior is considered to be highly stable over years, leading Olweus (1979) to argue that it is as stable as IQ. It is impressive that frequency of anger and dominance goals show a comparable or, in the case of peer and teacher ratings, a higher level of stability than those seen in this study for externalizing behavior. It may be that relational goals and an affective organization in which anger predominates are stable within the person and that behaviors indicative of externalizing psychopathology are more variable depending on environmental constraints and demands. The finding that affiliation goals are less stable may either reflect that such goals are harder to measure or that these goals do indeed change over time. Agreement Across Different Informants In Table 14.2, agreement across informants on relational goals is shown. It is possible to see from this that peers and teachers agree remarkably well on children's relational goals for dominance. The agreement across informants on affiliation goals was significant for 4 out of 6 correlations, TABLE 14.1 Correlations Examining the Stability of Relational Goals, Anger Frequency and Teacher's Ratings of Externalizing Psychopathology Over a 1Year Interval
Peer
Teacher
Child
Dominance
.79***
.67**
.34*
Affiliation
.21a
.25a
.37**
Anger frequency
.75***
.36*
.30*
Externalizing psychopathology ap < .10. *p < .05. **p < .01. ***p < .001.
.35*
Page 279 TABLE 14.2 Agreement Across Informants on Children's Dominance and Affiliation Goals
Peer Teacher (Time 1)
Peer Child (Time 1)
Teacher Child (Time 1)
Peer Teacher (Time 2)
Peer Child (Time 2)
Teacher Child (Time 2)
Dominance
.61***
.22a
.55***
.72***
.49**
.45**
Affiliation
.30*
.37*
.30*
.37*
.48**
.28*
a
p < .10. *p < .05. **p < .01. ***p < .001.
but it was generally weaker than the agreement seen for dominance goals. Again these levels of crossinformant agreement are comparable to those reported for the betterestablished construct of externalizing behavior (Achenbach, 1991a). The extent of this agreement across informants supports the validity of these measures for assessing children's relational goals. Even though peer and teacher goal ratings depend on inferences about internal states rather than on directly observed behavior, agreement across informants is impressive. It is also important that ratings made by the self show agreement with how others interpret goal states. Relation Between Dominance and Affiliation Goals and Children's Externalizing Problems Correlations between children's relational goals (as rated by peers, teachers, and children themselves) and externalizing symptoms are presented in Table 14.3. It is possible to see that across all accounts high dominance goals and low affiliation goals are associated with high levels of externalizing psychopathology at time 1. It is also possible to see that time 1 dominance goals are associated with time 2 teacher ratings of children's externalizing TABLE 14.3 Correlations Examining the Relation Between Relational Goals at Time 1 and Externalizing Psychopathology at Time 2
Time 1
Affiliation Goals
Peer
Teacher
Self
Peer
Teacher
Self
Externalizing psychopathology
.67***
.44**
.40** .15 .54***
.10
Anger expression teacher rating
.62***
.64***
.34*
.10
Time 2
*p < .05. **p < .01. ***p < .001.
Dominance Goals
.35*
.20
Page 280
psychopathology. Further affiliation goals as rated by peers at time 1 are associated with time 2 externalizing, although children's own ratings and teacher ratings of time 1 affiliation goals are not associated with children's externalizing. Thus children who show higher levels of dominance and lower levels of affiliation at time 1 show higher levels of externalizing behavior 1 year later. A more stringent test of the hypothesis that relational goals underlie externalizing psychopathology would be to examine whether goals at time 1 predict a change in externalizing behavior from time 1 to time 2. Are children who are rated as more dominant and less affiliative at time 1 likely to develop more difficulties with externalizing behavior over time? In order to answer this question, we carried out three hierarchical regression analyses (for each informant's report of goals). Children's externalizing behavior at time 2 was our criterion variable. By entering time 1 externalizing scores into the regression equation first, any subsequent predictors of time 2 externalizing were explaining variance in the change in externalizing scores between times 1 and 2. Affiliation goals were entered second into the equation, followed on the next step by dominance goals. In this way we were able to determine whether dominance goals contributed uniquely to the prediction of externalizing behavior, after variance attributable to affiliation goals had been taken into account. Results of these analyses can be found in Table 14.4. The first line shows the results of the first step for the three regressions: regressing time 2 externalizing on time 1 externalizing. As step 1 was the same for all analyses, this is not repeated in Table 14.4 for each separate regression. It is possible to see that for peer accounts of goals, affiliation goals predicted 17% and dominance goals predicted a further 16% of the variance in the change in externalizing from time 1 to time 2. Using teacher ratings, affiliation goals were not found to predict externalizing behavior but dominance goals predicted 11% of the variance in the change in externalizing from time 1 to time 2. Using selfratings, affiliation goals were not found to predict externalizing behavior but dominance goals predicted 8% of the variance in the change in externalizing from time 1 to time 2. We also tested the opposite causal model in which externalizing psychopathology predicted changes in goal ratings between time 1 and time 2. This is a test of the hypothesis that the presence of psychopathology affects a change in a person's relational goals. We found no evidence to support this direction of effect. Externalizing was never a significant or near significant predictor of time 2 goals after taking account of time 1 goals and using dominance or affiliation as criterion variables. Results of this series of analyses do suggest that children with externalizing psychopathology show a bias in their relational goals. High externalizers have relational goals in which being dominant in interaction is highly
Page 281 TABLE 14.4 Summary of Hierarchical Regression Analyses Examining the Role of Affiliation and Dominance Goals in Predicting Externalizing Behavior at Time 2 after Accounting for Time 1 Externalizing Behavior
B
SEB
b
Change in R2
Step 1
Externalizing behavior—Time 1
.34
.15
.35
.12
*
Step 2
Affiliation goals—peer report
9.73
3.24
.51
.17
**
Step 3
Dominance goals—peer report
12.19
3.75
.57
.16
**
Step 2
Affiliation goals—teacher report
.17
1.26
.02
.00
Step 3
Dominance goals—teacher report
4.5
1.95
.41
.11
Step 2
Affiliation goals—selfreport
.51
1.16
.07
.00
Step 3
Dominance goals—selfreport
3.80
1.99
.31
.08
a
p < .10. *p < .05. **p < .01.
*
*
Page 282
valued and being affiliative and cooperative is less important. They value pushing their ideas forward and making other children feel small. It is less important for them to pay attention to others and listen to their ideas as a means of building a cooperative partnership. Furthermore, it is not only dominance goals that predict development of more externalizing behaviors over time. Low affiliation goals also contribute to the development of externalizing problems. Returning to the concept of the goalcorrected partnership, one of the reasons why it may be difficult to construct this kind of partnership is that children develop a biased stance toward the other in which goals of dominance overshadow goals of cooperation. The intentions of the other are also appraised negatively. This underlying goal structure influences all immediate encounters with the environment. This is the underlying goal structure associated with higher levels of anger and aggression. Language of Mind We examined correlations between children's internal state talk and teacherrated externalizing psychopathology. Higher levels of externalizing psychopathology were almost significantly related to lower use of emotion terms, r (30) = .30, p < .06, and significantly related to lower use of desire terms, r (30) = .45, p < .01. Externalizing psychopathology was not associated with lower use of cognition terms. After we controlled for word count, children who used fewer desire terms showed significantly higher levels of externalizing behavior. That is to say, children with externalizing psychopathology discussed their own and other people's goals less frequently than other children. Finally, we were interested in whether children with psychopathology talked less about what was in others' minds than about what was in their own. If children are generally less skilled at cooperative endeavors we might expect them to be less able to enter into the mind of the other. We assessed this by partitioning the mental state talk into that which was about the self and that which was about the other. Then we calculated the proportions: mental state talk about the other/total amount of mental state talk. We found an association between externalizing psychopathology and talk of other minds. The higher the externalizing behavior the less the talk about other minds, r (29) = .40, p < .05. Here we find another aspect of functioning that may negatively affect the development of the goalcorrected partnership. Children with externalizing psychopathology are less skilled in using mental state talk, particularly mental state talk that is about goals (desire terms). They also spend less time talking about others' goals than about their own. We do not know whether these patterns of mental state discourse are primarily motivational or whether they represent deficits in language use. It is certainly the case that
Page 283
children with externalizing behavioral problems are more likely than nondisordered children to show cognitive impairments in development (see Moffit, 1993), which may have an impact on the way in which mental state talk develops. It may also be that cognitive vulnerabilities are exacerbated by contextual risks (such as poor family relationships) that make it less likely that these children will reach out to others, and feel motivated to develop the tools that will allow them to develop the goal corrected partnership. In summary, we found that children's relational goals were stable over time and that high dominance and low affiliation goals predicted externalizing psychopathology. We also found that children with externalizing psychopathology showed lower levels of mental state talk, particularly in the use of desire terms. How Do Difficulties in the Development of the GoalCorrected Partnership Develop? What leads some children to value dominance, devalue affiliation, and use less mental state talk, particularly about the goals of others? In this study, we did not examine the relation between goal orientations, or mental state talk and environmental risks. We can, however, speculate on the origins of such difficulties by drawing on research related to environmental factors that increase the likelihood that children will develop externalizing psychopathology. Patterns of aggressive interaction in children start up when children are very young. Patterson and colleagues (Patterson, Dishion, & Reid, 1993) have found that coercive interactions between parents and children become established in the preschool years and predict increases in aggressive behavior over time. These interactions involve escalating acts of aggression between parents and their children in which parents are coercing children to get children to meet parental goals and children are coercing parents to get their own goals met. Out of such a pattern of interaction it is not hard to imagine that children develop a goal structure in which they think it unlikely that their own goals will be met through cooperation. They come to value dominance more than affiliation. They also perceive others' intentions towards the self as hostile. Another contextual factor that we know to be a major risk factor in the development of externalizing disorder is children's exposure to angerbased conflict between their parents. Children exposed to high levels of their parents shouting at one another and becoming violent are more likely to show anger during peer interaction and higher levels of externalizing behavior more generally (Jenkins, 1999a; Jenkins & Smith, 1991). Angerbased marital conflict is characterized by parents blaming one another, power struggles in which one tries to convince the other of their position, winlose outcomes in which one person is seen as more dominant than the
Page 284
other, blocking of another's goals by being overtly and covertly resistant (Brody, Arias, & Fincham, 1996; Christensen & Heavey, 1990; Cummings & Davies, 1994; Gottman, 1993; Margolin, John, & O'Brien, 1989). Watching parents interact frequently in this way may increase the chances that children think that relationships are fundamentally adversarial. Neuropsychological deficits are also likely to play a part in the development of these goalbased patterns (Moffitt, 1993). Children with lower IQs or those with specific cognitive impairments are more likely to develop externalizing disorders than those who do not show these early patterns of vulnerability. It is also interesting to note, however, that even children who show early developmental difficulties can show cognitive gains and no behavioral problems if they live in supportive and stimulating environments (Werner & Smith, 1982). Those in more stressful environments (e.g., more negative parentchild relationships, divorce, poverty) show an exacerbation of their cognitive difficulties over time. For children who are more vulnerable in their cognitive development, and who live in coercive environments, the task involved in accurately representing their own and other people's goals and learning the language of internal state may prove particularly challenging. The same analysis of temperament is possible. Children come into the world showing individual differences in how strongly they react to goal blocks and how long they can tolerate frustration of goals (Emde et al., 1992; Goldsmith, 1993). Some environments will exacerbate such tendencies to anger whereas other environments may be conducive to the modification of such goal orientations. We want to quote from a recent novel about a 10yearold boy written by Roddy Doyle (1993).1 Interactions between this child and his father, between the two parents and between the child and his sibling illustrate both the existence of these relational goal structures and the kinds of interactions that may contribute to their development within a family. Paddy Clarke Ha Ha Ha takes place in Ireland. The novel is written in the first person from the point of view of the boy, Paddy. Over the course of the novel the parents' marriage is breaking down. A lot of angerbased marital conflict is portrayed from the child's point of view. Paddy describes a coercive pattern of interaction between his parents that is characteristic of their relationship: a dynamic in which the father refuses to respond to the mother. The father controls the interaction through his silence and derives satisfaction from his wife's powerlessness. In this first excerpt, Paddy, the child, is narrating. He is talking about how his father (also called Paddy) is refusing to speak to his mother. Then he reports dialogue between his mother and his father with his mother starting the discussion. Paddy and his brother observe the interaction. 1
We would like to thank Secker and Warburg and Roddy Doyle for permission to quote from Paddy Clarke Ha Ha Ha.
Page 285 He waited before he answered, always he did, pretended he hadn't heard anything. She was always the one that tried to get them to talk. He'd answer just when I thought she'd have to ask again, to change her voice, make it sound angry. It was agony waiting for him. Paddy? —What? —Did you hear me? —Hear what? —You heard me. —Heard what? She stopped. We were listening; she saw us. He thought he'd won; I thought he did. (p. 222)
This embittered struggle between husband and wife, structured around dominance and submission, is one source for Paddy's learning about goals within intimate relationships. This interactional pattern is seen again in Paddy's interaction with his father. In this relationship he also feels that his father is powerful and dominant, putting him in the position of being powerless and submissive. In this excerpt Paddy has run home to tell his father about a medal that he won at school. He is excited and wanting praise from his father. Before his father will give him this he makes him wait, as a means of demonstrating his power. The interaction between father and son is not hostile or nasty, but the dynamics around power and submissiveness are the organizers of the interaction. I won a medal. I came second in the hundred yards except it wasn't nearly a hundred yards; it wasn't even fifty. It was a Saturday, the school sports, the first one the school had ever had . . . I walked home; I didn't want to run. I went to my da first. —Get out; not now. He didn't look up. He was reading the paper. He always talked about Backbencher on Saturdays, telling my ma what Backbencher'd said, so he might have been reading Backbencher. He clicked the paper, straightened it up. He wasn't angry or anything. I felt thick. I should have gone to my ma first; it would have been easier then, what had happened. I went to the door; the bones in my legs were rubbery. He was in the drawing room. Peace and quiet, that was what he got in there, the only place in the house. I didn't mind waiting, not really, but he hadn't even looked up. I was going to shut the door quietly. He looked. —Patrick? —Sorry. —No; come in. The paper fell forward and folded over; he let it. I let go of the handle. It needed oiling. I came back in. I was scared and
Page 286 pleased, bits of each. I wanted to go to the toilet; I thought I did, that kind of feeling. I asked him something. —Are you reading Backbencher? He smiled. —What have you got there? —A medal. —Show us; you should have told me. You won. —Second. —Nearly first. —Yeah. —Good man. —I thought I won. —Next time. Second's good though. Put it there. He held out his hand. I wished he'd done it the first time. It wasn't fair the way he made you nearly cry before he changed and did what you wanted him to. It didn't always happen that way but it happened enough for him to have parts of the rooms to himself, for the house to be different at the weekends. I could never run to him; I had to check first. I blamed the paper. (p. 205)
The piece begins with the boy's excitement about telling his father some good news and this sinking feeling as his father is not responsive to him. His father lets him know that the relationship is on his terms and once the young Paddy has understood this, his dad is able to respond to him. His father's goal within that moment of interaction with his son is the same as that in the moment of interaction with his wife: to show who is boss. Dominance is the primary goal, and cooperation is relegated to a secondary goal, if it is present at all. Also interesting to note is that the author depicts the 10yearold child as able to understand these goals explicitly and to talk about them, an observation that found support from our empirical work. The final excerpt describes an encounter between Paddy and his brother. Here again the theme is dominance and submission. There is affection between Paddy and his brother, Francis, but Paddy's concern for his brother easily flips over into Paddy needing to show who is boss and force Francis to his will. In this excerpt, in contrast to the other excerpts, the winlose dynamic is expressed physically. Paddy is trying to get his brother to eat a cookie. —Francis. D'you want this? It was a biscuit, only a biscuit. I wanted it as well but I wanted him to take it. I was giving it to him. He wouldn't even look at it. I grabbed him.
Page 287 —Open your mouth! His lips vanished as he closed down his mouth. He got ready to be pulled around, stiff and dead. —Open your mouth! I held it in front of his eyes. —See. He shut them, crammed shut. I got the biscuit and I got his head and I pushed the biscuit at his mouth, and I pushed until it fell apart and I couldn't hold it. (p. 239, Reprinted with permission)
He continues to battle with Francis, resorting to holding his nose so that Francis will have to open his mouth. Francis continues to refuse and Paddy becomes more frustrated. —I won't hit you again, okay, ever. I didn't expect an answer or anything. I waited a bit. Then I kicked him. And I thumped him. Twice. Then I felt my back go freezing: someone was looking. (p. 242)
What we see here is Paddy's compulsion to have his brother submit to his will. Although he has affection for him and much of their interaction is affiliative and cooperative, it can easily tip, as here, into Paddy being dominant and his brother having to submit to his will. We see similar goal dynamics in three relationships. It is likely that a synergy is operating here. Paddy observes an interaction pattern between his parents that he also experiences directly in his relationship with his father. He plays out the same pattern in his relationship with his brother. As he observes and experiences himself in relationships, he is building up a cognitive structure for what relationships are about. Costanzo and Dix (1983) describe a social perception process in which children construct meaning by watching and experiencing interaction: ''Interpretations of persons and events are acquired through prior contentspecific social experience and applied as rules or values believed to govern the types of conduct being observed" (p. 64). As these repetitions in a goal orientation occur across different relationships, perhaps they gain strength and become more compelling and pervasive, resulting in a relational schema in which the child thinks that the main dynamic governing intimate relationships is a power dynamic. Rutter (1979) demonstrated that when several environmental risk factors occur together (such as the parents being highly conflictual with one another and a parent being coercive and hostile towards a child), the risk of externalizing disorder is greater than just adding the two risks together.
Page 288
Although we have spent this chapter talking about how continuities in people's goals in relationships exist and explain changes in behavior over time, we know that even children showing extreme externalizing problems develop alternative patterns of intention and action (Graham & Rutter, 1973; Robins, 1978). One factor that is probably central to explaining discontinuities in development is the formation of new supportive relationships. Quinton and Rutter (Quinton & Rutter, 1988) have shown that women who had severely deprived early childhoods and were consequently at much greater risk for mental health problems and parenting difficulties in their adult life were able to avoid such outcomes by finding a supportive partner. Similarly, children in highstress circumstances are less likely to show disturbance if they have a good relationship with a sibling or a grandparent (Jenkins & Smith, 1990). What is it that a supportive partner provides? Although person X may have continuities in her own goals toward others and her perception of Y's goals towards her, Y brings his own goal structure to the relationship. His own goals in relationship may be highly cooperative so that her attempts to engage him in conflict are nonproductive. He may bring a particularly benign interpretation to her attempts to have power in the relationship, such that he does not perceive her actions as hostile. As she assimilates his treatment of her to her existing schemas, her own goals may take on a different balance. Change in psychotherapy may involve a similar mechanism. As the psychotherapist fails to fit in with the goal structure, the structure shifts a little bit, allowing for new interpretations and the development of new goal structures. Most psychotherapies involve talk about relationships and goals in relationships. Perhaps through such talk new metacognitive structures emerge that facilitate successful cooperation in the goalcorrected partnership. A supportive sibling for an externalizing child or a spouse for an adult may allow a new belief structure to emerge in which cooperation and joint planning become sufficiently rewarding to substitute for previous goals. Conclusions Some children do not develop an intentional stance toward others that facilitates cooperative action and joint planning. Their own goals toward others and their perceptions of others' goals toward themselves are biased. They show an adversarial goal structure that leads to frequent experiences of anger and aggression. The use of internal state language to negotiate difficulties in the goalcorrected partnership is also limited and makes it more difficult for children to make and sustain joint plans that satisfy the goals of both agents in a cooperative interaction. Such difficulties develop through exposure to adverse environmental circumstances such as hostile
Page 289
parentchild relationships and marital conflict. Although this cognitive structure for the interpretation of action may have been accurate when seeking to understand events in their families, such interpretations may be quite inaccurate in more general contexts. We have also suggested that through an examination of children who show intentional structures that are less optimal, we gain some insight into the function of this representational system more generally. Children with externalizing disorders are oppositional. They are acting against other people. Their representations of the self and other discourage cooperative action. Optimal development consists of having a representational system that facilitates and enables cooperative action and planning. Such cognitive structures for representing our own and others' intentions and goals have developed to enable complex cooperation across multiple agents. These enable us to conceive of, and execute, plans for the achievement of goals that would not have been available to individuals acting on their own. References Achenbach, T. M. (1991a). Integrative guide for the 1991 CBCL/418, YSR, and TRF profiles. Burlington, VT: Department of Psychiatry, University of Vermont. Achenbach, T. M. (1991b). Manual for the Teacher's Report Form. Burlington, VT.: Department of Psychiatry, University of Vermont. Astington, J. W., Harris, P. L., & Olson, D. R. (1988). Developing theories of mind. New York: Cambridge University Press. Averill, J. R. (1982). Anger and aggression. An essay on emotion. New York: Springer. Baldwin, M. (1992). Relational schemas and the processing of social information. Psychological Bulletin, 112, 461484. Beeghly, M., & Cicchetti, D. (1994). Child maltreatment, attachment and the self system: Emergence of an internal state lexicon in toddlers at high social risk. Development and Psychopathology, 6(1), 530. Bowlby, . (1971). Attachment and loss, Volume 1. Attachment. London: Hogarth Press (Rep. by Penguin, 1978). Brody, G. H., Arias, I., & Fincham, F. D. (1996). Linking marital and child attributions to family processes and parentchild relationships. Journal of Family Psychology, 10, 408421. Case, R. (1991). The mind's staircase. Hillsdale, NJ: Lawrence Erlbaum Associates. Caspi, A., Elder, G. H., & Bem, D.J. (1987). Moving against the world: Life course patterns of explosive children. Developmental Psychology, 23, 308313. Christensen, A., & Heavey, C. L. (1990). Gender and social structure in the demand/withdraw pattern of marital conflict. Journal of Personality and Social Psychology, 59, 7381. Costanzo, P. R., & Dix, T. H. (1983). Beyond the information processed: Socialization in the development of attributional processes. In G. T. Higgins, D. N. Ruble, & W. W. Hartup (Eds.), Social cognition and social development: A socicultural perspective (pp. 6381). New York: Cambridge University Press. Cummings, E. M., & Davies, P. T. (1994). Children and marital conflict: The impact of family dispute and resolution. New York: Guilford Press.
Page 290
Dodge, K. A., & Coie, J. D. (1987). Socialinformationprocessing factors in reactive and proactive aggression in children's peer groups. Journal of Personality and Social Psychology, 53, 11461158. Doyle, R. (1993). Paddy Clarke Ha Ha Ha. London: Secker and Warburg. Dunbar, R. (1996). Grooming, gossip and evolution of language. London: Faber and Faber. Dunn, J. (1988). The beginnings of social understanding Cambridge, MA: Harvard University Press. Emde, R. N., Plomin, R., Robinson, J., Corley, R., DeFries, J., Fulker, D. W., Reznick, J. S., Campos, J., Kagan, J., & ZahnWaxler, C. (1992). Temperament, emotion and cognition at fourteen months: the MacArthur Longitudinal Twin Study. Child Development, 63, 14371455. Goldsmith, H. H. (1993). Temperament: Variability in developing emotion systems. In M. Lewis & J. M. Haviland (Eds.), Handbook of emotions (pp. 353364). New York: Guilford. Gottman, J. M. (1993). The roles of conflict engagment, escalation, and avoidance in marital interaction: A longitudinal view of five types of couples. Journal of Consulting and Clinical Psychology, 61, 615. Graham, P. J., & Rutter, M. (1973). Psychiatric disorder in the young adolescent: A followup study. Proceedings of the Royal Society of Medicine, 66, 1226 1229. Huesmann, L. R., Eron, L. D., Lefkowitz, M., & Walder, L. O. (1984). Stability of aggression over time and generations. Developmental Psychology, 20, 1120 1134. Hutchins, E. (1995). Cognition in the wild. Cambridge, MA: MIT Press. Jenkins, J. M. (1999a). Interparental conflict and children's emotions: the development of an anger organization. Manuscript submitted for publication. Jenkins, J. M. (1999b). Psychopathology and shortterm emotion: the balance of affects. Manuscript submitted for publication. Jenkins, J. M., & Greenbaum, R. (1996). Metacognition about emotion and child psychopathology. In N. Frijda (Ed.), Proceedings of the IXth Conference on the International Society for Research on Emotion. Toronto, Canada: International Society for Research on Emotion. Jenkins, J. M., & Smith, M. A. (1990). Factors protecting children living in disharmonious homes: Maternal reports. Journal of the American Academy of Child and Adolescent Psychiatry, 29, 6069. Jenkins, J. M., & Smith, M. A. (1991). Marital disharmony and children's behaviour problems: Aspects of a poor marriage which affect children adversely. Journal of (Child Psychology and Psychiatry, 32, 793810. Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford University Press. Malatesta, C. Z., & Wilson, A. (1988). Emotion/cognition interaction in personality development: A discrete emotions, functionalist analysis. British Journal of Social Psychology, 27, 91112. Margolin, G., John, R. S., & O'Brien, M. (1989). Home observations of married couples reenacting naturalistic conflicts. Behavioral Assessment, 11, 101118. Moffitt, T. E. (1993). Adolescent limited and life course persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674701. Oatley, K. (1992). Best laid schemes: Psychology of the Emotions. New York: Cambridge University Press. Oatley, K., & JohnsonLaird, P. N. (1987). Towards a cognitive theory of emotions. Cognition and Emotion, 1, 2950. Olweus, D. (1979). Stability of aggressive reaction patterns in males: A review. Psychological Bulletin, 86, 852875. Patterson, G. R., Dishion, T., & Reid, J. (1993). Antisocial boys. Eugene, OR: Castalia. Quinton, D., & Rutter, M. (1988). Parenting breakdown: The making and breaking of intergenerational links. Aldershot, UK: Avebury.
Page 291
Robins, L. N. (1978). Sturdy childhood predictors of adult antisocial behavior: Replications from longitudinal studies. Psychological Medicine, 8, 611622. Rutter, M. (1979). Protective factors in children's responses to stress and disadvantage. In M. W. Kent &J. E. Rolf (Eds.), Primary prevention in psychopathology, Vol. 3: Social competence in children (pp. 4974). Hanover, NH: University Press of New England. Salovey, P., & Mayer, J. M. (1990). Emotional intelligence. Imagination, Cognition and Personality, 9, 185211. Stern, D. (1985). The interpersonal world of the infant. New York: Basic Books. Stern, D. (1994). One way to build a clinically relevant baby. Infant Mental Health Journal, 15, 925. Tomkins, S. S. (1979). Script theory: Differential magnification of affects. In H. E. Howe & R. A. Dienstbier (Eds.), Nebraska Symposium on Motivation, 1978 (pp. 201236). Lincoln: University of Nebraska Press. Werner, E. E., & Smith, R. S. (1982). Vulnerable but invincible: a longitudinal study of resilient children and youth. New York: McGrawHill.
Page 293
PART IV INTENTIONALITY AND LANGUAGE
Page 295
Chapter 15— The Language of Intention: Three Ways of Doing It Janet Wilde Astington University of Toronto The notice says, "Do not feed the penguins." I, however, feed them peanuts. Now peanuts happen to be, and these prove, fatal to these birds. Did I feed them peanuts intentionally? Beyond a doubt: I am no casual peanut shedder. But deliberately? Well, that seems perhaps to raise the question, "Had I read the notice?" Why does it? Or ''on purpose?" That seems to insinuate that I knew what fatal results would ensue. Again, why? —Austin (1966/1979, p. 275, emphasis added)
All three terms highlighted above—intentionally, deliberately, on purpose—are ways of characterizing human action as intentional. Intentional, that is, in the narrow sense of being performed deliberately or on purpose, rather than in Brentano's (1874/1960) sense of being directed at some state of affairs in the world, although intentions are intentional in this broader sense too: When I say that I intend to do suchandsuch, I represent some future action in the world. In this chapter I focus on intentionality in its narrower sense, involving the idea of action performed on purpose. The concept of intention has been much discussed in philosophy and psychology and here I will do no more than consider the language of intention, that is, the ways in which we use words (such as those emphasized in the Austin quote) to anticipate and describe our own and others' actions. In particular, I am concerned with how children's acquisition of such terms may mediate their developing understanding of intention. By "mediate" I mean that the child's concept will be determined by the categories and
Page 296
the relations among the categories lexicalized in the child's natural language (Nelson, 1996). I would not argue that children have no understanding of intention until they can talk about it. I do, however, believe that the ability to talk about it significantly influences and increases their understanding. Ordinary Language Philosophy This belief is encouraged by midcentury ordinary language philosophy, whose practitioners claimed that language itself creates categories and distinctions that do not, indeed cannot, exist for us without language. Their method aims to give descriptive analyses of concepts as revealed in ordinary language, that is, in everyday, nontechnical discourse. J. L. Austin was one of the most important members of this group and the first part of my title is chosen to acknowledge one of his little known papers, "Three Ways of Spilling Ink" (Austin, 1966/1979), from which the opening quotation is taken. The manuscript was incomplete at the time of Austin's death and was assembled by Lynd Forguson from Austin's lecture notes. There are two parts to the ordinary language philosopher's method (which is not empirical although Austin sometimes said it was). First, one thinks of examples of what would be said or what could be said in a certain instance and how it would be interpreted. Second, one examines the grammar, etymology, and dictionary definitions of terms that might be used in these instances. We can see the method in action in "Three Ways of Spilling Ink" in which Austin investigated the concept of intention: In considering responsibility, few things are considered more important than to establish whether a man intended to do A, or whether he did A intentionally. But there are at least two other familiar words important in this respect. Let us distinguish between acting intentionally and acting deliberately or on purpose, as far as this can be done by attending to what language can teach us. A schoolteacher may ask a child who has spilled the ink in class [this is England in the 1950s, where inkwells were an important part of classroom equipment]: "Did you do that intentionally?" or "Did you do that deliberately?" or "Did you do that on purpose (or purposely)?" It appears at first sight to matter little which question is asked. They appear to mean the same or at least to come down to the same in this case. But do they really? There are in fact techniques available for distinguishing between these expressions. . . We may consider, for instance, for a start: (i) imagined or actual cases . . . (Austin, 1966/1979, pp. 273274)
This is the first part of the method, which Austin considered perfectly satisfactory. Ryle (1963) defended the methods used in ordinary language
Page 297
philosophy by asking why we need to count noses if we've got one ourselves; that is, one's own intuitions provide perfectly satisfactory data (at least, if one is a philosopher). As Forguson (1985) said, The nonempirical nature of [ordinary language philosophers'] descriptive analyses was not thought by them to constitute an obstacle to understanding our concepts, because university professors are competent speakers of English or some other natural language, they are ordinary people outside the philosopher's closet, and hence can be expected to bring their ordinary understanding with them into the closet when they are going about their analytical work. Moreover, professional philosophers are reflective, sensitive to nuance and to fine distinctions of usage. What better informants could there be?
Psychologists might think differently, of course. Be that as it may, in "Three Ways of Spilling Ink" we can see the method in action, with reference to intention. Austin provided a number of imagined examples, including the penguinfeeding one given at the beginning of the chapter. He then went on to apply the second part of the method, which considers the grammar, etymology, and so on of the words used in the examples. Austin pointed out that we might say, "I am deliberating," using the continuous or progressive aspect, but intend is not used with progressive aspect, rather one says, "I intend . . ." The adjectives are also differently formed: deliberate, purposeful, intentional—each takes a different suffix. Austin likened "deliberate" to "considerate,'' saying that there is a suggestion that some operation has been performed, such as the process of deliberation (about an action, for example), or consideration (of someone's feelings, for example). On the other hand, "purposeful," Austin suggested, by analogy to "thoughtful" and "careful," indicates that something may or may not be present (purpose, thought, care); that is, there is reference to things we may or may not be doing in performing an action. Third, he said, "intentional" qualifies an action more directly, more intimately. Austin continued in similar fashion for a few more pages. One may or may not be persuaded by his arguments. What conclusion might we reach? How shall we distinguish intentional, deliberate, and on purpose, and, more important, what light will that throw on the concept of intention? First, intentional: Austin said that "As we go through life doing, as we suppose, one thing after another, I in general always have an idea . . . of what I'm up to . . ." (Austin, 1966/1979, p. 283). Terms with the intention root are connected with this notion, that is, awareness of current actions. On the other hand, when we characterize an action as deliberate we imply that there has been deliberation, that the person has stopped to think, "Shall I or shan't I?" before deciding to act. Thus, terms with the deliberate root imply thinking prior to action. And last, when something is done for a purpose the purpose is something to be achieved or effected as a result of
Page 298
what is done. Note that Austin switched in these final pages and said for a purpose, not, as before, on purpose; there may well be a difference. The paper then tailed off with no real conclusion, which was not Austin's usual style but he never completed this particular manuscript. I have quoted it at some length in order to illustrate ordinary language philosophy in action and to start thinking about the language of intention. In his paper Austin referred to the work of Anscombe, whose book has been very influential in subsequent work on intention (Anscombe, 1957). It was she who first distinguished between expressing an intention for the future and currently acting intentionally, an important distinction that has been taken up by others. For example, Searle (1983) distinguished between prior intention and intentioninaction: "All intentional actions have intentions in action but not all intentional actions have prior intentions. I can do something intentionally without having formed a prior intention to do it, and I can have a prior intention to do something and yet not act on that intention" (p. 85). For Searle an action is comprised of intention and movement; the intentioninaction causes the bodily movement. Without the intentioninaction it is just a movement and not an action. Actions can be caused by prior intentions, which is where one thinks and then acts (see Fig. 15.1, #1). However, actions need not be caused by prior intentions; one can just act, as in an unplanned intentional action (Fig. 15.1, #2). Alternatively, prior intentions need not actually bring about actions. That is, one may plan some action and then not carry it out; this is an unfulfilled prior intention (Fig. 15.1, #3). Such unfulfilled intentions, incidentally, are the ones referred to in the saying, "The path to hell is paved with good intentions"—that is, one thought of the good deed but did not perform it. Bratman (1987) made a similar distinction between intentioninaction and prior intention (although he does not use these terms) when he said: "We do things intentionally and we intend to do things. Our commonsense psychology uses the notion of intention to characterize both our actions and our mental states" (p. 111). I illustrate these distinctions with one example (see Fig. 15.2). Consider three little girls, one (A) is in the kitchen, looking for something to take outside to feed the birds, another (B) is outside, throwing crumbs down for the birds to eat, and a third (C) is walking along eating a cookie, with crumbs dropping behind her, which the birds are pecking up. A and B illustrate the distinction between intention and action, while B and C illustrate the distinction between intentional and accidental acts. One could say that both A and B intend to feed the birds, but C does not. In Searle's terms, A has a prior intention and B has intentioninaction (which may have been caused by prior intention). In Bratman's terms, intention characterizes A's mind and B's action (and maybe also B's mind).
Page 299
Fig. 15.1. Illustration of relation between prior intention and components of intentional action (from Searle, 1983).
Let us return, for a moment, to Austin's three ways of spilling ink, intentionally, deliberately, or on purpose, and attempt to provide the conclusion he did not reach. "Deliberately," suggesting forethought involved, implies prior intention. Austin was uncharacteristically vague about "on purpose," seeming to mean "with purpose" and sometimes using ''purposely" or "purposefully." The suggestion is that the action is taken for some reason; that is, prior intention is implied. "Intentionally," on the other hand, does not carry the implication of prior intention, although the action so characterized may result from a prior intention. Thus, one might conclude that all three terms characterize intentional action, but only deliberately and on purpose imply prior intention; intentionally does not necessarily do so, although it may in context. The Language of Intention In this section I consider what is involved in the language of intention more broadly. The "three ways of doing it" that my title refers to are not
Page 300
Fig. 15.2. Illustration of the distinction between intention, action, and accident.
Austin's three ways of spilling ink, which give a very specific focus on one aspect of the language of intention. My title "Three Ways of Doing It" also refers to Austin's earlier work, How To Do Things With Words. This book, first published in 1962, is based on lectures given in the 1950s (Austin, 1962/1975). It is here that Austin spelled out his ideas about speech acts (Feldman, chap. 16, this volume). Speech acts are things we do with words, such as making promises, asking questions, and giving orders, as well as the obvious one, making statements. Austin argued that we do not use language just to make statements that can be judged true or false. We also use language in speech acts where truth is not an issue; rather, the issue
Page 301
is whether the speech act is performed completely, correctly (that is, appropriately), and sincerely. Such is the language of intention. Concerning intention, what is it that we want to "do with words"? What do we want to achieve? There are at least two things: express our intentions, and describe and characterize our own actions and those of other people. Let me elaborate on this: We express prior intentions, whether or not we carry them out, and we ask other people about their intentions; that is, we get them to express their prior intentions whether or not they carry them out. Furthermore, in cases where the prior intention is not carried out, we may talk about our own and others' unfulfilled intentions. We also talk about our own and others' actions, whether something was done intentionally or accidentally. Our excuses may depend on our claim that an action was not intended. Our moral judgments involve judgments of intentionality (Piaget, 1932/1977); for example, praise or blame for other people's actions typically makes the assumption that what they did was intentional. Thus, the "language of intention" has to cover all of this. I propose that there are "three ways" we use to achieve these ends: 1. Intention is implied; I mean in this case that it is implied by a linguistic expression, not by some action. 2. Intention is made explicit; this is the lexicalization of intention when intention terms are used. 3. Intention is avowed; that is, we "give our word" and a commitment is made. I elaborate each of these ways in turn. Intention Implied We may use the simple future to express intention, as in "I will do x" or "I shall do x," although the future tense does not necessarily mark intention. Certainly it may express intention, as in "I'll do it for you," but it may not, as in, "Help, I'll drop it!" The lack of intention is here implied by the verb drop (and help). Like the use of future tense, the auxiliary verb ''be going to" may or may not express intention. "I'm going to visit my uncle" suggests an intention to do so, whereas "Help, I'm going to fall!" implies the lack of intention by use of the verb fall (and help). Modals may also express intention, as in, "I would like to do it," but this may be just an expression of a wish. Miller and JohnsonLaird (1976) argued that although there is a lexical field of intention terms (see my second "way of doing it"), intention cannot be isolated in a simple lexical field. It is expressed all over the place in different semantic fields, for example, as a component of the meaning of
Page 302
some agentive verbs—that is, verbs that describe actions for which it makes sense to ask, "Was the action intentional?" Some agentive verbs incorporate intentionality directly into their meaning. Compare: "B died"
die is nonagentive
"A killed B"
kill is agentive
"A murdered B"
murder is agentive (+ intention)
There are, however, many "neutral" agentive verbs (e.g., break, move) that allow observers to report what they see without committing themselves as to whether the action is intentional or not. Adverbs can imply intention, when they are used with these neutral verbs: for example, "he moved skillfully," where skillfully marks a manner of intentional doing. Even prepositions (e.g., by, with) can be used to express intention (Halliday, 1970). Compare: The window was broken by a ball. The window was broken with a ball. The window was broken with a hammer. Here "by a ball" suggests that the window was broken accidentally, but "with a ball" suggests it was not an accident, but rather suggests there was an agent and perhaps an intention (compare "with a hammer"). Even if you are not convinced by this example (and some people are not), my point is that intention is deeply embedded in our language. This is before we ever get to specific lexical items expressing intention, which I consider now. Intention Explicit; The Lexicalization of Intention Some verbs express intention explicitly: for example, intend to . . . , plan to . . ., mean to . . ., try to. . . . In the present tense, the first three of these mark a prior intention, "I intend/plan/mean to do it." In contrast, try marks intentional action, not prior intention: "I'm trying to do it." Try can be used in future tense to mark prior intention: "I will try to do it" (notice this is weaker than "I intend to do it"). In the past tense, these verbs may indicate that an action was performed intentionally; that is, there was an intentioninaction, which may have been with or without prior intention. ''I meant to do it" marks such intentional action. However, the verbs may also indicate an unfulfilled, failed intention: "I meant to do it, but I forgot," or "I tried to do it, but I couldn't." Only context allows us to disambiguate "I meant to do it" as implying intentional action or unfulfilled intention. Lastly, these verbs in past tense are frequently negated to mark a lack of intention, as in, "I didn't mean to do it."
Page 303
Adverbs formed from these verbs, and other adverbial phrases, can mark intentional action: intentionally, on purpose. Adverbs and adverbial phrases can also mark unintentional action: accidentally, by accident. Some adverbial terms may be used to indicate that there was prior intention. Such are Austin's terms: deliberately, for example. Expressions of prior intention can vary in strength. Compare "I hope to do it," or "I'll try to do it," with "I intend to do it,'' or "I fully intend to do it." One would judge the action more likely to be performed if one of the second two expressions were used than one of the first two. But can someone trust that my action will be performed if I say, "I intend to do it"? Can I not go farther than this? Yes: This is the third of the "three ways." Intention Avowed; Commitment Made One might say that this is "going all the way"; that is, we "give our word" and a commitment is made. Saying "I intend to do it" doesn't commit me to doing it, but saying "I promise to do it" does so commit me. As Austin (1946/1979) wrote: "Promising is not something superior, in the same scale as hoping and intending, even to merely fully intending: for there is nothing in that scale superior to fully intending" (p. 99). There is a qualitative difference between "I intend to do it" and "I promise to do it." In promising to perform an action, one undertakes an obligation to do so. Austin continued: "When I say 'I promise', a new plunge is taken: I have not merely announced my intention, but, by using this formula (performing this ritual) I have bound myself to others and staked my reputation in a new way." There are other terms that obligate one in the same way as promise: for example, "I guarantee it" or "I give you my word." One can say, "I intend to do it but I don't promise to," but it is odd to say, "I promise to do it but I don't intend to." One can, of course, promise to do it and not intend to; this is an insincere or false promise. However, one cannot sensibly say it; this is an instance of Moore's paradox where speech act and mental state are in explicit conflict, as in, "It's raining but I don't believe it's raining" (C. E. Moore, English philosopher, 18731958). Children's Acquisition and Use of the Language of Intention Thus, there are three ways in which intention is expressed in language: implicitly, explicitly, and avowedly. An important question, to which I now turn, is when do children acquire and understand these expressions? Answering this question will provide information on children's developing
Page 304
concept of intention. The appropriate use of a linguistic term may not be a necessary condition for understanding the corresponding concept, but it is a sufficient one (Geach, 1957). "Appropriate" is the important qualifier here. Children may use words without fully understanding them and only careful experimental work will show the precise boundaries of the child's concept. The abstraction and complexity of the notion of intention guarantee that the concept will be developed over some considerable time—providing clear examples of cases of "use before meaning" (Nelson, 1996, p. 145). That is, children may not possess full adult control of some intention terms until years after they first produce the same terms in familiar contexts. Using and Understanding Implicit Expressions of Intention Almost as soon as children start to talk, they express their intentions in the first way I described, where intention is implicit in a linguistic expression. That is, they use terms like will and gonna to refer to their own and others' future actions. In Wells' (1979) study of children in Bristol, England, 50% of the sample had used will by 30 months of age and gonna by 33 months. However, as Olson and Kamawar (chap. 9, this volume) point out, it may be we who interpret such utterances as expressions of intention. Children may have little concept of intention at this point. Although will and gonna can refer to future action, the same terms can refer to future events. Children of this age use the terms in both ways, but may have no ability to distinguish between future actions that can be intended and future events beyond their control. Some time ago R. Brown (1973) said that when Adam, Eve, and Sarah started to use gonna (and also wanna and hafta) at about 2 years of age, the terms "were used to name actions just about to occur, a kind of immediate future which was often also a statement of the child's wish or intention" (p. 318). Recently, Montgomery (1998) analyzed Sarah's and Adam's transcripts, now available in the CHILDES database (MacWhinney & Snow, 1990), looking specifically at the use of intention terms (gonna, will, mean, try, and on purpose). The firstused terms are will and gonna, to express the future fulfillment of some act or some event. For example, Sarah at 3 years 0 months said, "I gonna get more water," and at 3 years 2 months said, "It gonna slip." Actions may be intended, whereas events may have some physical cause. Montgomery suggested that the children's early use of will and gonna is to express the future with no distinction made between intention and cause: Instead of saying that children are imputing intentions when they begin using auxiliaries, it would be more accurate . . . to say that they are using auxiliaries to state that an action or end is expected in the future. The
Page 305 distinction between future of present cause and future of present intention could be viewed as a developmental acquisition following from this subsuming rule. (Montgomery, 1998, p. 6)
Gee and Savasir (1985), who made an intensive study of will and gonna in the speech of two young 3yearolds playing together, agreed that both terms are used for future reference. They argued, however, that the two terms are not freely interchangeable. Will is most frequently used with firstperson singular in what they call "undertakings," that is, to sustain cooperative activity in immediate ongoing action (e.g., "I'll get some dolls out of my box" [p. 149]). Gonna is used in planning and may refer to actions beyond the immediate context or to the more distant future (e.g., ''Well I'm gonna skate when I'm a big sister" [p. 155]). However, even though the two terms are not interchangeable, they are not distinguishable on the basis of whether reference is made to fulfilment of cause or of intention. Montgomery's argument that 3yearolds do not distinguish between intention and cause parallels an argument I made that children of this age do not distinguish between intention and desire (Astington, 1993, chap. 6). As Brown (1973) said, Adam, Eve, and Sarah appeared to use gonna and wanna interchangeably, to refer to an immediate future that expressed the child's intention or wish. That is, intention may be expressed using wanna, a term in common use from 24 months or earlier. Bartsch and Wellman (1995) agreed that "announcing a desire to act . . . is often announcing an intention to act" (p. 68), and in their analyses found that the use of wanna was "often intimately linked to intention or goaldirected action," as for example: Abe (29):
Daddy! Fill my squirt gun.
Father:
Why? . . .
Abe:
Because I want to play with it. (Bartsch & Wellman, 1995, p. 70)
In sum, at this early stage (2 to 3 years of age) children use will and gonna to express predictions and intentions, and use gonna and wanna to express intentions and desires. Thus, one could argue that at this stage there is a conflation of prediction, intention, and desire. Furthermore, and most important in my view, it may be the acquisition of lexical terms allowing one to distinguish prediction, intention, and desire that helps the child to make these distinctions. Of course, older children and adults use will and gonna to express predictions and intentions, and gonna and wanna to express intentions and desires, but they can clarify what they mean if it is called into question. For example, "I'm gonna do it" could be disambiguated thus: "I'm gonna do it; I'll try not to, but I probably will" =
Page 306
prediction; "I'm gonna do it; I really intend to do it" = intention; "I'm gonna do it; at least, I'd like to" = desire. It is the acquisition of terms such as try and intend that facilitates the disambiguation. I am not proposing that the direction of causality is always from a linguistic to a conceptual distinction. However, in the case of concepts of abstract phenomena like prediction, intention, and desire, I believe that language plays a crucial role. Young children must experience desire and intention, or at least, they must have the experience of wanting something, and then of trying to do something to satisfy their desire. Harris (1991) claimed that even at a very early age children are aware of these experiences through introspection. Such awareness is crucial to his argument that young children do not need to think of mental states, like desire and intention, as hypothetical postulates within a theory of mind. In his view, young children are intuitively aware of their own experience, for example, of wanting or intending something. This is also what Olson and Kamawar (chap. 9, this volume) refer to, in saying that young children have "feelings." For Olson and Kamawar, feelings become mental states when the same state is recognized in others. In my view, the linguistic label plays a vital role here (Astington, 1996). The young child wants something but does not get it, or tries to do something but fails. The child has the experience, and the parent or caretaker talks about it, perhaps to comfort the child, incidentally using desire and intention terms. For example, "You want candy but we don't have any" or "You tried to reach it but you're not big enough yet" and so on. Most important, adults talk about their own and other people's mental states too, using the same terms, not always directly to the child but in conversation that the child overhears. For example, Dunn (1988, pp. 9091) described a scene where a 21monthold has accidentally kicked and hurt her sister and the mother intervenes: Mother to child:
Say sorry to Clare?
Child to sister:
Sorry (kisses and strokes her).
Mother to sister:
It was a mistake. She didn't mean to, darling. It was an accident.
Thus children can map the feelings they have onto others' experiences. They come to see themselves as possessing desires and intentions, and come to see that others do too. Between 2 and 3 years of age there is a marked increase in children's talk about desires, and notably, this is when children start to talk about others' desires, even though there is no increase in mothers' talk about desires over the same time period (J. R. Brown & Dunn, 1991). Also at this age children explain their own experiences in psychological terms, and ask for reasons for others' actions and reactions (Dunn & Brown, 1993; Hood & Bloom, 1979).
Page 307
Using and Understanding Explicit Expressions of Intention As children start to explain and excuse actions in the preschool years, they use terms such as mean, try, or on purpose (Bretherton, 1991; Dunn, 1991) that I discussed previously: intention made explicit by lexicalization. Dunn said that excuses and blaming based on intent were not common in her transcripts, but she gave some examples, even from children under the age of 3 years (Dunn, 1991, pp. 101102): 1. Child (26 months) climbing on his mother: Mother: You're hurting me! Child: Sorry. Sorry. I don't mean to. 2. Child (33 months) refusing to get down off settee: Mother: Get off! Child: Look. I'm not . . . [stands up on settee again]. Mummy. I'm not . . . I'm not standing on there. I'm trying to get a paint off [rubs TV]. I'm going to try and get a paint off. All I'm trying to do is . . . there. 3. Child (33 months) crying, saying her brother has bitten her: Mother: He bit you on the head? Child: Yes. Mother to brother: Philip, is that true? Brother: No. Child to mother: Yes! On purpose!
Although Dunn said that such expressions are not common, the frequency with which the terms are used is less important than the fact that they are used at all. Bartsch and Wellman (1995) showed that frequency of metacognitive term use by preschoolers is related to linguistic competence (measured by MLU), independent of age, but the first use of a term is related to age, independent of MLU. It may be use rather than frequency that is important as a marker of conceptual development. The child's concept of intention is gradually developed, as is the child's stock of intention terms. Young children may appear to understand intention when one term is used, and not when another intention term is used. In Bratman's (1987) terms, "commonsense psychology uses the notion of intention to characterize both our actions and our mental states" (p. 111), and it seems likely that children understand intention as characterizing action before they understand that intention can characterize mind. The term try, which marks action as intentional, is understood and used early on. In one study with Melanie Mann (unpublished data), we found that
Page 308
3yearolds could judge which of two characters was trying, although in a similar task they could not judge which of two characters meant to do something. We described brief scenarios, illustrated by pictures, involving the Sesame Street characters Bert and Ernie. The scenarios came in pairs, one featuring Bert and the other featuring Ernie. In the "trying" stories, both characters had the same desire, one acted but failed to satisfy it, and the other did not act but nevertheless got what was wanted. For example: "Bert is at the apple farm. He wants an apple off the tree. He can't reach it. He climbs on a box. He still can't reach it." The set of pictures illustrating this story was left in view and the second story was told: "Ernie is at the apple farm too. He wants an apple off the tree. It's too high up. He doesn't know what to do. Then the apple falls off the tree, onto the ground beside Ernie." With both sets of pictures in view, the child was asked the test question: ''Here's Bert, and here's Ernie. Which guy was trying to get an apple off the tree?" In the final frames, Ernie is shown with his apple and Bert is emptyhanded; nonetheless even 3year olds knew that Bert was the guy who was trying to get an apple. Trying is associated with acting. The term mean, however, may be more difficult. In some tasks, 3yearolds do appear to understand mean. If a character has a goal, acts, and achieves it, 3year olds can judge that he meant to do it, whereas if he achieves a different, unwanted goal they judge that he didn't mean to do it (Shultz, 1980). I argued that children's performance in tasks like these may be based on matching goals and outcomes, and judging matches as intended and mismatches as not intended (Astington, 1991). When that strategy is not available, 3yearolds have more difficulty judging which character "meant to do it"—for example, in a situation where both characters do the "same" thing, one with intention, and the other as a side effect of doing something else. This is like the situation illustrated in Fig. 15.2: In both pictures B and C, the birds eat the crumbs, but one can infer that only the character in B meant the birds to eat the crumbs (at least, one could infer it given more context). We contextualized situations like this for the "meant to" stories in the Bert and Ernie task. In these situations, the result was the same in both cases but only one character intended to bring it about. The matching strategy could not be used because the characters' intentions had to be inferred from the pictures and the story. For example: "Ernie has some bread. He takes it outside. He throws crumbs down. The birds peck them up." The set of pictures illustrating this story was left in view and the second story was told: "Bert's got some bread too. He walks along eating it. Some crumbs fall behind him. The birds peck them up." With both sets of pictures in view, the child was asked the test question: "Here's Ernie, and here's Bert. Which guy meant the birds to eat the crumbs?"
Page 309
Fig. 15.3. Mean scores at three age levels on intention and control tasks.
We tested eightythree 3 to 5yearolds on four pairs of stories: two with the try test question, and two with meant. The children also received two control story pairs, asking them to judge which guy was happy, in a case where the judgment could be made by matching goal and outcome. The order of story pairs and of scenarios within the story pairs was appropriately balanced across the children. Figure 15.3 shows 3, 4, and 5yearold children's performance on two "try" stories and two "meant" stories and two control stories. There was a significant difference between age groups only for "meant,'' F(2, 80) = 10.9, p < .001. Although try appears easier than mean here, one could argue that the tasks are not comparable in a number of ways; that is, the stories differ in more than just which verb is used. Lee (1995) gave 3 to 5yearolds a task with stories similar to the "try" ones in this study and counterbalanced "was trying to" and "meant to" in the test questions. She found, for the same story content, that the question with try was easier than the question with mean. It is likely that children's production and comprehension of expressions of intention continue to develop through the early school years. In an earlier study (Astington, 1986), I found that it was not until after 7 years of age that children were reliably correct on a task requiring comprehension of intends to, means to, plans to, to characterize prior intention, and intended to, meant to, on purpose, to distinguish accidental and intentional action. In this study, prior intention was marked in eight different ways: intends to, means to, plans to, and also, will, is going to, wants to, would like to, and thinks he'll Children were judged to be able to distinguish prior in
Page 310
tention from action if they were correct on six of the eight items (the probability of correctly choosing 6/8 = .132). Interestingly, comprehension of the verb intend most clearly distinguished between those who could recognize prior intention and those who could not. All the children in the former group understood the "intends to" item, even though it was not required for inclusion in that group, because they needed only to be correct on any six items. Using and Understanding Avowals of Intention Explicit avowals of intention are seldom included in inventories of children's speech acts, and instances of promising are rarely recorded (Snow, Pan, ImbemsBailey, & Herman, 1996). Nonetheless, the term promise certainly is used in conversation between parents and children. For example, the other day I saw a father running after his small child, calling out, "Come back. . . . You promised not to go beyond the big tree." Undoubtedly, parents do say "Promise me . . ." to hold the child to some particular course of action, and children say ''But you promised . . ." to remind the parent of some commitment made earlier. I am sure that even young children know that the term promise is used to emphasize the seriousness and importance of an utterance. Beyond that, however, their understanding is not the same as that of adults, and a long time is needed to acquire the mature concept. This is one of the clearest cases of "use before meaning" (Nelson, 1996). In earlier studies (Astington, 1988b, 1988c, 1990) I showed that it is not until adolescence that children understand that promising is only a speech act; that is, the utterance itself is the promise, even if the promised action is not performed. In one study (Astington, 1988b) 5 to 13yearolds heard six stories each containing a target utterance of promising, or predicting, or asserting. Half the children were not informed of the outcome; the other half heard stories in which a promise was kept, a promise was broken, a prediction was fulfilled, a prediction was unfulfilled, an assertion was true, and an assertion was false. For each story, the children's task was to decide whether the speaker of the target utterance had made a promise, and to justify their response. The most obvious change with age was that the older children clearly rejected cases of predicting as examples of promising, whether the prediction was fulfilled, unfulfilled, or the outcome was unknown, and they justified their responses by acknowledging that the speaker could not actually promise to bring about the outcome. As one 11yearold said, "Well, he isn't the one who controls the rain, so he couldn't promise 'cos it could rain in the afternoon." Children under 10 years of age ignored such facts and focused on outcomes in making their judgment.
Page 311
Even children who were not informed of the outcome imagined an outcome and based their judgment on that. Thus, the majority of 5 to 9yearolds said that a speaker had promised if the promise was kept or the prediction fulfilled, and judged that the speaker had not promised if the promise was broken or the prediction unfulfilled. They did, however, show some awareness of the distinction between promising and predicting. Children who had been informed of the outcome were asked a followup question to see whether they thought the speaker was responsible for bringing about the outcome. Responses to these questions showed that some 7yearolds and most 9yearolds did distinguish between promises and predictions in terms of the speaker's responsibility for the outcome, even though they did not make any distinction between them in judging whether the speaker had promised or not. Remarkably, even 11 and some 13yearolds misjudged actual promises, and said that the speaker had not promised in cases where the promise was broken. The older children in the group who were not told of the outcome sometimes thought they did not have sufficient information to make a judgment. When asked "Did he promise?" at the end of a story in which a brother promises to take his sister to the swimming pool next week, a 13yearold replied, "It's hard to tell unless you know what the guy's thinking." And a 12yearold said that she thought the speaker had promised, but then continued: ''If he was a person who usually kept promises like that, then I think it would be a promise, but if he didn't keep promises, then I'd probably say no, probably not." It is important to note that adults, given the same stories and questions, did not focus on outcomes, and judged that the speaker had promised in all cases of promising, even when the promise was broken or no outcome was given. It seems that until the adolescent years,' children regard the promise as the utterance plus the act. Alternatively, perhaps they regard only sincere promises as cases of promising, and they look for the act to assess the intention: If the promised action does not occur, or is likely not to occur, they judge the promise as insincere and thus not really a promise. However, as the adults agreed, promises are promises, whether they are kept, broken, or the outcome is unknown. Even an insincere promise is a promise; only the speaker knows that it is insincere. The most important feature of children's concept of promising, distinguishing it from that of adults, is that they do not think of promising simply as something done with words; for them it is not promising without the promised performance. As one particularly articulate 5yearold said, when justifying her response that the speaker had not promised in a story where the promise was later broken, "I know what promise is, you tell them you'll do it and you do it, that's a promise . . . . You have to do it if it's a promise and if you don't do it it's not a promise, it's a sort of joke but it's not a very nice joke."
Page 312
This incomplete understanding would not affect children's ability to make promises, although it does suggest that they might excuse their own—even perhaps others'— lack of performance of a promised act by saying that they didn't promise. Children from at least 5 years of age (the youngest age tested) can make promises in an elicited production task, although 5yearolds are unlikely to use the explicit performative verb promise (Astington, 1988a). Why Does Language Matter? In some circumstances, we can and do express and interpret intention nonlinguistically. One need only shop in a country where one does not speak the language, or watch a silent movie, or even a movie with the sound turned off, for this to become obvious. We infer intentions from what people do as well as from what they say. Donaldson (1978) emphasized the importance of this in early childhood—"the child does not interpret words in isolation—he interprets situations" (p. 88)—but this importance no doubt continues. Such considerations lead to my final question: Why does language matter; is language necessary? That is, what does it add to the behavioral expression and preverbal understanding of intention? Are there some things we cannot do without language (beyond making picayune distinctions between purposely and deliberately)? I would argue that the third of the "three ways" I have described is impossible without language: that is, intention avowed, commitment made. One cannot make a promise without language. In order to make a commitment to a future action that another can rely on, language is necessary: "I give you my word." Perhaps one might argue, for example, that an exchange of rings in a marriage ceremony is wordless and accomplishes the same end. However, this exchange accompanies the verbal vows, and the rings themselves are symbolic, as language is. Searle (1983) argued that intention is selfreferential. That is to say, the satisfaction conditions of intention refer to the intention. The propositional content of an intention is not [x happens], nor even [I do x], but [I do x in order to fulfill my intention]. Promising is similarly selfreferential; that is, I have an obligation to perform the promised act because I said that I would. Such selfreference—reference to what is said—is impossible without language. It is not necessary to use the explicit performative verb, promise, in order to make a promise, but it is necessary to use language. I agree with Searle (1969) that "I may only say 'I'll do it for you', but that utterance will count as and will be taken as a promise in any context where it is obvious that in saying it I am accepting (or undertaking, etc.) an obligation" (p. 63, emphasis mine). If it is not obvious that I am undertaking an obligation, I
Page 313
can make it so by using the performative verb promise. This is often the way the term is used, to make clear the extent of one's commitment, for example: "I will do it." "Do you promise?" "Sure." Austin (1962/1975) suggested that explicit performatives are likely to come later in the history of a language; that is, "I will" would appear earlier than "I promise that I will." This occurs in development too; children use "I will" to make a commitment at a younger age than they use "I promise" (Astington, 1988a). Austin suggested that it is complex social interaction that requires explicitness. Indeed, Rosaldo (1982) suggested that the reason promising is not found in languages such as Ilongot may be because making a commitment that another can rely on is not needed when one's actions are determined by one's position in the social hierarchy. Austin went on to argue that explicit performatives actually create new distinctions: Sophistication and development of social forms and procedures will necessitate clarification. But note that this clarification is as much a creative act as a discovery or description! It is as much a matter of making clear distinctions as making already existent distinctions clear. (Austin, 1962/1975, p. 72)
That is to say, when I use the term promise, as in "I promise I will," I make it clear that this is not just a prediction; it is my sincere intention that I am now obligated to carry out. The distinction between predicting and promising is not something that exists and that we describe; rather, it is only possible to make this distinction when we "do things with words." Finally and more generally, I would argue that even though intention can sometimes be expressed and interpreted nonlinguistically, a fully developed concept of intention is impossible without language. In everyday conversations, children hear linguistic terms that characterize actions as intended, premeditated, accidental, and so on. They start to produce these terms in familiar contexts, although they may not have a full understanding of the same terms in experimental tests of their comprehension. However, over time, meaning will develop from use, as Nelson (1996) showed in detail for a variety of semantic fields. The child's concept of intention is not all or none. It will develop gradually and will reflect and be reflected in the language of intention. Acknowledgments I thank Philip D. Zelazo for his helpful comments on this chapter, and the Natural Sciences and Engineering Research Council of Canada for its support.
Page 314
References Anscombe, G. E. M. (1957). Intention. Oxford, England: Basil Blackwell. Astington, J. W. (1986). Children's comprehension of expressions of intention. British Journal of Developmental Psychology, 4, 4349. Astington, J. W. (1988a). Children's production of commissive speech acts. Journal of Child Language, 15, 411423. Astington, J. W. (1988b). Children's understanding of the speech act of promising. Journal of Child Language, 15, 157153. Astington, J. W. (1988c). Promises: Words or deeds? First Language, 8, 259270. Astington, J. W. (1990). Metapragmatics: Children's conception of promising. In G. ContiRamsden & C. Snow (Eds.), Children's Language (pp. 223244). Hillsdale, NJ: Lawrence Erlbaum Associates. Astington, J. W. (1991). Intention in the child's theory of mind. In D. Frye & C. Moore (Eds.), Children's theories of mind (pp. 157172). Hillsdale, NJ: Lawrence Erlbaum Associates. Astington, J. W. (1993). The child's discovery of the mind. Cambridge, MA: Harvard University Press. Astington, J. W. (1996). What is theoretical about the child's theory of mind? A Vygotskian view of its development. In P. Carruthers & P. K. Smith (Eds.), Theories of theories of mind (pp. 184199). Cambridge, England: Cambridge University Press. Austin, J. L. (1946). Other minds. Proceedings of the Aristotelian Society, 20. (Reprinted in Philosophical papers, 3rd ed., pp. 76116, J. O. Urmson & G.J. Warnock, Eds., 1979, Oxford, England: Oxford University Press [page refs. are to this ed.]) Austin, J. L. (1966). Three ways of spilling ink (L. W. Forguson, Ed.), Philosophical Review, 75. (Reprinted in Philosophical papers, 3rd ed., pp. 272287, J. O. Urmson & G.J. Warnock, Eds., 1979, Oxford, England: Oxford University Press [page refs. are to this ed.]) Austin, J. L. (1975). How to do things with words (2nd ed.). Cambridge, MA: Harvard University Press. (Original work published 1962) Bartsch, K., &Wellman, H. M. (1995). Children talk about the mind NewYork: Oxford University Press. Bratman, M. E. (1987). Intention, plans, and practical reason. Cambridge, MA: Cambridge University Press. Brentano, F. (1960). The distinction between mental and physical phenomena. In R. M. Chisholm (Ed.), Realism and the background of phenomenology (pp. 39 61). New York: Free Press. (Original work published in German in 1874) Bretherton, I. (1991). Intentional communication and the development of an understanding of mind. In D. Frye & C. Moore (Eds.), Children's theories of mind (pp. 4975). Hillsdale, NJ: Lawrence Erlbaum Associates. Brown, J. R., & Dunn, J. (1991). 'You can cry, mum': The social and developmental implications of talk about internal states. British Journal of Developmental Psychology, 9, 237256. Brown, R. (1973). A first language: The early stages. Cambridge, MA: Harvard University Press. Donaldson, M. (1978). Children's minds. Glasgow: Fontana. Dunn, J. (1988). The beginnings of social understanding. Cambridge, MA: Harvard University Press. Dunn, J. (1991).Young children's understanding of other people: Evidence from observations within the family. In D. Frye & C. Moore (Eds.), Children's theories of mind (pp. 97114). Hillsdale, NJ: Lawrence Erlbaum Associates. Dunn, J., & Brown, J. R. (1993). Early conversations about causality: Content, pragmatics and developmental change. British Journal of Developmental Psychology, 11, 107123.
Page 315
Forguson, L. W. (1985). Unpublished commentary on Astington, J. W., 1985, Children's understanding of promising. Unpublished doctoral dissertation, University of Toronto (OISE). Geach (1957). Mental acts. London: Routledge & Kegan Paul. Gee, J., & Savasir, I. (1985). On the use of will and gonna: Toward a description of activitytypes for child language. Discourse Processes, 8, 143175. Halliday, M. A. K. (1970). Language structure and language function. In J. Lyons (Ed.), New horizons in linguistics (pp. 140165). Harmondsworth, England: Penguin. Harris, P. L. (1991). The work of the imagination. In A. Whiten (Ed.), Natural theories of mind: Evolution, development and simulation of everyday mindreading (pp. 283304). Oxford, England: Basil Blackwell. Hood, L., & Bloom, L. (1979). What, when and how about why: A longitudinal study of early expressions of causality. Monographs of the Society for Research in Child Development, 44(6, Serial No. 181). Lee, E. (1995). Young children's representational understanding of intention. Unpublished doctoral dissertation, University of Toronto (OISE). MacWhinney, B., & Snow, C. (1990). The Child Language Data Exchange System: An update. Journal of Child Language, 17, 457472. Miller, G. A., &JohnsonLaird, P. N. (1976). Language and perception. Cambridge, MA: Harvard University Press. Montgomery, D. E. (1998, June). A Wittgensteinian perspective on children's acquisition of mental terms: The case of intentions. Paper presented at the Annual Meeting of the Jean Piaget Society, Chicago, IL. Nelson, K. (1996). Language in cognitive development. New York: Cambridge University Press. Piaget, J. (1977). The moral judgement of the child. Harmondsworth, England: Penguin. (Original work published in French in 1932) Rosaldo, M. Z. (1982). The things we do with words: Ilongot speech acts and speech act theory in philosophy. Language in Society, 11, 203237. Ryle, G. (1963). Ordinary language. In C. E. Caton (Ed.), Philosophy and ordinary language (pp. 108127). Chicago: University of Illinois Press. Searle, J. R. (1969). Speech acts: An essay in the philosophy of language. Cambridge, England: Cambridge University Press. Searle, J. R. (1983). Intentionality: An essay in the philosophy of mind. Cambridge, England: Cambridge University Press. Shultz, T. R. (1980). Development of the concept of intention. In W. A. Collins (Eds.), Minnesota symposium on child psychology (Vol. 13, pp. 131164). Hillsdale, NJ: Lawrence Erlbaum Associates. Snow, C. E., Pan, B. A., ImbemsBailey, A., & Herman, J. (1996). Learning to say what one means: A longitudinal study of children's speech act use. Social Development, 5, 5884. Wells, G. (1979). Learning and using the auxiliary verb in English. In V. Lee (Ed.), Language development (pp. 250270). London: Croom Helm.
Page 317
Chapter 16— Intentionality and Interpretation Carol Fleisher Feldman New York University Intentionality is one face of a twopart complementary process. The other face is interpretation. In the construction of every intentional act or utterance, cultural patterns of interpretation that will be necessary for communication are also engaged. Every speaker is also at times a listener. Although intentionality is expressed for others, it is also personal. One of the two basic constituents of intentionality is a psychological mode that packages personal meaning for public consumption. The interpretive exercise is a challenge, requiring one to understand the inner life of another from public expression. The form that this understanding takes, at the first step, is assigning an intention to an action or utterance. Looking at intentionality this way, we can see that there must be many kinds of intentionality because there are surely many kinds of personal meaning, and many kinds of interpretations. Nevertheless, the many kinds have an important structural commonality, and they even share a great deal with simple intentions. Intentions are begging for more focused attention. They seem to be the criterial feature that distinguishes human, or at least animate, action from motion in the physical world. But what are intentions? Acts of meaning or desires and beliefs? Or simply desires? Or are they dispositions to act? And what is their relationship to intentionality, a more abstract, underlying mental attitude toward the world's objects. Are intentions necessary for intentionality? The same is even truer of intentionality. It seems an essential feature of humans, an ability to organize the world's objects under a point of view
Page 318
that no physical, or perhaps even animal, system shares. Indeed, it seems so essential to humans that it has been identified by some evolutionists as the key element of the great evolutionary leap to humankind. But the same puzzle arises here. What is intentionality? Is it a stance or mental attitude toward a mental object, toward a physical object, or what? And what is required to have a mental attitude, or what does intentionality involve that simple intentions lack—a belief, or consciousness, or what? Are simple intentions the mere wanting (to do), while intentionality can only be achieved when reason is (perhaps also) involved? Is the difference between simple intentions and intentionality a mere matter of scope—intention distinguishes (human) physical motion, whereas intentionality is what can mark a human being as human even at rest? I do not resolve the elusive connection between these two systems by distinguishing between them more sharply. Rather the approach here is to put them both into a single pot of intention/intentionality, and then try to analyze that. For one thing is plain: Intention and intentionality have a good deal in common—both distinguish the human from the physical world in some way. Rather than disentangle them, in this chapter, I consider the common universe of intentions and intentionality. My main point is the unsurprising one that the intention/intentional universe contains a vast number of interestingly different kinds of events. All these events have a further singular property: Corresponding to every kind of intention/intentionality, there is a kind of interpretation that is essential to it, as I explain. Every kind of intention/intentionality reflects two systems, one intellectual and the other social. In real life, expressed intention, whether the simple intentions of action, or the more heady intentionality of intentional action or even intentional utterance, carries both personal meaning and social meaning. The genius of the "intention/intentionality—interpretation" system is that it allows both the personal and the social to be fully represented in a single signal. The manner of the expression of the personal, within this system, becomes canonical enough to be communicable and part of the coin of friendship and connection with others, and yet, somehow, manages not to lose contact with the personal idiosyncracies that compose our ideas about self. This miracle, through which personal intention is given socially accessible expression, is what makes human life in groups as we know it possible, with the social meeting of minds. The communicability of intentions is a function of a complementary process of interpretation that is twinned to every intention. Interpretation is essential to intentions for a special reason: In most cases neither the intention of action nor the intentionality of what might be action, inaction, or utterance are explicitly expressed in them, nor therefore, directly observable. They have to be arrived at by the "observer" via an inferential
Page 319
process that is under the control of a theory the observer has in mind (which is, of course, a theory about what the observed other has in mind, and therefore is itself intentional, as well as being metamental). The theory the observer has in mind must be an interpretive theory that relates the myriad intentions and intentionality common in everyday life to their patterns of expression in various contexts. And, as it is part of common cultural coin, it is a theory also known to the observed person. They collaborate. The actor/interpreter acts in a fashion consistent with the action/interpretation pattern known also by the observer. The two—action and interpretation—are thus coordinated in every individual action and interpretation, and this is how actor and interpreter are able to competently perform their separate roles at any particular moment. Indeed, Anscombe (1958) defines intentional actions in terms of the processes of comprehension they give rise to. For her, "intentional actions are actions to which a certain sense of the question 'why?' has application" (p. 11). For Anscombe, to seek to understand the intentions of others is to assume that their actions have meaning, and make an effort to discover it. To succeed in this effort after meaning is to interpret the other person's actions correctly—namely, to identify the right intention(s) behind them. Analyses of the concept of intention and intentionality have appeared in the theoretical work of nearly all the major psychologists working on the topic of theory of mind, and several philosophers have given the matter serious attention, perhaps most notably John Searle in his book, Intentionality. Searle (1983) began with Brentano, who first formulated the idea of intentionality in the form we know it in today. Brentano (quoted in Edwards, 1967) wrote: Every mental phenomenon . . . includes something as an object within itself, although not always in the same way. In presentation something is presented, in judgment something is affirmed or denied . . . in desire something is desired, etc. This . . . is exclusively characteristic of mental phenomena. No physical phenomenon manifests anything similar. (p. 201)
Some time after Brentano's first formulation in 1874, the concept of intentionality ceased for a period to be important to a psychology then in the grip of behaviorism. But it has come back with force in the last decade, propelled by the new literature on theory of mind. There are currently several different versions of intention and intentionality within this literature, and no canonical version. But I believe that this state of affairs is not pernicious partly because the many views of intention (and of intentionality) are not mutually exclusive, although they may collide on details. To me it seems that they largely share a family resemblance. Let me list ideas about simple intentions first:
Page 320
1. All actions express or contain intentions. 2. Intentions are dispositions to act. 3. Intentions are desiresinaction, but need not involve a belief. 4. Intentions are states of desireplusbelief because the belief that the desire could be realized in a certain way in the occurring conditions is necessary to convett a simple desire into a disposition to act. 5. Intentions can take the form of an intentioninaction, or, more grandly, an a priori intention, in which case they may be conscious, too. 6. Intentions are meanings in the special sense that they are the meaning of an action, or its interpretation. 7. Intentions may be expressed in speech. Next I discuss intentionality. Some notions of intentionality are: 1. They take the form of an "aboutness." By "aboutness" we usually mean that some mental state is directed at something else, usually an object in the mind. But it might also, or even alternatively, be that 2. The mental state is directed to a real thing in the naive real world, a version preferred by many of the new breed of philosophers interested in "cognitive science." 3. A third notion, closely related to the first two, says that whatever it is about, every intentional event must contain a psychological component that is about something, and about it in a special way that is something like a hierarchical inclusion, or perhaps better, a function over an argument. 4. A fourth notion requires a high level of sophistication in the psychological component, particularly that it must include a good deal of real world knowledge—about people especially. 5. A fifth says that true intentionality cannot be expressed without language, perhaps not without complex utterances containing (at least) a propositional attitude and a proposition. For both simple intentions and for intentionality, the nature of the beast is not clear, or not unique. In both cases, there are some serious discrepancies among the claims enumerated here. Indeed I have tried to point to some of them within the list above. And there are, no doubt, difficulties that could follow from this. But again I do not focus on the discrepancies and differences. Rather, my plan is to be inclusive, considering that there may be useful riches in the brew. There is a good possibility that depending on how one comes at it, and what phenomena one has
Page 321
in mind, one may tend to emphasize some theoretical aspects rather than others, which may lead one in turn to generate one version of the intention/intentionality rather than another. For example, there is surely a place for intentions in action and intentions in utterance, for meaning in intentionality as well as intention, and perhaps even for intentionality toward physical objects as well as mental, although for psychology it would seem that the mental object is the more basic of the two. (But would, for example, shouting at a broken television work as an example of an intentionality toward things that bypassed the more usual mental object?) In any case, there is a good deal of diversity among observed forms of intentions/intentionality, and some may be related to these seeming discrepancies. Better for the moment to be inclusive, for if the intention/intentionality system is somehow essential to humanness, there could be as many ways of intending as there are ways of being human in the world. Moreover, there are some important areas of agreement. Here is the undisputed core as I see it: With respect to intentionality, all views agree that it is composed of a special twopart unit that contains both a person's attitudes and a piece of the world toward which these attitudes are directed or that they are "about." This special unit both looks into the mind and out to the world, and links them together via such simple intentions as directedness of the mental state to something else. Thus, simple intentions are basic to intentionality. Furthermore, this special hybrid unit composed of a linkage of world and mind can be unexpressed or expressed, and if expressed, expressed both in action and in speech and other symbolic systems, where its expression is more elaborated and so its workings are perhaps easier to see. Schematically, we can represent this basic pattern as a general case in which each element is named with the most neutral term I could find for it (see Fig. 16.1). A generic notion called "intentional events" is decomposed into generic parts. Intentional events always include a person who has a psychological mode ranging over contents, a psychological mode, and the contents of it. The psychological mode is a relation of the person to the contents. Both the psychological mode and the contents can vary in
Fig. 16.1. Generic intentionality and three intentional systems.
Page 322
interesting ways, and it is from their combinatorial that the range of intentional events is generated. Below the general case in Fig. 16.1 are listed three more specific forms of intentionality: the intention in the head, the intention in action, and the intention given symbolic expression using speech or other, say visual, symbolic systems. The essential requirement of the expressive system is that it be shared by an interpretive community, an insight I owe to David Olson (1994). I return to these specific intentional systems in a moment. Next, the chart proceeds to decompose the generic notion of intentional ''event" into its three main subtypes: states (internal events), acts (events expressed behaviorally), and speech (events expressed in words). These three are meant to capture the possibilities for the form of expression of intentionality in everyday life, their form of packaging for the social other who will provide the complementary interpretation. The packaging ranges from no overt indication of intentionality (states), to an overt indication (in action) that there must be an intentional state, to finally the definite indication of intentionality in utterance. They range therefore in how definitely they indicate the existence of an intentional state nearby. The matter of what specific intentional state that is, and how explicitly it is expressed, is a somewhat separate matter, for even when intentionality is expressed in utterance, there is a range of explicitness, and the further project of determining what specific intentionality is involved can be a major one. Nevertheless, in general there is a rough relation between overtness of marking of Intentionality and explicitness of the intentional mode. In general, actions are more explicit than mere states, and speech more explicit than acts. With language comes the possibility of a precise specification of the mental modes (which has further consequences we return to later—it may invite a greater precision in thought as well, and may also have consequences for opening up the range of possible contents). Thus, by chance, the scheme corresponds roughly to level of explicitness as well as of overtness. But that is an accident resulting from the loose correlation between overtness of marking and explicitness. The levels are meant to be levels of overtness of marking. Let us return now to the generic pattern, and consider its components: psychological mode, and contents. These two components can be used to generate a good many different kinds of intentionality, even if we do not take level into account. Turning to contents for a moment, we can try to get some notion of the range of possible kinds of contents. Some common kinds are quotidian contents, or propositions about everyday life; professional contents specialized to each work community; transcendental contents of the kind found in religions; fantastic contents, as in stories about outer space; and finally metacontents, which contain an intentional event as contents. Perhaps each kind of content contributes its distinctive character to a
Page 323
distinct kind of intentionality corresponding to it (as, perhaps, does each psychological mode). Over the various kinds of contents range different kinds of psychological modes for perceiving them: wishes, hopes, doubts, denials, affirmations, and beliefs, among them. The presence of some psychological mode is normally part of intentionality. Nevertheless, sometimes no real mental state can be found in the position reserved for one, as in the case of a null intention. One important case of null intentions occurs when an agent wants to deny responsibility for an action they performed—"I felt like I did not apply, but rather that I was recruited." The case of the null psychological mode, which I suppose serves as a placeholder that makes the whole expression come off as intentional in a sense, is nevertheless very odd. What makes it odd, I think, is that intentionality is normally as much part of a larger project of selfconstruction as it is part of an act of communication. With a null mode, the heart goes out of the selfmaking project, even if the communicational goal is still satisfied. In principle, any psychological mode can be paired with any kind of content, but certain relations between modes and content will seem more natural than others. This is bound to vary across cultures and persons, and perhaps even times and social context, but there may be some pattern to it all the same. At the very least, certain pairings may achieve canonicality within a social group, and others may come to be seen as bizarre. The canonical could elicit approval and solidify human connections within the group, or, and this could occur even within the group, it could come to be seen as banal. And such canonical pairings could serve as markers of outsider status when the person is with people outside his group, if the canonical pairings of his own group are a part of general cultural knowledge. Some settings require canonical pairings, and innovations could be seen as irreverent or rejecting. Other settings may invite innovation, and see them as creative or brilliant. Part of our standard cultural knowledge is knowing what kind of situation one is in—one requiring conformity or not. If it does, we should either know what the sitution's canonical pairings are (so that if we violate them can we do so knowingly) or keep quiet. And if it does not, we had best avoid too much canonicality at the risk of seeming banal, or even inauthentic. Some persons may tend to the canonical or the anticanonical, or even uncanonical, and many people may tend toward a single pairing of mode and content in some few areas. This can be very rigid or narrow indeed, when a person can only think of certain contents under a single mental state, or a certain mental state always activates the same contents. "If it fills me with awe, it must be from Venice." Although such general tendencies of a person, along with any specific pairings of mode and content distinctive to them, may be more or less successful from a communicational
Page 324
point of view, they are very important indicia of the self. These selections of general style and specific quirk are important tools in the work of selfmaking and self presentation. I noted above that the modes may be merely states, or they may be expressed either by acts, or by utterance. One may have a psychological attitude that is a mere state of being, one that is not expressed in action or speech at that moment, or even, perhaps, at any time at all. I assume that such mental attitudes underlie or accompany most expressions of mental attitude, whether in action or speech, and that they are basic. And yet both forms of expression, action and speech, probably contribute something to the individuation and definition of such states, something that develops the unexpressed state into a more specific mode. Both action and language are systems with patterns of their own. Once decisions are made about how such a state is to be expressed in such a pattern, a certain specificity is introduced that may actually add something to the latent meaning. This line of reasoning leads to the conclusion that although the unexpressed state may be the most basic form of the psychological constituent of intentionality, their full expression in speech may be the most fully realized form of the psychological modes. Accordingly, in considering the psychological modes, it may make sense to jump two levels to the expression of psychological mode in utterance, and see what the variety of utterance types may contribute to the manykindness of intentionality. I should also note that not all contents seem to be possible at all levels either. For example, quotidian contents seem particularly suitable for action, whereas metacontent seems particularly unsuitable for it. But not all nonquotidian contents require language, and some may be very tied to action. For example, transcendental contents may be particularly tied to the action of wordless prayer or meditation, and meditation may involve metacontents as well, even if metacontents seem generally more suited to speech than action. And professional contents might be blueprints, or even images of blueprints in the mind. There is no match of abstractness of content to level, but there are definite signs that contents and level must be suitably paired. In contrast, the modes seem to have their fullest realization in speech. As noted earlier, psychological modes may be named by an intentional utterance or they may merely be hinted at. For example, we can doubt by saying so, or by uttering a declarative sentence in a certain context and tone of voice. We can say we regret the small size of the portion or we can speak of its only having been a small portion. Elsewhere (Feldman, 1974), I described a continuum of explicitness in intentional expressions in English that begins with simple emphatic stress, and moves on to such syntactic devices as clefting, that foregrounds a propositional element, e.g., "It was John who married Mary," or "It was Mary who John married," as clefted versions of "John married Mary." In these cases, although we know
Page 325
the speaker was making a point of the clefted element, we still don't know what she had in mind to say about it. From there, with increasing explicitness, we go to the intentional adverbials even, only, just, and still. Then we go to the modal verbs before we reach lexicalized reports of mental state such as the propositional attitudes. Implicit expressions of psychological mode tell the hearer that there is an intention around somewhere, but they do not say what it is. The more implicit the form, the more it invites, even requires, interpretation to complete it. Thus, the painter Marcel Duchamp, trying to get away from what he called "retinal" art that appeals to the eye but leaves nothing for the mind to do (Dorment, 1997), titled his masterwork (The Large Glass), "The Bride Stripped Bare by Her Bachelors, Even [my italics]." Implicit intentions are probably the normal form of intentional expression. For, in fact, fully lexicalized expressions of intention do not leave room for what I take to be the interpretive query basic to the intentional domain—namely, "Why did he do (or say) that?" where the answer is an intentional state such as a wish, hope, or belief. Although of course we can go on, once a basic intention is in hand, to ask "why" about that too—e.g. ''Why does he hope that?"—this "why" seems to me less fundamental. Indeed, its answer may be of a different kind, not a psychological mode but, for example, conditions of childhood as in psychoanalytic explanation, or an appeal to realworld constraints. Moreover, the virtue of implicit psychological modes are that they protect the speaker against too much selfexposure to a listener who may turn out to be unsympathetic. The manner of symbolizing gives rise to a few more intentional kinds. When intentionality is encoded in a symbolic system, intentions can be expressed by speakers literally on the one hand, or metaphorically, mockingly, or ironically on the other. Moreover, they can be offered earnestly or formulaically, giving rise to the (nonpernicious) pseudointentions of ritual greetings. The basic schema would keep an interpreter busy enough, but in fact interpretations seldom are restricted to oneaction or onestatement packages, for the very good reason that understanding any single intention is often helped by considering the action or speech occurring before and even after it. We cannot afford not to interpret each individual action or utterance to discover its intention, but equally, we cannot avoid considering other intentional acts or utterances around it and how our target action or utterance may fit into the series formed by them. Taking this series into account invites us to construct larger interpretations that consider series of intentions, and the means by which one follows another—the ands, buts, ors, thens, and neverthelesses of narrative. The interpretation of such sequences of intentions often rises to the level of narrative, a matter I cannot take up here, except to note that in narrative the complex intentions are
Page 326
put together in a patterned way. These, then, could lead to an interpretation in terms of another intentional kind that could be called patterned intentions, or even, narrative intentions. Interpretation of larger units relies very heavily on culture. It does not begin until the interpreter embraces the intending agent as a member of a familiar cultural community. Then it proceeds to infer the right intention by making use of cultural knowledge of all kinds: what people of various kinds are like, how situations affect them, what it feels like to be related to different kinds of contents, and how, in their culture, these various considerations are put together into the coherent wholes of the culture, which is where standard stories and story genres come in as crucial interpretive equipment, creating such possibilities as fictional, realistic, and autobiographical intentionality. In this chapter, I have explored many intentional kinds that derived from my proposed schema. In addition, I have pointed to the absolute essentialness of considerations about interpretation for an understanding of intention. I have tried to show that distinguishing among kinds of intentions requires bringing in considerations about the kinds of interpretive processes that complete each of the distinct intentional states. The basic schema of intentional events is that they are composed of a person and of the person's mental state about something. The relation between the person and the contents, the psychological mode, is the first constituent; the contents are the second. The two together constitute an intentional event. If an intentional event is made public in intentional acts or in intentional speech, an observer can try to explain the why of it by interpretation. This might involve decomposing it into the same constituents from which it was constituted in the first place: psychological mode, and contents. The person will be recruited in the act of interpretation as well, by specifying the cultural universe in which the psychological mode and contents are found. I note that interpretation is done by a person who need not be an interlocutor, but can merely be an unengaged observer, can even be the self, usually in a setting in which the two persons—the intender and the interpreter—are somehow connected, whether at a basic level by sharing a space, or, more commonly, by being members of a cultural community. Interpretation is seen everywhere intentions are found. There is nothing fancy about it. Consider Menzel's (1974) chimps following each other, even heading each other, on the way to food. These are surely interpretations of acts, interpretations that take the form of attributing foodgetting intentions, interpretations that take place even in the absence of a constructed cultural framework. What stands in for culture here? Perhaps it is the shared ecological framework whose affordances for survival are matched by shared intention.
Page 327
Such primitive interpretations can also arise for humans. For example, when a group of strangers find themselves trapped in limited space with scarce resources, it is easy to see them each as principally occupied with his or her own resourcecollecting intentions, and the group as a whole defined principally by a clash of competing individual intentions to get the resources. Everyone shares a single psychological mode; only the contents would differ, and even these would be largely overlapping. Under pressure, the conditions may determine intentional states to such an extent that individual differences in the play of mind are overwhelmed. Primitive intentions like these are completed by primitive interpretations. Interpretation here would be limited to discovering for each person precisely what resources they are trying to get, or their precise intentional object, everyone's modes being alike and dictated by the conditions. The experience of a loss of self and humanity under desperate conditions must be tied up with this. More often in human affairs, both intentions and their interpretations are very complex, drawing on vast pools of common cultural knowledge. The mental processes involved, even for quotidian intentions, can be very demanding, perhaps far more difficult than those called for in everyday scientific thinking, for example. This may have in part to do with the artistry required to recruit a suitable pool of background knowledge for making an interpretation, a task made essential when the psychological mode is not expressed explicitly, as is often the case. The lexicalized propositional attitudes are the most explicit expression of a good many psychological modes that often, perhaps even usually, have other, less explicit forms of expression. These properties—the artistry of interpretation and implicitness—tend to make interpretation a hard project, and yet eventually, as among experts, interpretation can attain the clarity of legal reasoning and literary criticism. Therefore, it should not be surprising that it takes a while for the developing child to learn how to do it. But what this late onset may not take sufficiently into account is the potential for further growth in interpretive thinking. If ever there was a good candidate for significant cognitive growth across a lifetime, this would be it. Acknowledgments I am grateful to the Spencer Foundation for its support of this work through its grant to Jerome Bruner, "MeaningMaking in Context." I made extensive use of John Searle's analysis of intentionality in his book of that title, of Jerome Bruner's work on forms cognition that include interpretive thinking, and of David Olson's insights about the importance of an interpretive community. If I have contributed anything here, it is because I was standing on the backs of giants.
Page 328
References Anscombe, G. E. M. (1958). Intention. Oxford: Basil Blackwell. Dorment, R. (1997, April 4). Stripped bare at last: Wit, intelligence, desire, the true legacy of Marcel Duchamp. Times Literary Supplement, pp. 34. Edwards, P. (Ed.). (1967). The Encyclopedia of Philosophy (Vol. 4). New York: Macmillan. Feldman, C. (1974). Pragmatic features of natural language. In M. LaGaly, R. Fox, & A. Bruck (Eds.), Papers from the Tenth Regional Meeting of the Chicago Linguistic Society (pp. 151160). Chicago: Chicago Linguistic Society. Menzel, E. (1974). A group of young chimpanzees in a oneacre field. In M. Schrier & F. Stolnitz (Eds.), Behavior of nonhuman primates (Vol. 5). New York: Academic Press. Olson, D. R. (1994). The world on paper. Cambridge, England: Cambridge University Press. Searle, J. R. (1983). Intentionality. Cambridge, England: Cambridge University Press.
Page 329
Chapter 17— The Intentionality of Referring Jerome Bruner New York University When one speaks of the intentionality of some sign or symbol, one usually has in mind the manner in which it "stands for" something else, as in the Peircean distinctions among indexes, icons, and symbols (Peirce, 19471962). I examine, rather, the intention inherent in acts of indicating: how somebody communicates to another person that there is something particular at the focus of his attention that he wishes to bring to the attention of that other person, in return for which he wants some indication that the other has, as it were, "got the message." On this account, the sender must (a) indicate that he is seeking to draw another's attention to something, (b) indicate what it is that he is trying to draw another's attention to, (c) receive some sign whether he has succeeded or not, and (d) if not, figure out with or without the other's help what further steps need to be taken. For the highly socialized human adult, this fourpart act typically becomes so routinized that we take it for granted, like walking or using the phone. Worse still, we may even imagine that it is simply "innate," thereby obscuring the support conditions that bring it to fruition and sustain it, ignoring the technical virtuosity involved in mastering and using signs and symbols in our acts of communicating with others. Because my own concern is with the development of acts of referring, I concentrate on how this fourpart act, as I've called it, develops in concordance with other features of human development. It is only when we deroutinize acts of indicating or referencing by unbalancing the accessibility of information available to participants in a
Page 330
referential exchange (e.g., Clark, 1992, chap. 4), or by letting young children manage it on their own, or (even more revealing) by setting up seemingly easy referring tasks for higher apes (Crawford, 1937; SavageRumbaugh et al., 1993) that we realize how contingent, fragile, and collaborative acts of indicating are. Alas, the philosopher's pristine account of signs, significates, and interpretants contains no frustrated people trying vainly to make their indicating intent or its object clear, nor does it deal with odd target domains from which referenced objects must be selected. It is free of messy polysemy and puzzling contexts to be "read." Such accounts have altogether too little to say about how ordinary mortals bring off a meeting of minds under ordinary conditions. Peirce (19471962) made an effort to deal with the problem by invoking an interpretant as mediating between a sign and its referent, but his account of it leaves too much underspecified for a psychologist's needs. The struggle to share with another what one has in mind, how one constructs and communicates one's interpretant, begins early. Even very young children are sensitive in conversational exchanges to an adult's "What?" questions (Feldman, 1996; Kasermann & Foppa, 1982), responding to them with expansions of their utterances, as if appreciative of their interlocutor's difficulties. They seem to be sensitive to a problem domain and have some early intuitions about how to cope with it. So too with the distinction between what is said and what is meant: Although the general distinction is early grasped (and early exploited pragmatically!), its particulars never cease posing a problem in interpretation, even with respect to what the speaker is referring to. Gricean implicatures, subtle use of context, even ordinary ellipsis turn the referring problem into something more complicated and "psychologistic" than standard textbook accounts. I proceed with the conviction, then, that for a psychologist it is probably better to treat the intention to refer and the act of referring as rather more mysterious than it is commonly taken to be. The Beginnings Let me begin with the ontogeny of the matter. From a very tender age, human beings seem "motivated" to bring things at the focus of their experience to the attention of others. At the outset they seem to do so without wanting to do anything further about it than to achieve some mutuality of awareness. This mutuality of awareness, this primitive intersubjectivity, seems to precede (and to be a necessary condition for) more specific referential acts. Early indicating takes a variety of forms. The young infant very early on uses reaching and grasping movements to signal an object of desire to his
Page 331
or her caretaker—even to signal desire itself. And when some desired object has a canonical locus (like fruit juice in the fridge), an infant may reach importunately toward that canonical locus with appropriate vocalization. But around the same time that ordinary "demand" gesturing appears, infants also show signs of wanting to share attention with another just, as it were, for its own sake and with no discernible extrinsic motive. Such selfsustained referencing is strikingly exemplified in what psychologists have come to call joint attention (Moore & Dunham, 1995): Infant and caregiver look jointly at a common object, then look back to each other eyetoeye with evident enjoyment. It may even be, as some argue (e.g., Brazelton & Als, 1979), that such shared awareness of things in the world requires direct eyetoeye contact as a precursor. For both mother and infant seem initially to derive great pleasure from just looking directly into each other's eyes (Stern, 1985). Indeed, if the adult of the pair fails to reciprocate in such an exchange, the child may become acutely distressed (Stechler & Latz, 1966). And so do the parents of autistic children when the latter fail to reciprocate to efforts to make eyetoeye contact. But once normal eyetoeye contact is established in the opening months of the child's life, it soon becomes elaborated to include a "third object." Typically, the mother introduces an object close to the mutual line of regard, makes some conventional attentiongetting vocalization (Fernald, 1989, 1991), and turns to the object in question. The child follows the mother's line of regard to the object, the two looking at it in unison, after which they resume direct eyetoeye contact with seemingly renewed delight (David & Appell, 1961). Stern's (1985) expression for such bouts of sharing, attunement, seems well chosen. By the fourth or fifth month, the child is able to track an adult's line of regard without prior priming in an eyetoeye format. In the Scaife and Bruner (1975) study, for example, the child sits facing an experimenter who exclaims (with "marking" intonation), "Oh look!" and thereupon turns to an object. Far in excess of chance, even 6 montholds turn in that direction to search for a visual target. If they fail to find one, moreover, they often turn back to inspect the adult's gaze direction again (Butterworth &Jarrett, 1991). Gazefollowing increases steadily over the following months. But note that it is still a comprehension rather than a production skill: They are reliable gazefollowers before they point for the first time, even though they already use directional reaching to indicate objects of desire. At around 7 to 9 months, the child finally begins pointing to bring objects or events to another's attention. Pointing is a very special gesture, not remotely like reaching in order to signal a desired object. It is specific to our species. It seems intended not only to bring something to another's attention but also, as it were, to mark it as nonordinary in some way. I can illustrate this last matter by a child's first observed act of pointing that we
Page 332
were fortunate enough to catch on video. Richard, whose communicative development my assistant and I (Bruner, 1983) had been following almost weekly since his sixth month, was now into his eight month and off with his family on a summer holiday in the country. His first fullfledged point, quite an "excited" one, was toward a small flock of birds flushed up as he was toddling about a small field adjacent to the summer cottage where he was staying. His pointing (if I may be permitted some "thick interpretation") seemed as much jubilant as communicative. He followed it immediately with a twice repeated vocalization, "Burr burr," looking directly at the research assistant in attendance. A highspirited and responsive young woman, she in turn lit up, smiled, and supplied the typical expansion and idealization: "Yes, Richard, it's a bird,'' with exaggerated prolongation of and stress on the final lexeme. One even gets the impression, watching the video, that Richard's excitement over the flushed covey of birds was sharpened by his own wellformed firsttime act of pointing to which my assistant so spontaneously and fulsomely responded. The occasion seemed rather like an epiphany for both of them. In the months following, there were similar (although less dramatic) episodes, sometimes accompanied by a naming attempt, sometimes only by attentiondrawing vocalizations such as the protodemonstrative da. They seemed always addressed to markworthy objects or events: familiar objects in unfamiliar settings, unfamiliar ones in familiar settings, things that were "out of the ordinary," things that were being searched for and then come upon. Richard did not point at humdrum objects. Nor, let it be noted, did he point at objects he merely desired: For that he continued to "reach demonstratively" or to use vocal signals with implorative contours. In all instances observed (whether on or off video), Richard's pointing was in the presence of somebody from whom he expected an appropriate recognition response. Lest it be overlooked, let me comment on one thread that runs through all the foregoing observations. It is not only that infants and young children (as we like to tell our undergraduates) are "naive realists" who believe that there is a world out there to which they have direct access. None of us ever quite gets past that. Rather, they believe too that they and their conspecifics experience the same world. When they find no object along the adult's line of regard, they turn back to recheck adult's gaze, then look back out again. That is to say, the young child can be provoked into hypotheses about objectsintheworld by others seeming to be looking at (or listening to) them. The world of things, then, is also putatively an interpersonal world as well. Although this early intersubjective representation of the world has become a standard topic in developmental texts (e.g., Astington, 1993), its implications for a theory of reference have not been sufficiently examined.
Page 333
Grasping Intentions Let me turn now to another part of the story. New evidence indicates that during their second year, infants "recognize" the intentions impelling adult acts. They seem to register not just the "surface" properties of another's act, but to impute intentions to it. This is nicely demonstrated by Meltzoff's (1995) stunning finding that young infants imitate not what others do, but what they take them to be intending or trying to do. When an adult's act has been thwarted or blocked by obstacles, infants imitate the "intended act" rather than the thwarted act observed. It is akin to adults observing the famed HeiderSimmel (1944) animated film, unable to resist seeing the gyrating geometric figures as acting "intentionally." There seem, then, to be two kinds of intentions that the infant/toddler seems able to recognize. One is epistemic—recognizing that another is attending to something— some object, event, state. The other is instrumental—recognizing the goaldirected intentions inherent in the acts of others. Combining the two is, I believe, an essential ingredient of meaningmaking—a matter with which we shall be much concerned in what follows. This is because the two come together in a unique way in narrative structuring which, as we shall see presently, is of prime importance in the development of shared reference. Parents and caregivers typically (indeed, almost irresistibly) treat infants as if they had both epistemic and instrumental intentional states—goals, beliefs, desires, feelings, "things in mind." Parents also stoutly believe that infants try to communicate their intentions to them. But they place the burden of proof for interpreting these efforts principally on themselves—perhaps an early version of the SperberWilson (1986) presupposition of relevance. Parents, moreover, are so prone to such beliefs that they even report delight in "discovering" that their infants have "real minds just like everybody else" (Gopnik & Meltzoff, 1997)—as if they were sorely in need of confirmation. Tomasello (1992) has indeed proposed that treating infants as if they had an intentionendowed subjective life may be a crucial condition for their developing one. And the work of SavageRumbaugh and her team (1993) certainly suggests that this attitude is an essential ingredient in the "enculturation" of pygmy chimpanzees. We must not overlook this striking concordance: that human beings are powerfully predisposed to "see" intentional states in the acts of others, and at the same time, to believe that others are similarly predisposed. These predispositions, unique to our species, operate in both the instrumental and epistemic domains: We "see" others as intending to achieve extrinsic goals and as intending to understand the world. And we expect them to do the same by us. In effect, this creates a community held together by a network of interacting intentions. Such a community provides the
Page 334
necessary condition for the development of both culture and language, both in the phylogenetic sense and in the ontogeny of those "born" into the human species. It is this framework, as I make clear in a moment, that underlies the development of reference from its early forms of joint attending to its later full expression in lexico grammatical speech. Take early joint attending first. By my argument, another's directed "gaze" is taken by the young child as a sign that "stands for" an object out there "in the world." When another's line of regard is followed and an object is duly "found," that object becomes the putative referent of that sign. Its "existence'' is verified not only by "finding" the object sought, but also by checking for an appropriate vocal or gestural assent from the party whose line of regard was being followed. The child's primitive interpretant, under the circumstances, seems to be some sort of notion or "theory" of mindintheworld1 along the lines just spelled out: that other minds are tuned to the same world I am tuned to, that what they are tuned to can be inferred from their direction of looking, and that one can confirm whether there is a common focus both "objectively" (by "finding" their target object out there) or "consensually" (by receiving an assenting response from the person whose "mind" one is trying to "read"). Return now to what we described earlier as the infant's two modes of seeing the actions of an adult—as revealing epistemic and/or instrumental intentions. I propose that the first steps on the way to meaningmaking consist of bringing the two forms of apparent intent into concordance with each other. At the outset, the object of communicative interacting between parent and child seems to be either to provide the two with "direct" intersubjective contact or to fill the child's needs. In time, the act of indicating—the child's or the adult's directed to the child—becomes increasingly contextualized (e.g., with the "marking" aspect of the child's early pointing, or in peekaboo [Bruner & Sherwood, 1976] where the child is "playing with" object permanence, or when the mother and infant enact scenes with dolls or puppets that require close specification of actions and protagonists). Indicating also becomes contextualized in a flow of observed instrumental action. The young child sometimes even counts on an adult applying the "principle of relevance" as when the child holds out his or her hand for something whose identity is determined by where the child and adult are in a mutually regulated activity—as in "passing games" (Bruner, 1978). 1
I do not mean to make heavy weather out of the word "theory." I agree with Nelson, Henseler, and Plesa (in press) and Lillard (1998) that the word "theory" begs the question about the kind of cognitive structure required to accomplish the small miracle involved in the act of referring. I do not have strong convictions about the controversy stirred up by the "theorytheory" position of Carey (1985) and Gopnik and Wellman (1994) and I am using "theory" in an asif sense.
Page 335
Indeed, it may well be that one reason why narrative play is so alluring to young children is that it provides opportunities for the child to indicate (often very excitedly) the element of violated expectation in the overall storied action—the toy farmer falling off the tractor, the doll meeting an unexpected outcome, and so on. Recall that these setbacks (Aristotle's old "peripeteia") are what, in adult practice, justify the telling of a story—what gives a story its "sense" (HerrnsteinSmith, 1978; Labov & Waletsky, 1967). You cannot "refer'' to a hero, a villain, a victim without imbedding him or her in the "sensemaking" plot. Underlying all of this, of course, is the interesting interaction, noted earlier, between the instrumental and epistemic, an interaction that lies at the heart of narrative structuring. For referents in a narrative—objects, agents, and the rest—are specifiable by the functions they serve, and these functions are specifiable only in terms of intended actions toward a goal (Bruner, 1983, 1991). What makes sense and reference so inseparable in narrative is that one cannot specify, say, a "hero" or "villain" or even some crucial object without specifying something about their instrumental, intentional role in a primitive plot. It is through these communal, narrativized exchanges that sense and reference come together to provide meaning in a public, culturally relevant way. Sense, as it were, is the cognitive framing for consensual reference. So how do children learn to deal with these complexities? Surely they must need some help from others already adept at it. I suggested some years ago (Bruner, 1983) that aid comes in the form of a Language Acquisition Support System (LASS) that exists in a linguistic community to help children figure out, among other things, what is being singled out in the messy game of reference. I chose that acronym as a contrast to Chomsky's LAD, his innate Language Acquisition Device. For even if grammatical rules were innatebutasleep, to be wakened by overheard utterances, even poorly formed ones, it would still require massive collaborative assistance by adults to get the infant started in the business of figuring out what was being meant by what was being said—what interpretants were needed to form a bridge between a sign and its significate(s). Let me, finally, give a couple of examples of LASS at work in collaborative activities that assure that the expert and novice are paying joint attention to each other's mental contents and, thereby, achieving intersubjective accord toward a world of wouldbe referents. We already know that before the child gets into lexicogrammatical speech he is already adept at signaling and "reading" epistemic and instrumental intentions in others, and that, moreover, his parents and caregivers credit him or her with this gift and act accordingly. I think the crucial problem for the adult is to provide occasions and settings in which it becomes possible for the child to grasp both what is being talked about,
Page 336
in what sense it is being intended, and how it all relates to the setting in which the exchange is occurring. Now to the examples. One of them is scaffolding, the other formatting. Scaffolding (Bruner, 1983; Wood, Bruner, & Ross, 1976) consists of constructing simple, noiseless opportunities for a child to grasp the senseandreference of various signs. The adult uses gestures, facial expressions, signs, and (further along) grammatical constructions that she knows the child already understands, and then frugally adds new sign elements that can easily be understood in the light of these. The trick, as it were, is to reduce the number of degrees of freedom that the child must manage in making sense of the message—reducing almost to the point of redundancy. We do this not only for the young child, but also for nonnative speakers and confused strangers. These same tactics carry over into matching our grammar to the child's grammatical competence, which most of us do exquisitely without even trying, so natural does it seem (McNeill, 1970). Adults do not have to be taught to play their part in LASS. We know full well not to load our utterances with heavy presuppositional loads, tortuous ellipses, burdening anaphora. Formatting operates by rigging the context in which communicating occurs so as, again, to reduce the load on the interpretant. It imbeds scaffolding in routinized, highly recurrent settings (or "formats")—as in childhood mealtime, dressing, and bedtime rituals; "book reading" routines (Ninio & Bruner, 1978); "greeting/farewell"; and "polite request''—formats that provide familiar settings for mother and child to make more easily known to each other what they have "on their minds" and, above all, what is to be accomplished. These recurrent settings establish the situatedness of meanings in particular contexts. Such situatedness, as it were, hands the child a comprehensible context on a silver platter, concretely provides him or her with the interpretant for making meaning of what is being referred to and in what sense. As with scaffolding, formatting reduces degrees of freedom to a minimum. And, by the way, formatting is not restricted to our interactions with children: It is the heart of what is variously and loosely called "on the job training," apprenticeship, "handson experience." After some familiarization, the postal clerk quickly grasps what is going on in the post office, the aspiring cook what is being done and said in the kitchen. Anat Ninio and I were first struck by the power of formatting in a study of early referential indicating or "labeling" (Bruner, 1983; Ninio & Bruner, 1978; Ninio & Snow, 1996).Jonathan's mother was "reading" him a picture book, which, to begin with, is a rather formidable exercise in directing and becoming a partner in a young child's attention. Here is how they proceeded. They were jointly leafing their way through a familiar children's picture book, looking for things to bring to each other's attention and to look at jointly. If the mother chose a new labeling target, she would begin
Page 337
with a characteristic "What's THAT, Jon/a/than?" stressing THAT with a distinctive, elevated intonation, carefully syllabifying her son's name as well. If he replied with a brief, namelike label, she would reply to his version with her usual "That's right, Jonathan, that's a [standard label]," at the same time confirming his utterance while idealizing it into standard English. Over repeated occasions, however, she raised the ante, as if expecting him to get closer to the correct phonological form, and refusing to accept his proffered label until he did. Soon enough, however, they began entering seriously into the sensereference game. She showed him a penny bearing the Queen's image, this time with a "Who's THAT, Jon/a/than?" "Nini," he replied—his name for his grandmother. "No, Jonathan, that's the QUEEN." He insisted, and his interpretant prevailed. She gave in. Once they had hit on a jointly acceptable label, even if he was off phonologically, she would signal on subsequent readings that she knew that he knew by an altered version of "What's that, Jonathan?" this time with ''that" an octave down in pitch and his proper name given without syllabic marking, declaring the achievement of consensually confirmed joint reference. Once a label was stabilized, she began the next step, drawing attention to the properties such a joint object would have to have to be a both senseworthy and verifiably referential. "Yes," she would say to Jonathan's correct label, "and what's the doggie DOING?" with the final word carrying the rising intonation this time. [The pictured dog was conspicuously barking.] Or "What's the doggie EATING?" if the pictured dog were chewing on a bone. Sense and reference were being served together: Here's a dog in a familiar setting, and this is what dogs are verifiably like when we refer to them. Jonathan and his mother both took great pleasure in these daily "reading" sessions. They were more playful than pedagogical, with much laughter, eyetoeye contact, and even a certain amount of "epistemic flirting," as in the episode of deciding whether it was the Queen or Jonathan's grandmother on the coin. Formatting is typical of parents and their young children. But it is also used wherever a complex, symbolically mediated routine needs to be assembled or reassembled—whether in coaching a complex sport or in rehabilitating aphasia. For in acquiring language or any other rulegoverned skill, one needs to grasp the ways in which the elements involved are "situated" in an accessible context. And if one has no reliable representation of the context to fall back on, it helps mightily to have it there perceptually. Let me conclude, then, with these brief remarks. Acts of indicating provide the means for extending individual human attention to an interpersonal, intersubjective level. Such acts would not be possible but for some preadapted or innate capacity to appreciate that our conspecifics share a
Page 338
common world and that one can, even at the very start, determine what in that world is at the focus of another's attention by following their line of regard. Joint attention at the outset seems to be an activity sui generis, for its own sake, and with its own intrinsic reward. From early on as well, infants perceive the behavior of others as instrumental, aimed at achieving ends in that world. They also can signal their own demands to others. Caregivers aid both these processes through scaffolding and formatting. Well before lexicogrammatical speech becomes the main game, the child (and the parent) have learned many routes into each other's mind, and thereby have come to be reassured that they are jointly inhabiting the same world. These routes serve the child well in getting into the language games to follow. References Astington, J. W. (1993). The child's discovery of the mind Cambridge, MA: Harvard University Press. Brazelton, B., & Als, H. (1979). Four early stages in the development of motherinfant interaction. Psychoanalytic Study of the Child, 34, 349369. Bruner, J. (1978). Learning how to do things with words. In J. Bruner & A. Garton (Eds.), Human growth and development: The Wolfson College Lectures 1976 (pp. 6284). Oxford, England: Clarendon Press. Bruner, J. (1983). Child's talk: learning to use language. New York: Norton. Bruner, J. (1991). Acts of meaning. Cambridge, MA: Harvard University Press. Bruner, J., & Sherwood, V. (1976). Early rule structure: The case of 'Peekaboo.' In R. Harre (Ed.), Life sentences: Aspects of the social role of language (pp. 5562). New York: Wiley. Butterworth, G. E., &Jarrett, N. L. M. (1991). What minds have in common is space: Spatial mechanisms serving joint visual attention in infancy. British Journal of Developmental Psychology, 9, 5572. Carey, S. (1985). Conceptual change in childhood Cambridge, MA: MIT Press. Clark, H. (1992). The arenas of language use. Chicago: University of Chicago Press. Crawford, M. (1937). The cooperative solving of problems by young chimpanzees. Comparative Psychology Monographs, 14(Serial No. 68). David, M., & Appell, G. (1961). A study of nursing care and nurseinfant interaction. In B. M. Foss (Ed.), Determinants of infant behavior (Vol. 1). London: Methuen. Feldman, C. (1996). You can't step in the same river twice: Repair and repetition in dialogue. In C. Bazzanella (Ed.), Repetition in dialogue (pp. 3244). Tubingen: Niemeyer. Also Beitrage zur Dialogforschung, 11 [special multilingual issue]. Fernald, A. (1989). Intonation and communicative intent in mothers' speech to infants: Is the melody the message? Child Development, 60, 14971510. Fernald, A. (1991). Prosody in speech to children: Prelinguistic and linguistic functions. Annals of Child Development, 8, 4380. Gopnik, A., & Meltzoff, A. (1997). Words, thoughts, and theories. Cambridge, MA: MIT Press. Gopnik, A., & Wellman, H. (1994). The theory theory. In L. A. Hirschberg & S. A. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture (pp. 120134). Cambridge, England: Cambridge University Press. Heider, F., & Simmel, M. (1944). An experimental study of apparent behaviour. American Journal of Psychology, 57, 243259.
Page 339
HerrnsteinSmith, G. (1978). On the margins of discourse: The relation of literature to language. Chicago: University of Chicago Press. Kasermann, M., & Foppa, K. (1982). Some determinants of selfcorrection: An interactional study of SwissGerman. In W. Deutsch (Ed.), The child's construction of language (pp. 77104). New York: Academic Press. Labov, W., & Waletzky, J. (1967). Narrative analysis: Oral versions of personal experience. In J. Helm (Ed.), Essays on the verbal and visual arts: Proceedings (of the 1966 Annual Spring Meeting of the American Ethnological Society (pp. 1244). Seattle: University of Washington Press. Lillard, A. (1998). Ethnopsychologies: Cultural variations in theories of mind. Psychological Bulletin, 123, 332. McNeill, D. (1970). The acquisition of language. New York: Harper & Row. Meltzoff, A. N. (1995). Understanding the intentions of others: Reenactment of intended acts by 18monthold children. Developmental Psychology, 31, 838850. Moore, C., & Dunham, P. J. (Eds.). (1995). Joint attention: Its origins and role in development. Hillsdale, NJ: Lawrence Erlbaum Associates. Nelson, K., Henseler, S., & Plesa, D. (in press). Entering a community of minds: "Theory of mind" from a feminist viewpoint. In P. Miller (Ed.), Feminist theory and developmental psychology. Ninio, A., & Bruner, J. S. (1978). The achievement and antecedents of labelling. Journal of Child Language, 5, 115. Ninio, A., & Snow, C. E. (1996). Pragmatic development. Boulder, CO: Westview. Peirce, C. S. (19471962). Collected works (Vols. 16). Cambridge, MA: Harvard University Press. SavageRumbaugh, E. S., Murphy,J., Sevcik, R. A., Brakke, K. E., Williams, S. L., & Rumbaugh, D. L. (1993). Language comprehension in ape and child. Monographs of the Society for Research in Child Development, 58(34, Serial No. 233). Scaife, M., & Bruner J. S. (1975). The capacity for joint visual attention in the infant. Nature, 253, 265266. Sperber, D., & Wilson, D. (1986). Relevance: Communication and cognition. Oxford, England: Blackwell. Stechler, G., & Latz, E. (1966). Some observations on attention and arousal in the human infant. Journal of the American Academy of Child Psychiatry, 5, 517 525. Stern D. (1985). The interpersonal world of the infant. New York: Basic Books. Tomasello, M. (1992). The social bases of language acquisition. Social Development, 1, 6787. Wood, D., Bruner, J. S., & Ross, G. (1976). The role of tutoring in problem solving. Journal of Child Psychology and Psychiatry, 17, 89100.
Page 341
AUTHOR INDEX A Achenbach, T. M., 277, 279, 289 Adams, F., 2, 11, 97, 114 Adamson, L., 43, 46, 60, 64, 74, 244, 264 Akhtar, N., 20, 41 Allen, C., 169, 191 Allesandri, S. M., 78, 85, 92, 93 Als, H., 331, 338 Amsel, A., 79, 92 Anand, K. J., 112, 114 Anderson, C., 163, 166 Anscombe, G. E. M., 244, 264, 298, 314, 319, 328 Appell, G., 331, 338 Appleton, M., 9, 11, 232, 240 Arias, I., 284, 289 Armstrong, D. M., 97, 114 Artuso, M., 156, 158, 165 Asendorpf, J. B., 86, 92 Ashmead, D. H., 244, 266 Astington, J., 4, 5, 6, 7, 8, 11, 11, 12, 17, 25, 31, 37, 38, 39, 102, 103, 113, 115, 116, 120, 123, 130, 131, 134, 151, 154, 158, 161, 163, 165, 166, 229, 230, 233, 239, 240, 241, 272, 289, 305, 306, 308, 309, 310, 312, 313, 314, 332, 338 Austin, J. L., 295, 296, 297, 300, 303, 313, 314 Averill, J. R., 273, 289 Azar, S. T. 262, 264 B Baars, B. J., 109, 114 Baillargeon, R., 170, 190, 191 Bakeman, R., 43, 46, 60, 64, 74, 244, 264 Baker, C. I., 201, 224 Baker, S. C., 158, 165 Baldwin, D. A., 19, 23, 38, 46, 60, 210, 221, 223 Baldwin, J. M., 1, 2, 4, 5, 11, 12, 95, 96, 98, 99, 101, 114, 154, 165, 229, 234, 240, 241 Baldwin, M., 273, 274, 289 Banerjee, M., 29, 41 Bargh, J. A., 80, 92 BaronCohen, S., 65, 74, 119, 131, 182, 192, 200, 220, 220, 221, 223, 224 Barresi, J., 44, 45, 46, 60, 61, 65, 74, 148, 151, 154, 160, 165 Barton, M. E., 20, 26, 41 Bartsch, K., 20, 31, 39, 144, 150, 229, 237, 241, 305, 307, 314 Bates, E., 244, 262, 265 Baudonniere, P M., 86, 92 Bauer, W. D., 262, 265 Beardsall, L., 37, 39, 232, 241 Beeghly, M., 275, 289 Bekoff, M., 169, 191 Bem, D. J., 273, 274, 289 Benigni, L., 244, 265 Benson, P, 199, 224 Bentivegna, C., 235, 241 Berg, E. A., 103, 115 Berridge, D. M., 8, 12, 233, 242 Berthenthal, B. I., 89, 91, 92 Bevan, R., 199, 224 Bialystok, E., 113, 114 Bidell, T. R., 102, 115 Bierschwale, D. T., 200, 213, 225 Bill, B., 19, 38 Bíró, S., 28, 39, 65, 74, 176, 192
Page 342
Bjorklund, D. F., 128, 131 Bloom, L., 306, 315 Bonvillian, J. D., 262, 265 Borton, R. W., 190, 193 Boshart, J., 149, 152 Bowerman, M., 113, 114 Bowlby, J., 270, 272, 289 Brakke, K. E., 330, 333, 339 Bratman, M. E., 298, 307, 314 Brazelton, B., 331, 338 Brentano, F., 1, 2, 12, 96, 114, 243, 265, 295, 314 Bretherton, I., 46, 60, 244, 265, 307, 314 Brierly, K., 199, 224 Brody, G. H., 284, 289 Brooks, P., 109, 115, 122, 127, 130, 131, 147, 148, 151, 152 BrooksGunn, J., 85, 86, 90, 92, 93, 116 Brown, J., 9, 12, 37, 39, 232, 233, 234, 235, 239, 241, 306, 314 Brown, R., 304, 305, 314 Bruner, J., 3, 4, 11, 12, 37, 39, 72, 74, 95, 114, 230, 232, 234, 241, 244, 245, 262, 265, 266, 331, 332, 334, 335, 336, 338, 339 Burgess, A., 159, 165 Buriel, R., 258, 266 Burton, R. V., 236, 242 Bushnell, E. W., 244, 266 Butterworth, G., 47, 60, 91, 92, 200, 201, 224, 244, 265, 331, 338 Byrne, R., 171, 183, 192, 194, 197, 220, 224, 225 C Call, J., 69, 72, 73, 74, 75, 183, 189, 194, 201, 225 Camaioni, L., 73, 75, 244, 265 Campos, J. J.. 244, 265, 284, 290 Carey, S., 163, 165, 170, 171, 176, 192, 193, 334, 338 Carlson, S. M., 91, 94, 119, 121, 122, 127, 128, 129, 131, 133, 134, 144, 148, 150, 151 Carpenter, M., 47, 59, 60, 63, 64, 65, 66, 74, 75 Carruthers, P, 98, 112, 114, 182, 192 Carter, A. 97, 108, 117, 120, 122, 125, 132, 145, 152 Carter, S. L., 259, 265 Caryl, P G., 171, 192 Case, R., 272, 289 Caspi, A., 273, 274, 289 Cech, C. G., 220, 213, 225 Chan, S. Q., 258, 266 Chance, M. R. A., 199, 224 Chandler, M., 236, 241 Cheney, D. L., 170, 171, 171, 172, 174, 182, 183, 186, 189, 192 Christensen, A., 284, 289 Cicchetti, D., 85, 93, 275, 289 Clark, H., 70, 74, 330, 338 ClarkeCarter, D., 238, 242 Clements, W. A., 149, 151, 183, 192, 193 Cochran, E., 47, 60, 200, 201, 224, 244, 265 Cohen, L. B., 49, 60, 61, 99, 115 Coie, J. D., 275, 290 Connell, D., 262, 266 Cooper, R., 190, 194 Corbetta, D., 244, 266 Corkum, V., 44, 46, 47, 49, 61, 99, 116, 200, 201, 224 Corley, R., 284, 290 Cosmides, L., 170, 191, 192 Costanzo, P. R., 287, 289 Cottington, E., 259, 266 Craik, F. I. M., 110, 115 Crane, T., 2, 12 Crawford, M., 330, 338 Crittenden, P. M., 262, 265 Crnic, K., 262, 265 Croft, K., 199, 224 Cronbach, L. J., 254, 265 Csibra, G., 28, 39, 65, 74, 176, 192 Cummings, E. M., 284, 289 Cutting, A., 8, 12, 233, 241 D Dale, P., 262, 265 Damon, W., 19, 39 Dantzer, R., 83, 92 Darwin, C., 20, 39, 81, 82, 92, 195, 224 David, M., 331, 338 Davidson, C. A. 213, 225 Davidson, D., 154, 157, 159, 160, 165 Davies, P T., 284, 289 Dawkins, R., 171, 192, 193 DeFries, J., 284, 290 de Villiers, J., 158, 165 de Villiers, P., 158, 165 de Waal, F., 197, 199, 220, 224
Page 343
DeLoache, J., 111, 115 Dempster, F. N., 128, 131 Dennett, D., 77, 78, 80, 92, 157, 164, 165, 169, 192, 243, 245, 265 Desjardins, R. N., 19, 38 Dewey, J., 98, 111, 115 Diamond, A., 128, 131, 134, 145, 151 DiBiase, R., 85, 92 Dishion, T, 283, 290 Dix, T., 259, 265, 287, 289 Dodd, B., 190, 192 Dodge, K. A., 275, 290 Doherty, M. J., 141, 151 Dolan, R. J., 158, 165 Donaldson, M., 312, 314 DonelanMcCall, N., 233, 241 Dore, J., 262, 265 Dorment, R., 325, 328 Doyle, R., 284, 290 Dretske, F., 170, 192 Dunbar, R., 272, 290 Dunham, P., 64, 75, 244, 266, 331, 339 Dunn, J., 8, 9, 10, 12, 13, 37, 39, 230, 231, 232, 233, 234, 235, 238, 239, 241, 242, 245, 271, 290, 306, 307, 314 Dunn, L. M., 255, 265 Duval, S., 80, 82, 92 Dworkin, D. H., 258, 266 E Eddy, T. J., 91, 94, 183, 186, 193, 200, 201, 202, 208, 210, 213, 216, 225 Edwards, P., 319, 328 Eilan, N., 96, 115 Ekman, P., 20, 39 Elder, G. H., 273, 274, 289 Eldridge, N., 248, 265 Eliot, G., 238, 241 Emde, R. N., 244, 265, 284, 290 Emery, N. J., 201, 224 Eron, L. D., 273, 290 Estes, D., 237, 241 Evans, S. L., 259, 265 Everett, B. A., 199, 224 F Feinman, S., 244, 265 Feinman, S., 244, 265 Feldman, C., 3, 7, 11, 234, 241, 300, 324, 328, 330, 338 Feldman, R., 246, 252, 253, 255, 260, 261, 265, 266 Fenson, L., 85, 86, 87, 88, 92, 262, 265 Fernald, A., 23, 40, 331, 338 Fernyhough, C., 238, 242 Field, T. M., 258, 259, 265 Fincham, F. D., 284, 289 Fischer, K. W., 85, 86, 87, 89, 91, 92, 94, 102, 115 Flavell, E. R., 17, 31, 39, 123, 131, 134, 151, 199, 210, 224 Flavell, J. H., 17, 31, 39, 123, 131, 134, 151, 199, 210, 224 Fletcher, P C., 158, 165 Fodor, J., 38, 39, 81, 92, 158, 165, 169, 170, 176, 188, 192 Foppa, K., 330, 339 Forguson, L. W., 297, 315 Frackowiak, R. S. J., 158, 165 Franklin, M. B., 244, 265 Freeman, N. H., 8, 12, 233, 242 Frege, G., 96, 115 Freud, S., 1, 12, 82, 83, 84, 92, 111, 115 Frey, R., 158, 165 Frith, C. D., 158, 165 Frith, U., 158, 165 Fritz, A. S., 236, 241 Frye, D., 3, 7, 10, 12, 13, 43, 60, 61, 66, 74, 96, 97, 99, 102, 106, 108, 109, 110, 115, 117, 119, 120, 121, 122, 125, 127, 128, 130, 131, 132, 133, 134, 135, 136, 137, 138, 143, 145, 146, 147, 148, 151, 152 Fulker, D. W., 284, 290 Furrow, D., 161, 166 G Gale, E., 158, 165 Gallup, G. G., Jr., 91, 92, 220, 224 García Coll, C., 252, 265 Gardner, H., 82, 92 Geach, 304, 315 Gebelt, J., 235, 241 Gee, J., 305, 315 Gelman, R., 171, 176, 190, 192, 194 Gergely, G., 28, 39, 65, 74, 176, 177, 192 Gergen, K. J., 252, 265 Gerstadt, G. L., 134, 151
Page 344
Giambrone, S., 220, 223, 225 Gibson, E., 64, 74 Gilbert, J., 128, 131 Gluckman, A., 72, 75 Goethe, J. W., 95, 115 GoldmanRakic, P. S., 110, 115 Goldsmith, H. H., 284, 290 Golinkoff, R. M., 244, 266 Gomez, J. C., 183, 192 Goodall, J., 192, 224 Goodnow, J. J., 236, 241 Gopnik, A., 2, 3, 4, 5, 6, 9, 11, 12, 13, 17, 18, 20, 25, 31, 34, 35, 36, 37, 38, 38, 39, 40, 46, 61, 68, 75, 95, 102, 113, 115, 119, 123, 131, 134, 137, 148, 152, 154, 161, 162, 165, 182, 190, 192, 210, 224, 333, 334, 338 Gordon, A. C. L., 7, 12 Gottman, J. M., 284, 290 Gouin Decarie, T., 47, 61 Graf, P, 162, 165 Graham, P. J., 288, 290 Gralinski, J. H., 90, 94 Grant, D. A., 103, 115 Grant, V. J., 252, 265 Gray, J. A., 98, 115 Green, F. L., 17, 31, 39, 123, 131, 134, 151 Grice, H., 70, 74 Griffin, D. R., 244, 265 H Hadwin, J., 29, 39 Hala, S., 236, 241 Halford, G., 126, 131 Hall, D. K., 252, 266 Halliday, M. A. K., 302, 315 Hann, D. M., 259, 265 Happe, F., 158, 165 Harding, C. G., 244, 266 Hare, B., 201, 225 Harnishfeger, K. K., 128, 131 Harries, M., 199, 221 Harris, P., 5, 11, 12, 17, 29, 38, 39, 49, 61, 86, 92, 138, 145, 151, 160, 165, 229, 237, 240, 241, 272, 289, 306, 315 Harrison, A. O., 258, 266 Harter, S., 89, 93 Hartl, M., 6, 13 Hartung, J. P., 262, 265 Hauser, M. D., 3, 4, 10, 170, 171, 172, 176, 178, 182, 191, 192, 193 Haviland, J., 235, 241 Heavey, C. L., 284, 239 Heider, F., 27, 39, 176, 193, 333, 338 Hekiman, E., 262, 264 Helfer, R. E., 263, 266 Helwig, C. C., 110, 117 Henseler, S., 334, 339 Herman, J., 310, 315 HernnsteinSmith, G., 335, 339 Heyes, C. M., 182, 183, 193 Hickey, P. R., 112, 114 Hietanen, J., 199, 224 Hix, H. R., 121, 128, 129, 131, 114, 148, 151 Hogrefe, J., 134, 151 Homer, B., 163, 165 Homskaya, E. D., 134, 151 Hong, Y. J., 134, 151 Hood, B., 170, 193 Hood, L., 306, 315 Howard, J. A., 85, 86, 93 Hudson, J. A., 235, 241 Huesmann, L. R., 273, 290 Hughes, C., 119, 128, 131, 133, 134, 143, 144, 149, 151, 232, 233, 239, 241, 242 Hughes, R., 159, 165 Hume, D., 195, 224 Humphrey, N. K., 220, 224 Hutchins, E., 269, 290 I Ilyenkov, E. V., 3, 12 ImbemsBailey, A., 310, 315 Irwin, J.M., 19, 38 J Jacques, S., 96, 97, 98, 100, 103, 104, 108, 109, 113, 115, 116, 117, 121, 129, 130, 131, 132, 146, 152 James, W., 11, 12, 112, 116 Jarrett, N., 200, 201, 224, 331, 338 Jarrold, C., 6, 12, 121, 132, 133, 152 Jaynes, J., 82, 93 Jeannerod, M., 144, 151
Page 345
Jenkins, J., 8, 10, 12, 37, 39, 113, 116, 158, 166, 233, 241, 274, 275, 283, 288, 290 Jenkins, R., 262, 265 Jennings, S., 85, 86, 87, 89, 94 John, R. S., 284, 290 JohnsonLaird, P. N., 273, 274, 290, 301 315 Jolly, A., 220, 224 K Kagan, J., 19, 39, 99, 101, 102, 116, 284 290 Kamawar, D., 10, 11, 157, 162, 166, 304, 306 Kamm, K., 244, 266 KarmiloffSmith, A., 222, 224, 238, 242 Karraker, K. H., 259, 266 Kasermann, M., 330, 339 Kavanaugh, R. D., 86, 92, 138, 151 Kaye, K., 245, 262, 266 Kearsley, R., 101, 117 Kempe, R., 263, 266 Kirkham, N. Z., 108, 116, 129, 130, 131 Knobe, J., 243, 266 Koos, O., 65, 74 Kopp, C. B., 90, 94 Krebs, J. R., 171, 192, 193 Kripke, S., 153, 166 Kruger, A. C., 43, 61, 64, 70, 74, 75 Kuhl, P. K., 190, 193 Kyriakidou, C., 8, 12, 233, 242 L Labov, W., 335, 339 Lamberty, G., 262, 265 Lang, B., 10, 108, 120, 134, 138, 139, 144, 148, 152 Largo, R. H., 85, 97, 93 Latz, E., 331, 339 Lau,A., 110, 117 Lazarus, R. S. 273, 274, 290 Leadbeater, B.J., 9, 13 LeDoux, J. E., 81, 82, 93 Lee, E., 309, 315 Leekam, S. R., 8, 13, 37, 40, 134, 152, 233, 242 Lefkowitz, M., 273, 290 Lempers, J. D., 199, 210, 224 Leonard, E. L., 99, 117 Leslie, A., 7, 12, 68, 74, 91, 93, 100, 116, 138, 149, 151, 171, 176, 182, 193, 237, 242 Levine, L. J., 29, 40 Levinson, S., 70, 75 Lewis, C., 8, 12, 233, 242 Lewis, M., 5, 8, 10, 78, 79, 80, 81, 82, 84, 85, 86, 90, 91, 92, 93, 101, 112, 116 Lillard, A., 334, 339 Lockhart, R. S., 110, 115 Lockman, J. J., 244, 266 Lorincz, E. N., 201, 224 Loveland, K. A., 85, 93 Lowe, M., 86, 88, 93 Lumb, A., 6, 12 Luria, A. R., 3, 12, 103, 113, 116, 129, 131, 134, 151 Luria, Z., 252, 266 LyonsRuth, K., 262, 266 M MacCallum, F., 6, 12 MacWhinney, B., 304, 315 Malatesta, C. Z., 274, 290 Malle, B. F., 243, 266 Mannie, S., 235, 242 Mans, L., 85, 93 Marcovitch, S., 101, 116 Mareschal, D., 156, 166 Margolin, G., 284, 290 MaridakiKassotaki, K., 8, 12, 233, 242 Markman, E. M., 19, 38 Marler, P., 172, 193 Martin, K., 259, 266 Mauthner, N. 119, 120, 125, 128, 132, 133, 143, 148, 152 Mayer, J. M., 273, 291 Mayringer, H., 144, 152 McAdoo, H. P., 262, 265 McCune, L., 86, 88, 89, 93 McEvoy, R. E., 119, 131 McKinney, J. P., 263, 266 McNeill, D., 336, 339 Mead, G., 72, 75 Meins, E., 238, 242 Meldon, A. I., 95, 116 Meltzoff, A. N., 2, 3, 4, 5, 9, 17, 18, 19, 20, 25, 26, 27, 28, 30, 33, 34, 36, 38, 39, 40, 46, 61, 68, 75, 86, 93, 95, 102, 116, 137, 182,
Page 346
190, 192, 193, 210, 224, 333, 338, 339 Menzel, E., 196, 224, 326, 328 Meredith, M. C., 86, 92 Messer, D. J., 244, 266 Michotte A., 176, 193 Miller, G. A., 301, 315 Miller, P H., 17, 39 Miller, R. T, 244, 265 Millikan, R., 170, 193 Mistlin, A., 199, 224 Mitchell, R. W., 171, 193, 197, 224 Moffitt, T. E., 283, 284, 290 Montgomery, D. E.. 304, 305, 315 Moore, C., 3, 6, 10, 11, 12, 44, 45, 46, 47, 49, 60, 61, 64, 65, 66, 68, 74, 75, 99, 116, 148, 151, 154, 160, 161, 165, 166, 200, 201, 224, 238, 242, 244, 266, 331, 339 Moore, M. K., 18, 19, 20, 33, 35, 38, 39, 95, 116 Morgan, R., 65, 75 Morissette, P, 47, 61 Morland, J. K., 252, 267 Moses, L., 6, 12, 17, 23, 25, 31, 38, 39, 40, 46, 60, 121, 128, 129, 131, 144, 148, 151, 221, 223 Mosier, C. E., 244, 266 Mounoud, P., 81, 91, 93 Mumme, D. L., 23, 40 Munn, C., 171, 193 Murphy, D. J., 244, 266 Murphy, J., 330, 333, 339 Murray, L. 56, 61 N Nádasdy, Z., 28, 39, 65, 74, 176, 192 Nagel, T., 98, 116, 160, 166 Nagell, K., 47, 59, 60, 63, 64, 66, 74 Nelson, K., 296, 304, 310, 313, 315, 334, 339 Newell, A., 81, 93 Newton, P E., 236, 242 Nichols, S., 110, 117 Nicolich, L., 86, 88, 93 Ninio, A., 336, 339 Norman, D. A., 144, 151 O O'Brien, M., 284, 290 O'Neill, D. K., 20, 40, 213, 225 Oakes, L. M., 49, 60, 61 Oatley, K., 269, 273, 274, 290 Ogan, T. A., 244, 267 Oldershaw, L., 262, 266 Olson, D., 2, 5, 7, 8, 10, 11, 11, 12, 17, 38, 157, 161, 162, 165, 166, 229, 240, 272, 289, 304, 306, 322, 328 Olweus, D., 278, 224 Oram, M., 199, 201, 224 Ortega, J., 199, 224 Osofsky, J. D., 259, 265 Ozonoff, S., 119, 131 P Pacherie, E., 144, 151 Pachter, L. M., 258, 266 Palfai, T., 96, 110, 115, 119, 121, 122, 127, 131, 133, 134, 135, 136, 138, 147, 151 Pan, B. A., 310, 315 Papousek, H., 78, 94 Patterson, G. R., 283, 290 Peirce, C. S., 329, 330, 339 Pennington, B., 119, 128, 131, 132 Perilloux, H. K., 213, 225 Perner, J., 4, 6, 7, 8, 10, 13, 17, 29, 37, 39, 40, 91, 94, 108, 116, 120, 124, 131, 132, 133, 134, 138, 139, 141, 142, 144, 145, 148, 149, 151, 152, 183, 192, 193, 194, 233, 242 Perret, D. I., 201, 224 Perrett, D., 199, 224 Peskin, J., 157, 166 Pethick, S., 262, 265 Phillips, S., 126, 131 Piaget, J., 4, 7, 13, 66, 69, 75, 80, 85, 91, 94, 95, 99, 101, 111, 116, 301, 315 Pine, C. J., 258, 266 Pinker, S., 170, 191, 193 Piñon, D. E., 103, 104, 117 Pipp, S., 85, 86, 87, 89, 94
Page 347
Plesa, D., 334, 339 Plomin, R., 284, 290 Plotkin, R. C., 262, 266 Pollock, C., 263, 266 Potel, D., 121, 132, 133, 152 Povinelli, D. J., 3, 4, 10, 43, 46, 91, 94, 156, 183, 186, 189, 191, 193, 200, 201, 202, 208, 210, 213, 216, 223, 225 Premack, A. J., 171, 177, 181, 183, 193 Premack, D., 171, 176, 177, 181, 183, 189, 193, 210, 225 Pribram, K. H., 134, 151 Provenzano, R. , 252, 266 Pure, K., 161, 166 Pyers, J., 158, 165 Q Quine, W. V. O., 157, 166 Quinton, D., 288, 290 R Rader, N., 64, 74 Radner, D., 244, 266 Radner, M., 244, 266 Ramadas, J., 163, 166 Ramer, A. L. H., 244, 265 Ramsay, D. S., 5, 8, 10, 85, 86, 87, 88, 92, 93, 101, 112 Rapus, T, 106, 117, 122, 128, 132 Ratner, H. H., 43, 61, 64, 70, 74, 75 Raver, C. C., 9, 13 Reaux, J. E., 202, 213, 225 Reddy, V., 9, 11, 231, 232, 240, 242 Reid, G. M., 252, 266 Reid, J., 283, 290 Reilly, J. S., 262, 265 Repacholi, B. M., 2, 3, 4, 5, 6, 9, 13, 20, 23, 31, 36, 40, 95, 102, 137, 148, 152 Rescorla, R. A., 79, 94 Reznick, J. S., 8, 10, 11, 97, 101, 102, 103, 104, 108, 117, 120, 122, 125, 132, 144, 146, 152, 230, 246, 247, 252, 253, 255, 260, 261, 262, 265, 266, 267, 284, 290 Ricard, M., 47, 61 Ricks, M., 85, 94 Robins, L. N. , 288, 291 Robinson, D. R., 262, 264 Robinson, J., 284, 290 Rochat, P., 65, 75, 91, 94 Rogers, S., 119, 131 Rogoff, B., 244, 266 Roitblat, H. L., 80, 94 Romanes, G. J., 195, 197, 225 Rosaldo, M. Z., 313, 315 Rosen, C. S., 91, 94 Ross, G., 336, 339 Roth, K., 163, 166 Rubin, J., 252, 266 Ruble, D. N., 259, 265 Ruffman, T, 8, 13, 37, 40, 233, 242 Rumbaugh, D. L., 330, 333, 339 Russell, B., 96, 111, 116, 196, 197, 225 Russell, J., 6, 12, 43, 61, 119, 120, 121, 125, 127, 128, 132, 133, 143, 144, 148, 151, 152, 238, 242 Rutter, M., 287, 288, 290, 291 Ryle, G., 296, 315 S Sa, W., 162, 166 Salovey, P., 273, 291 Samuels, M., 109, 115, 122, 127, 130, 131, 147, 148, 151, 152 Santos, L., 182, 183, 186, 193, 194 Sapp, F., 6, 12 Sarda, M., 25, 40, 138, 140, 152 SavageRumbaugh, E. S., 330, 333, 339 Savasir, I., 305, 315 Scaife, M., 244, 266, 331, 339 Schneider, K., 244, 266 Schwartz, B. B., 261, 266 Schwebel, D. C., 91, 94 Searle, J., 4, 6, 7, 13, 21, 33, 40, 77, 79, 81, 82, 94, 99, 102, 103, 116, 124, 132, 158, 166, 243, 266, 298, 299, 312, 315, 319, 328 Sellars, W., 155, 161, 165, 166 Semcesen, T., 108, 116, 129, 130, 131 Sevcik, R. A., 330, 333, 339 Seyfarth, R. M., 170, 171, 172, 174, 182, 183, 186, 189, 192 Shallice, T., 144, 151 Shapiro, B. L., 163, 166 Sharpe, S., 119, 120, 125, 128, 132, 133, 143, 148, 152
Page 348
Sherwood, V., 334, 338 Shipstead, S. G., 199, 224 Shultz, T., 25, 40, 138, 140, 152, 156, 166, 308, 315 Sigel, I., 111, 116 Silk, J. , 174, 192 Simmel, M., 176, 193, 333, 338 Singer, J. L., 91, 94 Skinner, B., 158, 166 Skinner, B. F., 111, 116 Slaughter, V., 6, 12, 17, 20, 31, 37, 39, 40 Slomkowski, C., 9, 12, 37, 39, 232, 233, 241 Smith, M. A., 283, 288, 290 Smith, P. K., 182, 192 Smith, R. S., 284, 291 Snow, C., 304, 310, 315, 336, 339 Sodian, B., 236, 242 Solomon, R., 259, 266 Sommerville, J. A., 110, 117 Spelke, E., 163, 165, 170, 176, 190, 192, 194 Spencer, J. P, 244, 266 Spiker, D., 85, 94 Spinazzola, J., 101, 117 Sroufe, L. A., 85, 93 Stahl, J., 262, 266 Stanger, C., 82, 84, 85, 86, 91, 93 Starkey, R, 190, 194 Stechler, G., 331, 339 Steele, B., 263, 266 Stein, N. L., 29, 40 Stenberg, C. R., 244, 265 Stern, D., 271, 273, 291, 331, 339 Stich, S., 158, 166 Stipek, D. J., 90, 94 Strichartz, A. F., 236, 242 Strosberg, R., 20, 41 Struhsaker, T T, 172, 194 Stummer, S., 10, 108, 120, 138, 139, 141, 142, 148, 152 Sugarman, S., 81, 94 Sullivan, K., 149, 152 Sullivan, M. W., 78, 82, 84, 85, 86, 91, 92, 93 T TagerFlusberg, H., 149, 152 Taylor, M., 17, 40, 91, 94 Tesla, C., 9, 12, 37, 39, 232, 233, 241 Thal, D., 262, 265 Theall, L. A., 202, 213, 225 Thelen, E., 244, 266 Thomas, R. K., 197, 225 Thomas, S., 199, 224 Thompson, C., 148, 151 Thompson, N. S., 197, 224 Tidball, G., 19, 38 Tidswell, T, 119, 120, 125, 128, 132, 133, 143, 148, 152 Tomasello, M., 3, 4, 9, 11, 19, 20, 26, 41, 43, 47, 60, 61, 63, 64, 69, 70, 71, 72, 73, 74, 74, 75, 95, 99, 102, 148, 183, 189, 191, 194, 201, 210, 225, 235, 242, 245, 262, 266, 267, 333, 339 Tomkins, S. S., 274, 291 Tooby, J., 170, 191, 192 Trevarthen, C., 56, 61, 65, 75, 95, 116, 244, 267 Tulving, E., 98, 110, 114, 116 Twentyman, C. T, 262, 264, 267 V Vásquez García, H., 262, 265 Vinden, P., 8, 13 Vinter, A., 91, 93 Vishton, P, 170, 194 Volterra, V., 244, 265 von Hofsten, C., 170, 194 Vygotsky, L. S., 3, 13, 95, 111, 113, 116 W Walden, T. A., 244, 267 Walder, L. O., 273, 290 Waletzky, J., 335, 339 Walters, G. C., 262, 266 Warkentin, V., 86, 92 Wasik, B. H., 252, 265 Watson, M. W., 85, 87, 94 Weiss, M., 82, 84, 85, 86, 91, 93 Wellman, H. M., 6, 13, 17, 20, 29, 30, 31, 39, 41, 46, 61, 119, 131, 144, 150, 161, 165, 229, 237, 241, 305, 307, 314, 334, 338 Wells, D., 25, 40, 138, 140, 152 Wells, G., 304, 310 Welsh, M., 128, 132 Werner, E. E., 284, 291
Page 349
White, A., 239, 242 Whiten, A., 171, 182, 183, 189, 192, 194, 197, 220, 224, 225 Wicklund, R. A., 80, 82, 92 Widmayer, S. M., 258, 259, 265 Willatts, P., 244, 267 Williams, J. E., 252, 267 Williams, S. L., 330, 333, 339 Wilson, A., 274, 290 Wilson, D., 333, 339 Wilson, M. N., 258, 266 Wilson, W. H., 126, 131 Wimmer, H., 4, 6, 13, 91, 94, 120, 124, 132, 133, 134, 144, 151, 152, 183, 194 Wittgenstein, L., 24, 41 Won, D., 23, 40 Wood, D., 336, 339 Woodruff, G., 171, 183, 193 Woodward, A., 49, 61, 68, 75 Woolley, J. D., 29, 30, 41 Y Youngblade, L., 9, 12, 13, 37, 39, 232, 233, 241, 242 Yuill, N., 29, 40 Z ZahnWaxler, C., 284, 290 Zambarano, R. J., 259, 265 Zeedyk, M. S., 246, 248, 249, 264, 267 Zelazo, P. R., 96, 97, 99, 101, 117 Zelazo, P. D., 1, 2, 7, 8, 9, 11, 13, 66, 75, 96, 97, 98, 99, 100, 101, 103, 104, 106, 108, 109, 110, 113, 115, 116, 117, 119, 120, 121, 122, 125, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 140, 145, 146, 147, 148, 151, 152 Zernicke, R. F., 244, 266 Zoll, D., 262, 266
Page 351
SUBJECT INDEX A AnotB error, 145156 Abstraction(s), 84, 90 Abulic dissociations, 106107, 110 Accomodation, 66 Action(s), 95 and intentionality, 95, 96 Action control, 96, 97, 102, 121, 125 Action monitoring, 144 Active intermodal mapping (AIM), 18 Affective expression, 79 Animal studies, 172174, 177179, 83187, 202209, 211213, 214220 vs. human studies, 195, 197, 222223 mental states in primates, 171 seeingasknowing, 202, 205, 208, 209, 210, 214 Appearancereality distinction, 37, 112, 161 Appearancereality tasks, see false beliefs Ascription, 153, 154, 155, 158, 161 acquisition of ability, 155 vs. description, 153, 159 and infant intentionality, 260261, 261 and intention, 163, 164, 245 and language, 159 understanding of, 162 Assimilation, 99 Attachment, 238 Attention, 330, 331 as a mental state, 201 Autism, 331 Autonomy, 258, 259 B Behaviorism, behaviorists, 44, 189 Beliefs, ascription of, 154, 155, 156, 157, 160 about beliefs, 154 children's understanding of, 18, 19 concept of, 154, 155, 157, 160 and surprise, 154, 155 Biases, 127 Brain, and consciousness, 82 and mind, 237 C Child abuse, 262, 275 Childchild interaction, 37, 233, 235, 276 Chimpanzees and communication of intention, 7273 and concepts about mental states, 201, 209, 213, 220 high and low level models, 201, 205, 208, 219 vs. human beings, 209, 220, 222 and social understanding, 73, 196197, 198, 211 Cognitive Complexity and Control theory, 96, 105, 109, 121, 122, 125126, 128, 129, 130, 135143 and theory of mind, 126, 135, 138 Communication, animal, 7273, 170175
Page 352
Complexity, 109, 110, 111, 113, 135, 136, 140 Conscious experience, use of, to understand experience of self and other, 84, 306 what it is like, 98, 160 Conscious intention, 80, 85, 86, 90, 91 Consciousness, and action, 96 intentionality and, 84, 98 minimal, 97, 99 and the psychologist's fallacy, 112 recursive, 100101, 111112 reflective, 104, 108109 of self, 81, 102, 112 Cooperative play, 233, 276, 277 D Deception, 120121, 122, 125, 126, 127, 129130, 144, 147, 157 in animals, 171, 196, 197 children's understanding of, 157 intention in, 236 Desires as intentional states, 3132, 331 understanding of, 24 Developmental psychopathology, 273274 externalizing, 274275, 277, 278, 279, 280, 283, 284, 289 and intention, 270, 282, 288 and interpretation of others' intentions, 274, 288 Differentiation, 237238 Dimensional change card sort (DCCS), 105106, 108, 122124, 126, 128, 129, 134, 135, 136, 138 E Eliminativism, 158 Epistemological development, 162, 163 Emotion judgments, 274275 Emotions, 230 and cognition, 237 conversation about, 232, 233, 235 and goal orientation, 273 and intentionality, 20, 29, 3031, 81, 282 and narrative, 234235 and social context, 238 and understanding of mental states, 20, 22, 238 role of culture in, 239240 individual differences in, 239 Evolution, 189, 191, 198 of social behaviors, 220, 221, 222 of theory of mind, 221 Executive function, vs. automatic control, 144145, 148 competence and performance, 121 and metarepresentation, 7, 120 tasks to assess, 122, 123, 124, 134, 138, 141 and theory of mind, 119120, 121, 23, 125126, 127, 130, 133, 135, 139, 141, 143149 Expectancies, 171, 174, 178, 179, 181, 182, 183, 187, 188 vs. belief, 188 Expectancy violation procedure, 176, 178, 183, 190 F False beliefs, 17, 37, 123, 126, 136, 156 and executive function, 138139, 141, 142, 143, 149 in primates, 187 and rule use, 136137, 138 tests of, 133134, 183, 184185 Feelings; see Emotions Folk psychology, 245 acquisition of, 17, 18, 29, 32 in children, 17 Formatting, 336337 Functionalism, 7, 171 G Gaze monitoring, 19, 46, 47, 64, 331, 332, 334 in chimpanzees, 199, 200, 221, 222 Goalcorrected partnership, 270271, 272, 282, 283 Goals, 269, 270, 275, 276, 278279, 280, 282, 282 H Habituation paradigms, 51, 52, 54, 57, 173, 176, 190191
Page 353
I Imitation, Baldwin and, 2, 5 behavioral reenactment procedure, 20, 2526, 27, 30, 102, 333 as an index of understanding of intention, 18, 20, 25, 26, 3435, 68, 333 and pretend play, 8687 and sharing behavioral states, 35 Inference(s), and observer interpretation, 319 knowledge acquired through, 27, 30, 126 Inhibition, 121, 122123, 128, 129, 134, 144, 145, 146, 147, 148, 149 Innate mechanisms, 170, 197, 237 Instrumentalism, 164165 Intention(s), 2, 57, 21, 44, 243, 269 and action, 79, 82, 320 ascription of, 231, 318319, 244245 and belief, 6, 22, 31, 320, 333 and brain, 82 in communication, 63, 64, 7072, 73, 230231, 304310, 329, 334 in cultural understanding, 72 and nonhuman primates, 7273 tasks to assess, 276 and desire, 6, 21, 30, 3132, 33, 148, 305, 306, 320, 333 development of, 4, 79, 85, 333 directionality and, 2124, 306 expressed in language, 233, 304, 305, 312, 313 developing understanding of, 309313 expressed in pretend play, 234 and interpretation, 275, 287, 317, 318, 319, 322, 325, 326, 327, 330, 333 levels of, 8081, 83, 84, 91 In narratives, 234236, 325326 as propositions, 2122, 312, 320 as representations, 2122, simple, 24, 320 understanding of, 27, 33, 47, 66, 7273, 272, 304, 307, 309310 Intentionaction distinction, 4, 24, 28, 79, 145, 298, 300 development of, 2425, 43 Intentioninaction, 34, 298299 Intentional action, 2, 7, 83, 124, 125, 140, 299, 302, 318, 319, 321, 322, 324 and description, 124 and verb tense, 302303 Intentional agent, 6366, 69, 71 and simulation, 6869 Intentional events, 321, 322 Intentional objects, 97, 99, 101 Intentional relations, 4447 constructivist theory of, 48 and firstperson/third person information, 45, 47, 4849, 5960, 156, 160 and novelty preference paradigm, 4950, 51, 54, 57 Intentional speech, 321, 322, 324, 325 Intentional stance, 7778, 157, 231, 245 Intentional states, 321, 324 direction of causality in, 7 Intentionality, 12, 2122, 44, 96, 229, 243244, 295, 317 ff. contents of, 323, 324, 326, 327 consciousness and, 82, 83, 84, 9697 and culture, 333334 emergence of, 244 and emotion, 80, 83, 84 functional explanation of, 8, 9, and goaldirected behavior, 12, 45, 2428, 64, 65, 67, 70, 80, 81, 82, 9697, 125, 244, 269, 272, 308, 333 parent perception of, 245 ff. and personal meaning, 317, 318, 321 and psychological mode, 322323, 325, 326, 327 and reference, 329, 333, 334, 336 and selfother knowledge, 23, 32, 33 and simple intentions, 317321 structural explanation of, 78, 9 Internal states, communication about, 231, 232, 233, 275, 277, 282, 284 in narratives, 234235 J Joint attention, 46, 6466, 210, 331, 212, 331, 334, 336, 337 as characteristic of intentional agents, 65, 244, 334
Page 354
emergence of, 65, 66 K Knowing, 158, 165, 189190 feeling of, 161 limits of, 160 L Labeling, 95, 100, 336337 Language, and action, 95, 96, 113114. in communication, 335 of internal states, 272273 and consciousness, 111113 and development/understanding of intentional states, 271, 335 knowledge of intention expressed in, 300301 avowed, 303, 310312, 313 implicit, 301302, 304 explicit, 302303, 307310, 313 and selfawareness, see personal pronouns and theory of mind, 158, 282283 Language acquisition, 11, 113, 245 Language Acquisition Support System(LASS), 335, 336 and scaffolding, 336 and formatting, 336 Levels of Consciousness model, 96, 100, 102103, 106, 108, 110111, 113 and executive function, 112113 and language, 111, 113 Longterm memory, 98, 100, 103 M Maturation, 2, 99, 237 Mental states, ability to report on, 229 ascription of, 72, 156157, 182, 272 and behavior, 223 communication and, 282 computational, 169 conversations about, 233 knowledge of, 210, 211 interest in, 231232 intentional, 21, 22, 169 reasoning about, 197 Metacognitive abilities, 82, 91, 158, 307, 288 Metarepresentations, 153 Mindreading, 231, causes for development of, 237, 238 primate, 183 Modularity, 65, 210, 238 Motion, animate vs. inanimate, 28, 29, 3435, 176, 177, 178181, 182 expectancy violation procedure, 176, 178 and inference about goals/beliefs, 177178 Mutual gaze, 331, see also Gaze monitoring; Joint attention N Naive realism, 332 Nativism, 38, 66, 80 Nine to 12month change, 46, 6366, 99 Nonlinguistic expression, of intention, 312, 313 O Object permanence, 190 Ordinary language philosophy, 296299 and intention, 296 P Parentchild interactions, 37 aggression in, 283, 284287 mutual goals in, 271 intentional communication in, 230231 and infant characteristics, 251253, 255259, 263 criteria for attribution of, 248249, 250, 251, 260261 and parent characteristics, 255, 259 tasks to measure, 246247, Perceptions, and knowledge, 155 Perseverative errors, 104, 105, 106, 107, 122, 128, 129 Personal pronouns, and selfrecognition, 8990 Planning, 122, 126, 273, 288, 305 Planfulness, 8
Page 355
and language, 113 Prediction (of behavior), vs. promise, 310311, 313 Prefrontal cortex, 7, 136 Pretend play, 8489, 91, 233234, 335 development of, 86, 138 and false belief, 138 intention expressed in, 85, 91, 234, 335 selfawareness in, 86 selfrecognition in, 8687, 88, 91 Prior intention, 299, 301 Primates, see also Animal studies, chimpanzees alarm calls from, 172 as intentional agents, 6970, 72 selfother identification, 6970 Procedural system, 98 Propositions, 157 and theory of mind, 157 Psychological distance, 111 R Ramp task, 135, 147 Recursion, and consciousness, 110, 112 levels of, 97 Referential communications, 334, 335, 337 development of, 331332, Representation, animal, 189 primate, 175, 182, 189 and reality, 159 understanding of, 141, 145, 149 Representational redescription, 222 Rules, 102, 214215, 216, 219 bias in, 127 embedded/higherorder, 121, 125, 126, 129, 136, 141, 146 inference versus action problems, 126127 and theory of mind, 126, 141 formulation of, 113 and language, 113, 129 perseveration and, 105, 106, 107, 122, 128, 129 Representation of, 107108, 109 selfreferring, 102, 104, 109 Rule use development of, 121, 122, 124 and inhibition, 146 tasks, 103, 105, 122, 123 S Saysomethingdifferent task, 141143 Scaffolding, 336 Schemata, 99, 141142, 145, 146, 147, 271, 274, 287, 325, 326 Seeingknowing distinction, 202, 205, 208, 209, 210, 214 and experience, 210 Selfawareness, 8283, 90 objective, 82, 8990 and pretend play, 85 Subjective, 82, 90 Selfconcept, 102 Selfconsciousness, 80, 81, 101104 and emotion, 84, 85 Selfknowledge, 68, 69 Selfother distinction, 18 Semantic system, 98, 101 Simulation theory, 160 Social cognitive modules, 65 Social context, 269 Social convention, 158 Social referencing, 2223, 24, 46, 64, 244 Social roles, primate, 175 Speech acts, 300301 T Theory of mind, 56, 1718, 37, 46, 9092, 110, 113, 334 theories of, 120 and epistemological development, 163 and executive function, 119120, 126, 133 ff. in primates, 183 social factors in acquisition of, 9, 37 Theorytheory, 161 Triadic interactions, 43, 46, 56, 64 W Wisconsin Card Sorting Task, 134, 136 Working memory, 100101
Page 357
CONTRIBUTORS Janet Wilde Astington, Institute of Child Study, University of Toronto Jerome Bruner, Department of Psychology, New York University Judy Dunn, Social, Genetic and Developmental Psychiatry Research Centre, Institute of Psychiatry, London Carol Fleisher Feldman, Department of Psychology, New York University Douglas Frye, Graduate School of Education, University of Pennsylvania Alison Gopnik, Department of Psychology, University of California, Berkeley Rachel Greenbaum, Institute of Child Study, University of Toronto Marc D. Hauser, Departments of Anthropology and Psychology, Harvard University Jenny Jenkins, Institute of Child Study, University of Toronto Deepthi Kamawar, Centre for Applied Cognitive Science, Ontario Institute for Studies in Education, University of Toronto Birgit Lang, University of Salzburg Michael Lewis, Institute for the Study of Child Development, Robert Wood Johnson Medical School, New Brunswick, NJ Andrew Meltzoff, Department of Psychology, University of Washington Chris Moore, Department of Psychology, Dalhousie University, Halifax, Nova Scotia David R. Olson, Centre for Applied Cognitive Science, Ontario Institute for Studies in Education, University of Toronto Josef Perner, Department of Psychology, University of Salzburg Daniel J. Povinelli, Laboratory of Comparative Behavioral Biology, University of Southwestern Louisiana
Page 358
Douglas Ramsay, Institute for the Study of Child Development, Robert Wood Johnson Medical School, New Brunswick, NJ Betty M. Repacholi, Macquarie University, Sydney, Australia J. Steven Reznick, Department of Psychology, University of North Carolina, Chapel Hill Sandra Stummer, University of Salzburg Michael Tomasello, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany Philip David Zelazo, Department of Psychology, University of Toronto