Cognitive Social Psychology: the Princeton Symposium on the Legacy and Future of Social Cognition

  • 24 940 8
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up

Cognitive Social Psychology: the Princeton Symposium on the Legacy and Future of Social Cognition

cover next page > Cover title: author: publisher: isbn10 | asin: print isbn13: ebook isbn13: language: subject public

2,330 767 3MB

Pages 535 Page size 842 x 1148.48 pts Year 2011

Report DMCA / Copyright

DOWNLOAD FILE

Recommend Papers

File loading please wait...
Citation preview

cover

next page >

Cover

title: author: publisher: isbn10 | asin: print isbn13: ebook isbn13: language: subject publication date: lcc: ddc: subject:

Cognitive Social Psychology : The Princeton Symposium On the Legacy and Future of Social Cognition Moskowitz, Gordon B. Lawrence Erlbaum Associates, Inc. 0805834141 9780805834147 9780585384719 English Social perception--Congresses, Social psychology--Congresses. 2001 HM1041.P75 1998eb 302/.12 Social perception--Congresses, Social psychology--Congresses.

cover

next page >

< previous page

page_i

next page >

page_i

next page >

Page i Cognitive Social Psychology The Princeton Symposium on the Legacy and Future of Social Cognition Edited by Gordon B. Moskowitz Princeton University

< previous page

< previous page

page_ii

next page >

Page ii Copyright © 2001 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 Library of Congress Cataloging-in-Publication Data Princeton Symposium on the Legacy and Future of Social Cognition (1998) Cognitive social psychology : 2001016136 the Princeton Symposium on the Legacy and Future of Social Cognition / edited by Gordon B. Moskowitz. p. cm. Includes bibliographical references and index. ISBN 0-8058-3414-1 (cloth : alk. paper) 1. Social perception—Congresses. 2. Social psychology—Congresses. I. Moskowitz, Gordon B. II. Title. HM1041 .P75 1998 302'.12—dc21 Books published by Lawrence Erlbaum Associates are printed on acid-free 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

< previous page

page_ii

next page >

< previous page

page_iii

next page >

Page iii CONTENTS Part I. The Motives Driving Social Cognition 1.A Relational Approach to Cognition: Shared Experience and Relationship Affirmation in Social 3 Cognition Curtis D. Hardin and Terri D. Conley 2.The Personal Need for Structure and Personal Fear of Invalidity Measures: Historical Perspectives, 19 Current Applications, and Future Directions Megan M. Thompson, Michael E. Naccarato, Kevin C.H. Parker, and Gordon B. Moskowitz 3.Behavioral Discrepancies and the Role of Construal Processes in Cognitive Dissonance 41 Jeff Stone 4.Self-Evaluation: The Development of Sex Differences 59 Eva M. Pomerantz, Jill L. Saxon, and Gwen A. Kenney 5.Evaluating the Self in the Context of Another: The Three-Selves Model of Social Comparison 75 Assimilation and Contrast Hart Blanton 6.Outgroup Favoritism and the Theory of System Justification: A Paradigm for Investigating the 89 Effects of Socioeconomic Success on Stereotype Content John T. Jost Part II. Building Blocks of Social Cognition: Representation and Structure 7.A Case for the Nonconscious Self-Concept John J. Hetts and Brett W. Pelham 8.The Role of Theories in Mental Representations and Their Use in Social Perception: A TheoryBased Approach to Significant-Other Representations and Transference Serena Chen 9.What We Theorize When We Theorize That We Theorize: Examining the “Implicit Theory” Construct From a Cross-Disciplinary Perspective Michael W. Morris, Daniel R. Ames, and Eric D. Knowles 10. From Cognition to Culture: The Origins of Stereotypes that Really Matter Mark Schaller and Lucian Gideon Conway, III 11. The Dynamic Relationship Between Stereotype Efficiency and Mental Representation Jeffrey W. Sherman 12. A Cornerstone for the Science of Interpersonal Behavior? Person Perception and Person Memory, Past, Present, and Future Leonard S. Newman

< previous page

page_iii

105 125 143 163 177 191

next page >

< previous page

page_iv

next page >

Page iv Part III. Subjective Perception and Motivated Judgment 13.On Partitioning the Fundamental Attribution Error: Dispositionalism and the Correspondence Bias Douglas S. Krull 14.Let's Not Forget the Past When We Go to the Future: On Our Knowledge of Knowledge Accessibility Diederik A. Stapel and Willem Koomen 15.Illusory Correlation and Stereotype Formation: Tracing the Arc of Research Over a Quarter Century Steven J. Stroessner and Jason E. Plaks 16.The Other Side of the Story: Transparency Estimation in Social Interaction Jacquie D. Vorauer 17.The Flexible Correction Model: Phenomenology and the Use of Naive Theories in Avoiding or Removing Bias Duane T. Wegener, Meghan Dunn, and Danny Tokusato 18.Exploring the Boundaries of Rationality: A Functional Perspective on Dual-Process Models in Social Psychology Akiva Liberman

211 229 247 261 277 291

Part IV. Control Over Cognition and Action 19.The Crossroads of Affect and Cognition: Counterfactuals as Compensatory Cognition Neal Roese 20.Goals and the Compatibility Principle in Attitudes, Judgment, and Choice C. Miguel Brendl 21.Preconscious Control and Compensatory Cognition Gordon B. Moskowitz 22.Implicit Stereotypes and Prejudice Irene V. Blair 23.Exerting Control Over Prejudiced Responses Margo J. Monteith and Corrine I. Voils

307

References Author Index Subject Index

389 475 497

< previous page

page_iv

317 333 359 375

next page >

< previous page

page_v

next page >

Page v PREFACE On January 8, 9, 10, and 11, Princeton University's psychology department welcomed the year 1998 with a conference that brought together a group of young researchers working in the area of social cognition. The idea for a conference struck me during the first Society of Experimental Social Psychology (SESP) meeting I attended. Although I had been surrounded by an incredible cohort of graduate students during my years at New York University, I was overwhelmed by the number of young researchers who had converged on Washington, DC, for that SESP meeting. My first thought, aside from getting a beer, was to contemplate getting this group that had been scattered across the globe together for a chance to meet as a cohort – both as a social occasion and as an opportunity to discuss where our research as a collective was heading (and forge collaborations with colleagues one might not otherwise have the opportunity to meet). With the financial backing of Princeton University, and the Langfeld fund within the psychology department, I was able to have such an event. Needless to say, without the support of my department at Princeton, this event would not have been able to occur, and each of us who participated is grateful to Princeton for having made this a possibility. Thus, little did my friend and colleague, John Darley, realize what he was initiating when he invited me to be his guest at that SESP meeting. I decided to pattern the conference after two conferences that I experienced only through the edited volumes that emerged from them. One was held at the University of Colorado in 1957 and was entitled Contemporary Approaches to Cognition. During that meeting, the participants (Bruner, Brunswik, Festinger, Heider, Muenzinger, Osgood, and Rapaport) set for themselves the task of outlining a historical introduction to the topic of cognition, detailing each contributor's orientation and recent research, and looking to Fritz Heider to discuss future trends. As they stated, the symposium “undoubtedly reflects the Zeitgeist, but our major aim is to influence it.” The second was held at the University of Western-Ontario in 1978 and brought together researchers who had focused their interests on the cognitive processes involved in person perception and person memory. Borrowing from Heider (1944), they dubbed this thematic concern with the processing aspects of impression formation and social judgment Social Cognition (Volume I of the Ontario Symposium series). Throughout the 1980's such research began its transition from the cutting edges of the discipline to occupying the heart of social psychology. Its entrenchment in the middle of the road was perhaps most clearly marked by the impact of Fiske and Taylor's Social Cognition text and three edited volumes – Wyer and Srull's Handbook of Social Cognition, Sorrentino and Higgins' The Handbook of Motivation and Cognition, and Uleman and Bargh's Unintended Thought . Throughout the 1990s a new crop of social cognition researchers, trained by the contributors to the volumes described earlier, have begun to exert their impact on the direction of the field. In the past 5 to 8 years, these young scientists have established labs of their own and conducted research that will help define the shape of social cognition for the future. However, this group has remained fairly isolated from one another despite working on issues and with methodologies that would greatly inform one another. Although their mentors have shared numerous opportunities to exchange ideas over the years, as of 1997 (when I was plotting the conference) this group had had little chance to gather as a unit and discuss their vision. But the opportunity, once made possible thanks to the Langfeld funds, then presented a problem. Who was I to invite to this conference given that 4 days allowed for a maximum of only 24 talks (at a grueling pace of 6 talks a day, 45-60 minutes each) and 24 participants? This was a dilemma I did not take lightly because, like anyone, I certainly was ill-suited to select from among my peers a handful who could attend. To solve this problem, I attempted a rational approach. I decided that I would like an edited volume to emerge from the conference and that it should attempt to define social cognition somewhat broadly (to include not simply people who study accessibility effects in person perception, but anyone with an interest in the ways in which goals and cognitive processes interact in mediating social responses; e.g., attitudes, impressions, actions, etc.). I started by making a list of topics in social cognition that I would want covered in a comprehensive book. I then made a list of every nontenured faculty person doing social cognition research I could track down and summarized their research interests, with a special focus on where their interests were heading (by virtue of their most recent publications). I then matched people with topics – a process that occupied a chunk of my time while visiting at Tolman Hall in Berkeley in the spring of 1997. I then began a process of selecting people based on the goal of having a group that, as a collective strung together in the proper order, could produce a coherent volume on social cognition. This was, of course, still an imperfect solution, because for

< previous page

page_v

next page >

< previous page

page_vi

next page >

Page vi many categories there were multiple people who could have been invited. In these instances, I tried to balance out the graduate training institutions where folks came from. Certainly there are many people I wish could have attended who were not invited. I admit to having a bias in some instances to inviting a person I knew when there seemed to be more than one equally plausible candidate based solely on research interests; but that was a last resort. Much time and effort went into an attempt at being rational. As we see in several chapters, attempts at rationality often fail, and I cannot claim to be exempt from this fact. This was the process, and it produced an exciting conference spilling over into late-night discussions over extravagant dinners and bottles of wine and beer. Although we had many participants, the aim was to make the conference feel small. It was spread out over many days, with numerous social events built in, and time allotted within each presentation for discussion. It was an amazing and stimulating experience for me, and I hope the excitement of the long weekend is captured by the chapters in this volume. This volume represents the efforts of our conference participants to not simply review their own work, but to place that work in a historical context and to take aim at the future. The volume that has resulted defines social cognition quite broadly, in that it brings together researchers from varied perspectives, but whose interests still cohere along some important dimensions. I am grateful to the contributors for taking this charge seriously and not just producing summaries of the latest results coming out of their labs. The book is organized in a fashion that allows each chapter to build off of each other. Although it is an edited volume, the aim was to allow it to operate as a text in social cognition as well. I want to thank all those who participated in and attended the conference. I especially want to thank the members of the psychology department at Princeton university who chaired the various sessions of the conference and attended the marathon sessions with the interest of a new graduate student. I know the conference participants appreciated their enthusiasm as much as I did, so a special thanks goes to those who chaired the sessions: Joel Cooper, Debbie Prentice, Mike Hogg, John Jemmott, Joan Girgus, John Darley, Dale Miller, and George Miller. There are several people who helped in coordinating and organizing this event who also need to be thanked. Adam Galinsky and Ian Skurnik were active participants in all the sessions as well as being tremendously giving of their time and energy for making sure the conference ran smoothly. I was very lucky to have two such incredible graduate students during my first years as a professor, and everyone in attendance got to observe just how lucky I was. Mary Ann Opperman must be credited for making the conference a reality. She worked on every detail and taught me a great deal. I am indebted to her for making the conference simply perfect (at least from my perspective). I want to also thank those people who have served as my mentors over the years – scholars who have guided me and taught me (by both their actions and inactions) in grad school and in the early years of my post graduate student life in social psychology. It is through the committement of such mentors that the work of all the contributors in this volume is made possible. In my personal case they include John Darley, Marcia Johnson, Joel Cooper, and Dale Miller at Princeton; Peter Gollwitzer, Ute Bayer, Jule Malzacher, and Veronika Brandstätter at the Max Planck institute; John Bargh, Shelley Chaiken, Tory Higgins, Diane Ruble, Jeff Tanaka, Yaacov Trope, and Jim Uleman at NYU; and finally, the students and postdocs who overlapped with me at NYU: Chuck Stangor, Tim Strauman, Wayne Winborne, Roman Thein, David Schechter, Mary Tota, Elizabeth Van Hook, Len Newman, Felicia Pratto, Abigail Panter, Akiva Liberman, Carlos Rivero, Lisa Spielman, Ronda Eisenberg, Tracy Wellens, Orit Tykocinski, Jenny Vookles, Erik Thompson, Chuck Hymes, Bob Roman, Tom Alfieri, Paula Raymond, Roger GinerSorolla, Eun Rhee, Magda Garcia, Eva Pomerantz, Stacey Lutz, Gerd Bohner, Doug Hazlewood, Beate Schuster, and Piroska Hunyadi. Finally, the thanks of all the contributors are extended to the folks at Lawrence Erlbaum for giving voice to our young group. — Gordon Moskowitz Summer 1999

< previous page

page_vi

next page >

< previous page

page_vii

next page >

Page vii CONTRIBUTORS Daniel R. Ames, Columbia University, Department of Psychology, 406 Schermerhorn Hall, MC 5501 1190 Amsterdam Avenue, New York, NY 10027 Irene Blair, Department of Psychology, University of Colorado, Boulder, Colorado, 80309, USA Hart Blanton, Department of Psychology, State University of New York at Albany, Albany, New York, 12222, USA Miguel Brendl, INSEAD, Fontainebleau Cedex, 77305, France Serena Chen, Department of Psychology, Institute of Personality and Social Research, Tolman Hall, University of California, Berkeley, Berkeley, California, 94720-5050, USA Terri D. Conley, Department of Psychology, 1282A Franz Hall, Box 951563, University of California, Los Angeles, Los Angeles, CA 90024-1563 Meghan Dunn, Federal Judicial Center, Research Division, One Columbus Circle, NE Washington, DC 20002 Lucian Gideon Conway, III, Department of Psychology, University of British Columbia, 2136 West Mall, Vancouver, British Columbia, V6T 1Z4, Canada Curtis D. Hardin, Department of Psychology, 1282A Franz Hall, Box 951563, University of California, Los Angeles, Los Angeles, CA 90024-1563 John J. Hetts, Ohio State University, Department of Psychology, 1885 Neil Avenue, Columbus, Ohio, 43210-1222, USA John Jost, Graduate School of Business, Stanford University, Stanford, California, 94305, USA Gwen A. Kenney, Department of Psychology, University of Illinois, 603 E. Daniel St., Champaign, Illinois, 61820, USA Eric D. Knowles, Department of Psychology, Institute of Personality and Social Research, Tolman Hall, University of California, Berkeley, Berkeley, California, 94720-5050, USA Willem Koomen, Social Psychology, University of Amsterdam, Roetersstraat 15, 1018 WB Amsterdam, Netherlands Doug Krull, Department of Psychology, Northern Kentucky University, Highland Heights, Kentucky, 41099, USA Akiva Liberman, National Institute of Justice, 810 7th street, NW, Washington, DC, 20531, USA Margo J. Monteith, Department of Psychology, 115 Kastle Hall, University of Kentucky, Lexington, Kentucky, 405060044, USA Michael W. Morris, Graduate School of Business, Stanford University, Stanford, California, 94305, USA Gordon B. Moskowitz, Department of Psychology, Green Hall, Princeton University, Princeton, New Jersey, 08544, USA Michael E. Naccarato, Backbone Media Inc., 550 Front Street West, Suite 206, Toronto, Ontario, Canada Leonard S. Newman, The University of Illinois at Chicago, Psychology Department, 1007 West Harrison Street, Chicago, Illinois, 60607-7137, USA Kevin C.H. Parker, Clinical Psychology Department, Kingston General Hospital, Kingston, Ontario, Canada

< previous page

page_vii

next page >

< previous page

page_viii

next page >

Page viii Brett W. Pelham, 356 Park Hall, Department of Psychology, State University of New York at Buffalo, Buffalo, New York, 14260, USA Jason E. Plaks, Department of Psychology, Schermerhorn Hall, Columbia University, New York, New York, 10027, USA Eva M. Pomerantz, Department of Psychology, University of Illinois, 603 E. Daniel St., Champaign, Illinois, 61820, USA Neal Roese, Department of Psychology, Simon Fraser University, Burnaby, British Columbia, V5A 1S6, Canada Jill L. Saxon, Department of Psychology, University of Illinois, 603 E. Daniel St., Champaign, Illinois, 61820, USA Mark Schaller, Department of Psychology, University of British Columbia, 2136 West Mall, Vancouver, British Columbia, V6T 1Z4, Canada Jeffrey W. Sherman, Department of Psychology, Northwestern University, 2029 Sheridan Rd., Evanston, Illinois, 60208-2710, USA Diederik A. Stapel, Social and Organizational Psychology, University of Groningen, Grote Kruisstraat 2/1, 9712 TS Groningen, Netherlands Jeff Stone, Psychology Department, University of Arizona, Tucson, Arizona, 85721, USA Steven J. Stroessner, Department of Psychology, Barnard College, Columbia University, 3009 Broadway, New York, New York, 10027-6598, USA Megan M. Thompson, Command Group, Defence and Civil Institute of Environmental Medicine, 1133 Sheppard Ave., West, P.O. Box 2000, Toronto, Ontario, M3M 3B9, Canada Danny Tokusato, Department of Psychological Sciences, Purdue University, West Lafayette, Indiana, 47907-1364, USA Corrine I. Voils, Department of Psychology, 115 Kastle Hall, University of Kentucky, Lexington, Kentucky, 405060044, USA Jacquie D. Vorauer, Department of Psychology, Duff Roblin Building, University of Manitoba, Winnipeg, Manitoba, R3T 2N2, Canada Duane T. Wegener, Department of Psychological Sciences, Purdue University, West Lafayette, Indiana, 47907-1364, USA

< previous page

page_viii

next page >

< previous page

page_1

next page >

page_1

next page >

Page 1 Part I. The Motives Driving Social Cognition

< previous page

< previous page

page_2

next page >

page_2

next page >

Page 2 This page intentionally left blank.

< previous page

< previous page

page_3

next page >

Page 3 1 A Relational Approach to Cognition: Shared Experience and Relationship Affirmation in Social Cognition Curtis D. Hardin Terri D. Conley University of California, Los Angeles Social psychology is poised to realize the synthesis envisioned in the ‘‘The Sovereignty of Social Cognition” by the late Thomas Ostrom (1984), who defined social cognition as “the struggle to understand the interdependence between cognition and social behavior.” In so doing, Ostrom broke from contemporary definitions of social cognition, which in the 1960s and 1970s was defined as the study of cognition about social objects and since the early 1980s has been defined as the study of the cognitive bases of social perception and behavior (cf. Fiske & Taylor, 1991; Heider, 1958; Kelley, 1967). Although Ostrom celebrated the identification of cognitive foundations of social behavior, he lamented the lack of research on social foundations of cognition. However, it can no longer be said that the social bases of cognition have been ignored. Indeed the recent decade has witnessed a blossoming interest in how social relationships affect even very basic information processing (e.g., Baldwin & Sinclair, 1996; Baumeister & Leary, 1995; Higgins, 1992; Jost & Banaji, 1994; Leary, 1995; Markus & Kitayama 1991; Schwarz, 1994; Steele & Aronson, 1995; Tice, 1992). We argue that pursuing theoretical integrations of cognitive and social activity will yield important new insights into the hallmark issues of social psychology, including attitudes, social perception, the self-concept, and stereotyping. To do so, we first identify in the history of social cognition research two fundamental human requirements – social connectedness and cognitive understanding – and argue that a full understanding of social-cognitive interdependence requires a renewed focus on how social interaction structures basic information processing. We propose that shared reality theory provides one such synthesis from its postulate that both relational and epistemic requirements are served by the interpersonal realization of shared experience (Hardin & Higgins, 1996). Moreover, research demonstrating the role of shared reality processes in the regulation of interpersonal behavior and individual cognition provides tentative promise for Ostrom's prescriptive charge for the social cognition endeavor. FOUNDATIONS OF SOCIAL COGNITION: RELATIONSHIPS AND EPISTEMICS Twisting through the history of research on social cognition, like the frayed strands of a double helix, is the observation of two great forces driving human behavior: One is the requirement to establish, affirm, and protect social relationships, and the other is the requirement to understand the self and its environments ( inter alia., Asch, 1952; Festinger, 1954a; Freud, 1922/1989; Heider, 1958; Higgins, 1981a; James, 1890/1950; Lewin, 1931; Mead, 1934; Newcomb, 1953; Schachter, 1959; Sherif, 1936; Sullivan, 1953). Interestingly, for the most part, the contemporary literatures on the pursuit and consequences of the relational and epistemic needs have evolved independently. For example, contemporary social cognition research has been characterized by a near exclusive focus on epistemics (see Thompson, Naccarato, Parker, & Moskowitz, chap. 2, this volume, for a review) – that is, how the cognitive system enables individuals to understand and thereby adaptively navigate an informationally complex world (e.g., Bargh, 1996; Higgins & Bargh, 1987; Markus & Wurf, 1986; Stangor & Lange, 1994; Swann, 1990). Meanwhile, research on the negotiation and maintenance of social relationships has occurred on the selfdescribed margins of mainstream social psychology (e.g., Baumeister & Leary, 1995; Collins, 1997; Hazan & Shaver, 1987; Murray & Holmes, 1997). This was not always the case, however, for the American psychological tradition arose out of the modern conception of human nature, including the assumption that individual thought is a product of social activity (e.g., Dewey, 1922/1930; Freud, 1922/1989; James, 1890/1950; Marx & Engels, 1846/1970; Wittgenstein, 1953). Indeed, with the benefit of hindsight, it

< previous page

page_3

next page >

< previous page

page_4

next page >

Page 4 appears that the theoretical alienation between social behavior and cognition is a relatively recent (and perhaps temporary) development. An expressed focus on their interdependence characterizes classical social-psychological theory, and recently contemporary theorists have reaffirmed an interest in social-cognitive interdependence. Epistemics and Relationships in Classical Psychology At its genesis, American psychology accepted as axiomatic the proposition that cognition and social activity are fundamentally interdependent. For example, William James argued that human knowledge is not an individual end, but emerges from an ongoing process of adaptive, cooperative, social activity (e.g., James, 1890/1950, 1907/1992, 1909). Indeed, James' postulate that good thinking is a collective activity (rather than a solitary activity) is one of the defining elements of the pragmatic philosophical tradition (e.g., Dewey, 1922/1930; Mead, 1929). In so doing, James broke radically from classical philosophy by identifying the search and validation of “truth” in practical, social terms rather than in the ephemera of platonic ideals: True ideas are those that we can assimilate, validate, corroborate and verify. False ideas are those that we can not . … Truth lives, in fact, for the most part on a credit system. Our thoughts and beliefs ‘pass,’ so long as nothing challenges them, just as bank-notes pass so long as nobody refuses them. But this all points to direct face-to-face verifications somewhere, without which the fabric of truth collapses like a financial system with no cash-basis whatever. You accept my verification of one thing, I yours of another. We trade on each other's truth. But beliefs verified concretely by somebody are the posts of the whole superstructure. (James, 1907, pp. 100–101, italics in the original) The social behaviorists built on James' insight with descriptions of ways in which the individual's cognitive world is constructed and regulated by social structure and interpersonal interaction (e.g., Berger & Luckmann, 1966; Cooley, 1902/1983; Dewey, 1922/1930; Mead, 1934; Stryker & Statham, 1985). In this tradition, the social-cognitive synthesis is realized by taking as axiomatic the necessity of social relationships for human survival and postulating that adaptive cognitive structure emerges from the internalization of social organization. For example, Cooley (1902) made a foil the individual solipsism of Emerson and others to emphasize instead that individual understanding is predicated on the internalization of the social world: A castaway who should be unable to retain his imaginative hold upon human society might conceivably live the life of an intelligent animal, exercising his mind upon the natural conditions about him, but his distinctively human faculties would certainly be lost, or in abeyance. George Herbert Mead articulated a mechanism by which society might be internalized, and in so doing argued that human cognition reflects an ever dynamic interplay among socially shared perspectives (Mead, 1930, 1932, 1934, 1938, 1956). Mead emphasized in particular the role of ongoing cooperative social interaction in the creation and maintenance of meaning. Mead's synthesis of self and society was also predicated on the necessity of human relationships for survival, but through the mechanism of the organized exchange of social gestures and concomitant perspective taking. Mead likened everyday social interaction to rule-governed play in organized games like baseball, in which each participant must continually take the various perspectives of each of the other participants to understand its place, responsibilities, and plans for action. For example, a catcher in baseball must take the perspectives of the batter, shortstop, outfielder, and pitcher (among others) to understand itself and the everevolving reality of the game. The cornerstone of Mead's theory is the concept of the significant symbol , on which rests the human capacity for language and cognitive representation. The significant symbol is realized when an individual evokes in itself, by a social gesture, the functionally identical response the gesture evokes in others. According to Mead, it is only through significant symbols, realized in ongoing social interaction, that people are able to remember the past, anticipate the future, and experience the present. From this perspective, the mind cannot be separated from cooperative social action, but instead arises out of it. Notably, Mead's objective relativism broke radically from the introspectionists and phenomenologists – and, indeed, from many varieties of subjective relativism popular among poststructural theorists of today – by postulating that the individual mind can exist only in relation to other minds with the capacity to share experiences. The social expression and validation of an idea is essential for both its clarity and objectivity, ultimately relieving individual experience of mere solipsism. One of the most important implications of Mead's theory is that, through significant symbols, individual experiences are socially constructed, existing only in relation to the organized social activities that support them, thereby creating meanings out of what would otherwise be an undifferentiated flux of stimuli

< previous page

page_4

next page >

< previous page

page_5

next page >

Page 5 (cf. Whorf, 1956; Wittgenstein, 1962). Yet because such cognitions are assumed to be linked directly to the shared perspectives of real people in actual social exchange, Mead's social-cognitive theory is a thoroughly material account of mind, and hence congenial to the scientific endeavor (Jost & Hardin, 1997). While theorists in the Jamesian tradition articulated a social-cognitive synthesis in which epistemics was the primary adaptive consequence of organized social interaction, a complementary synthesis was being articulated in the tradition of psychoanalysis in which social connectedness – and concommitant psychic integrity –was the primary adaptive outcome. Freud and his students postulated that psychological structure is not only grounded in family dynamics, but ultimately motivated by the human need to feel connected with significant others (e.g., Erikson, 1959; Freud, 1922/1989; Jung, 1934; Sullivan, 1953). Like the American pragmatists, Freud (1922/1989) grounded his theory in the necessity of society for survival, identifying separation anxiety as the most primitive human fear. For Freud, the primacy of social relationships renders connectedness an essential human motivation achieved through social identification, in which one attempts to take on the characteristics, desires, and behaviors of the object of identification. The primary means by which identification occurs is the attempt to understand others by taking their perspectives as well as attempting to elicit understanding from them. Hence, Freud proposed that social identification is the process by which humans regulate all important social attachments – a notion central to his theorizing. For example, in discussing his theory of psychological structure, Freud (1933/1965) wrote: The basis of this process is what is called an “identification” – that is to say, the assimilation of one ego to another one, as a result of which the first ego behaves like the second in certain respects, imitates it and in a sense takes it up into itself. …It is a very important form of attachment to someone else (p. 56). Although Freud was the first to emphasize the primacy of social identification in cognition, his students translated the insight into a language that may have better communicated the social character of healthy individual cognition. For example, each of Erikson's (1959) stages of human development involve social validation of one kind or another for successful resolution. Jung (1934) emphasized the necessity of individual consciousness to find resolution with shared cultural values. Bowlby (1969) wrote that ‘‘the young child's hunger for his mother's love and presence is as great as his hunger for food.” Indeed, Sullivan's (1953) interpersonal theory of psychology articulated psychodynamic theory in a way that resonates unmistakably with Mead's (1934) emphasis on perspective taking in cognition: By the end of childhood, the pressure toward socialization has almost invariably fixed a big premium on carefully sorting out that which is capable of being agreed to by the authority figure. This is the first very vivid manifestation in life of the role of consensual validation, by which I mean that a consensus can be established with someone else. (Sullivan, 1953, p. 224) Hence, two great traditions of classical psychology – Jamesian social behaviorism and Freudian psychodynamics – not only emphasized the interdependence of social and cognitive activity, but converged on a similar social-cognitive synthesis. Yet each synthesis had a distinct flavor. Although both emphasized the essential role of perspective taking in cognition, the social behaviorists focused most on its epistemic functions, whereas the psychodynamic theorists focused most on its relational functions.1 As discussed later, shared reality theory represents an attempt to integrate these two motivations within a single social-cognitive framework. Relationships and Epistemics in Classical Social Psychology Early social-psychological theorists also began with the social-cognitive assumption of interdependence, but did so from a perspective borne of a world ravaged by revolution, war, and genocide (e.g., Adorno, Frenkel-Brunswik, Levinson, & Sanford, 1950; Asch, 1952; Heider, 1958; Le Bon, 1896; Lewin, 1935; Sherif, 1936). Although the epoch produced a social psychology defined by an interest in the social foundations of knowledge, it represented a kind of neo-rationalism in which the power of society to structure individual thought was not merely acknowledged but conceived as a threat to individual rationality (cf. DeCartes, 1647/1970; Russell, 1912). Hence, both the relational and epistemic motivational strands of social cognition dominant in classical psychology survived, although most American social-psychological theorists viewed the two motivations as essentially binary and competing (for critiques, see Baumeister, 1987; Geertz, 1973; Fiske et al., 1998; Markus & Kitayama, 1994; Triandis, 1995; Turner, Hogg, Oakes, Reicher, & Wetherall, 1987). With few exceptions since, social psychologists have assumed that cognitive integrity is defined by the degree to which individuals are able to resist social influence, whether it be interpersonal, group based, or imputed through more broadly societal or cultural means. Consequently, social psychologists

< previous page

page_5

next page >

< previous page

page_6

next page >

Page 6 have studied social influence in terms of its power to delude the individual, as evidenced, for example, by the pejorative labels used to characterize it, including conformity (e.g., Milgrim, 1961), groupthink (e.g., Janis, 1982), deindividualization (Zimbardo, 1962), and social loafing (e.g., Latane & Darley, 1969) – and, indeed, by the fact that the preeminent journal of the fledgling discipline was entitled Journal of Abnormal and Social Psychology. Some contemporary theorists have criticized this perspective as peculiarly “Western” (e.g., Fiske et al., 1998; Markus & Kitayama, 1991; Triandis, 1995), characterizing it as an individualistic conception of personhood arising from cultural prescriptions for people to be unique, free, and unbounded by social relationships. However, we find it notable that proponents of intellectual individualism expressly assumed that their prescriptions were neither endorsed nor practiced by people in general (e.g., Descartes, 1647/1970; Nietzche, 1954/1977; Russell, 1912; Sartre, 1933). Moreover, intellectual individualism appears to resonate with the broader ideology of the JudeoChristian canon, which views individual resistance to worldly social influence as impossible but laudable (e.g., Bloom, 1992). From our perspective, however, it is notable that resistance to worldly influence in the Judeo-Christian tradition does not occur in a social vacuum, but rather is enacted to affirm one's relationship with God as well as the earthly community of believers. Finally, although social psychology is fairly characterized as subscribing to a paradigm in which the otherwise rational individual is pitted against irrational social influences, this perspective has produced with more clarity than any system preceding it demonstrations that people value social connection more than anything else – except, perhaps, air to breath, food to eat, and shelter from the elements (e.g., Bowlby, 1969). Nevertheless, in some instances, social-psychological theorists expressly recognized the adaptive interdependence between social influence and individual cognition by investigating the role of relationship concerns in the basic structure of thought. For example, Vygotsky (1962, 1978) turned Piagetian theory on its head in research demonstrating the social foundations of cognitive development. Lewin (1931, 1951) proposed that social influences form the forces at the very foundation of psychological structure and the psychological situation. Sherif (1936) assumed the generally adaptive function of the social motivation in cognition by demonstrating the role of group dynamics in the construction of coherent perception out of ambiguous sensation. Moreover, Sherif recognized that social interaction not only formed the basis for individual understanding, but also that shared understandings have social regulatory functions: “Established social values are standardized fixations which the individual incorporates in himself and which henceforth have a great deal to do with regulating his likes and dislikes, his closeness to or remoteness from other individuals, and his activities in satisfying his basic needs” (p. 125). Asch's (1951, 1956) research demonstrated that otherwise simple and unmistakable judgments created severe discomfort and even uncertainty among individuals faced with the incorrect but unanimous judgments of others. Although ironically Asch's work has been assimilated within the rubric of conformity, which is where in contemporary textbooks one now invariably finds discussions of his famous line experiments, he rejected the rationalist paradigm pitting social influence against adaptive cognitive functioning. Instead, Asch concluded that adaptive human understanding is predicated on social influence and shared experience: If we were unable to come to agreement about [our] surroundings, sensible interaction would lack all foundation. To the extent that we do so and to the extent that we can make known our experiences, the relations of identity or similarity of our perceptions become the condition for mutual action. These facts become also the intelligible basis for relations of difference in our experiences. …For this reason I am able to enter into relations with others. (Asch, 1950, pp. 128–129) Asch not only recognized the epistemic functions of socially shared cognition, but, like Sherif, emphasized its role in regulating relationships: That attitudes have such social roots and implications has consequences for their cognitive and emotional functioning, for the conditions of their growth and change. Their content and their persistence and change must be seen as an expression of the need to maintain viable group-relations. Only in this way can we fully understand the pull of social conditions in the formation and modification of attitudes and the fact that they vary lawfully with group membership. …For a Southerner to deny the prevailing views about Negroes requires a drastic intellectual reorientation and a serious snapping of social bonds. It would be tantamount to questioning the perceptions and cherished values of those nearest to him and of casting himself out of the group. (1950, pp. 577–578) Follow-up research employing Asch's (1951) line paradigm supported both the epistemic and relational functions of social consensus by demonstrating that conformity increases with the difficulty and ambiguity of the

< previous page

page_6

next page >

< previous page

page_7

next page >

Page 7 judgment task (e.g., Asch, 1951; Deutsch & Gerard, 1955), as well as by group attractiveness, cohesiveness, and interdependence (e.g., Back, 1951; Deutsch & Gerard, 1955; Gerard, 1954; Jones, Wells, & Torrey, 1958; Thibaut & Strickland, 1956). The repeated observation of the potency of epistemic and relational needs eventually elicited theoretical integrations. For example, inspired by Gestalt principles of perceptual coherence, Heider's (1946, 1958) balance theory described the social consequences of intrapersonal and interpersonal belief symmetry and asymmetry, postulating synthesis through the mechanism of cognitive consistency. According to balance theory, if attitudes and/or others are perceived to belong together in unit relation, then there is a pressure toward coherence within the unit relation. For example, people have more stable relationships with others who like their friends and dislike their enemies. Unbalanced states are unstable because they are met with attempts to restore balance by either (a) changing one's own attitude or the attitude of one's relationship partner, (b) misapprehending the attitude of oneself or one's partner, or (c) changing one's attitude about one's partner (e.g., Monsour, Betty, & Kurzwiel, 1993). Festinger's (1954a) social comparison theory provided yet another synthesis of relational and epistemic needs in a framework that continues to be influential today (e.g., Taylor & Lobel, 1989). Social comparison theory postulates that human gregariousness serves epistemic needs by facilitating the use of social standards in (self-) understanding. According to the theory, social comparison processes afford understanding of both “what is real” and “what is good,” each having different implications for social motivation and action. The goal of understanding what is real is facilitated by social comparisons with similar others because they are assumed to be in a better position to provide consensual validation that one's opinions are accurate. Hence, particularly in the case of opinions, group members will attempt to reduce discrepancies to the extent that they exist, thereby increasing uniformity within the group. However, particularly in the case of abilities, people attempt to acquire more and more of what is good, which motivates movement upward (and away) from the group norm. By the end of the 1950s, then, social-psychological research was rife with demonstrations of the power of the society to constrain beliefs and attitudes, particularly through attempts to establish, protect, affirm, and maintain social relationships. Yet Schachter's (1959) review of this literature led him to conclude: Despite the importance of the study of affiliative needs, almost nothing is known of the variables and conditions affecting these needs. We have no precise idea of the circumstances that drive men either to seek one another out or to crave privacy, and we have only the vaguest and most obvious sort of suggestions concerning the kinds of satisfaction that men seek in company. (p. 1) Although Schachter did not forward a theory of affiliation, he identified some conditions likely to produce affiliative behavior. Affiliation with others of similar experience is more likely under conditions of uncertainty and anxiety (e.g., Schachter, 1951, 1959) as well as with ingroup members when group beliefs are attacked (e.g., Festinger, Reiken, & Schachter, 1956). Moreover, when affiliation is delayed or precluded, individuals may even exhibit symptoms of insanity (e.g., Faris, 1934). Indeed, Schachter (1959) reported his own attempt to study the psychological effects of social isolation, but aborted the research when he found in a preliminary study that just three of five participants exhibited severe symptomology in 3 days, concluding that such research was economically impractical because it would require 10 to 12 days to reliably produce psychotic symptoms. In summary, classical social psychological research not only proceeded on the assumption of social-cognitive interdependence, but repeatedly demonstrated the operation of epistemic and relational motivations. The era produced at least two attempts to synthesize them within single theoretical frameworks, social comparison theory (Festinger, 1954a) and balance theory (Heider, 1958). After briefly outlining the presence of these themes in contemporary social cognition research, we describe our attempt to integrate them within the rubric of shared reality theory. Relationships and Epistemics in Contemporary Social Psychology Although the cognitive side of social cognition has dominated research for several decades – culminating in the 1980s definition of social cognition as the study of the cognitive bases of social behavior (e.g., Fiske & Taylor, 1984, 1992; Higgins & Bargh, 1987) – recent years have witnessed a decided resurgence in research on relational needs as well (e.g., Baumeister & Leary, 1995; Leary, 1995; Murray & Holmes, 1997). This emerging interest bridges the classical interest in relational demands to the contemporary emphasis on epistemics in ways that afford syntheses in the language of contemporary social cognition theory. In short, research shows that relationships are pursued with unusual purpose, and have profound emotional and cognitive consequences. For example, it is now well accepted that forming social relationships is

< previous page

page_7

next page >

< previous page

page_8

next page >

Page 8 easy and breaking them is difficult. Social attachments are formed in infancy (e.g., Bowlby, 1969) as well as under conditions in which history, interdependence, or instrumental interests are absent (e.g., Billig & Tajfel, 1973; Brewer, 1979; Tajfel, 1981). Mere proximity can produce stable social attachment (e.g., Festinger, Schachter, & Back, 1950) under both positive conditions and even extremely negative conditions (e.g., Elder & Clipp, 1988). A review of this kind of research suggests that shared emotions are crucial to the development of interpersonal relationships (Moreland, 1987). Moreover, research suggests that, once formed, social relationships are difficult to escape. Even participants in temporary groups deny that relationships within the group will end when the group dispenses (e.g., Lacoursiere, 1980; Lieberman, Yalom, Miles, 1973). Remarkably, people have difficulty escaping even abusive relationships (e.g., Roy, 1977; Strube, 1988). Relationship status has substantial emotional consequences. For example, young children exhibit extreme distress when separated from their caregivers (Bowlby, 1969, 1973). Recollection of social rejection elicits anxiety, and imagining social rejection increases physiological arousal (Craighead, Kimball, & Rehak, 1979). People feel happier, better off, and less depressed when they are in a network of close relationships than when they are socially isolated (e.g., Argyle, 1987). Indeed, the prospect of losing important relationships elicits anxiety and depression, and severed relationships elicit loneliness and grief (e.g., Leary, 1990). Interestingly, loneliness is related more to lack of intimate social relationships than lack of social contact per se (Reis, 1990; Wheeler, Reis, & Nezlek, 1983; Williams & Solano, 1983). Although severe anxiety is elicited by social exclusion, anxiety is relieved by social inclusion (Barden, Garber, Leiman, Ford, & Masters, 1985). A review of this literature led Baumeister and Tice (1990) to conclude that social exclusion is the single most important cause of anxiety. Finally, research suggests that the pursuit of social relationships has important cognitive consequences. For example, compared with standard attributional categories of locus, stability , and controllability , S. M. Anderson (1991) found evidence that the strongest attributional dimension was interpersonalness, reflecting the degree to which participants attributed causes of their behavior to their relationship status (e.g., “Because I'm married’’). People attribute more positive characteristics and fewer negative characteristics to ingroup members than out-group members (e.g., Forsyth & Schlenker, 1977; Howard & Rothbart, 1980; Leary & Forsyth, 1987; Zander, 1971). Partner-serving biases are as strong as self-serving attributional biases among members of happy marriages but not among those in unhappy marriages (e.g., Fincham, Beach, & Baucom, 1987; Murray & Holmes, 1997). Moreover, research suggests that relationship orientation appears to affect self-representation. For example, social relationships have been implicated in self-evaluation (Baldwin, Carrell, & Lopez, 1990), including vulnerability to negative stereotypes (Steele & Aronson, 1995). Close relationships appear to be bound up with self-representations (e.g., Andersen & Baum, 1994; Aron, Aron, Tudor, & Nelson, 1991; Hardin & Higgins, 1996), In summary, contemporary social-psychological research has not only made progress in identifying social consequences of basic information processing, but has corroborated the importance of relational pursuits identified in classical research. Moreover, the emerging identification of information-processing consequences of affiliation make the time ripe for new attempts to provide theoretical syntheses of epistemic and relational aspects of social cognition. SHARED REALITY THEORY: RELATIONSHIP AFFIRMATION IN COGNITION We have drawn liberally on the insights of the social cognition tradition by taking as axiomatic human epistemic and relational needs and integrating them in the context of contemporary communication theory. In particular, shared reality theory synthesizes epistemic and relational needs through a single social-psychological mechanism: the perceived achievement of mutual understanding, working intersubjectivity, or what we term shared reality (Hardin & Higgins, 1996). Put simply, shared experience links specific interpersonal relationships to specific cognitions, thereby simultaneously binding social relationships and maintaining the individual's grasp of a dynamic world. Predicated on the assumption that epistemic and relational needs are realized through the achievement of mutual perceptions of shared experience, shared reality theory postulates two fundamental axioms from which several useful predictions may be derived (illustrated in Fig. 1). In short, people share reality both to connect with others and to know. Firstly, as represented in Axiom I, social relationships are established and maintained to the degree that participants in the relationship achieve specific, mutually shared understandings of themselves or the world about them. Conversation creates friendship when conversants discover common experiences. With the birth of a child, for example, new parents find themselves sudden inhabitants of a society in which the

< previous page

page_8

next page >

< previous page

page_9

next page >

Page 9 shared joy and heartbreak of parenthood connects them to other parents, including a remarkable number of strangers in restaurants, grocery stores, and gas stations. By the same token, shared reality theory postulates that establishing and maintaining social relationships is impossible without the achievement of shared reality within the relationship. Conversation fails friendship when conversants discover no common experience. Secondly, according to the theory, cognition is as dependent on shared reality as relationships are. As represented in Axiom II, cognitions are established and maintained to the degree that they are recognized, validated, and shared with others. For example, it is difficult to sustain belief in a revolutionary new childrearing technique if, when communicated in ecstatic enthusiasm, one is met with dull stares or disagreement. Although few would argue that these postulates are principally impossible, the skeptic may find the heart of shared reality theory implausible. After all, why should human capacities as fundamental and ubiquitous as social connection and cognitive representation be dependent on something as slippery as mutual perceptions of shared reality? We believe the communication literatures provide some answers. As reviewed elsewhere (Hardin & Higgins, 1996; Higgins, 1981a, 1992), it is well accepted that communication (a) involves shared, rule-governed conventions concerning social roles and behavior (e.g., Austin, 1962; Cushman & Whiting, 1972; Gumperz & Hymes, 1972; Peirce, 1940; Rommetveit, 1974; Ruesch & Bateson, 1968; Searle, 1969; Watzlawick et al., 1967), (b) requires cooperative coorientation and mutual perspective taking (e.g., Cushman & Whiting, 1972; Grice, 1971; Mead, 1934; Merleau-Ponty, 1962; Rommetveit, 1974), (c) functions not only to transmit information, but also to create and define social relationships (e.g., Blumer, 1962; Bolinger, 1975; Garfinkel, 1967; Gumperz & Hymes, 1972; Hawes, 1973; Watzlawick et al., 1967), and (d) is a socially interdependent process in which the purpose and meaning of the interchange is collaboratively determined (e.g., Blumer, 1962; Burke, 1962; Garfinkel, 1967; Goffman, 1959; Hawes, 1973; Krauss & Fussell, 1998; Merleau-Ponty, 1962; Rommetveit, 1974; Watzlawick et al., 1967). Communication is not only central to the construction of meaning, but requires at every turn the establishment and maintenance of “common ground,” which is, the mutual perception that communicative participants are talking about the same thing and share conversationally germane knowledge (for reviews, see Clark & Brennan, 1991; Krauss & Fussell, 1996; Sperber & Wilson, 1986). The literature not only documents the innumerable strategies employed to achieve, verify, and Fig. 1.1: Shared reality theory Axiom I: The establishment and maintenance of social relationships requires shared reality. Axiom IA: Relationships are established and maintained to the degree that shared reality is achieved and maintained among relationship participants. Axiom IB: Relationships are abandoned to the degree that shared reality is not achieved and maintained among relationship participants.Axiom II: The establishment and maintenance of individual experience requires shared reality. Axiom IIA: Beliefs are established and maintained to the degree that they are socially shared. Axiom IIB: Beliefs are abandoned to the degree that they are unshared. sustain accrued common ground in everyday conversation, but also the severe social and psychological disruption engendered when common ground is ecologically or experimentally subverted (reviewed in Krauss & Fussell, 1998). One can demonstrate this for oneself by observing the consequences of suppressing the usual head nods, mmms, and uh-huhs elicited in everyday conversation. However, we suggest aborting the experiment quickly if you want the conversation to continue or, indeed, if you wish to remain on friendly terms with your partner. In short, we are convinced that the fundamental role of cooperative, interdependent, meaning-making activity is not only the foundation of common conversation, but also the foundation of social relationships and cognitive representation. Thus, shared reality is defined as interpersonally achieved perceptions of common experience. Shared reality includes explicit agreement and consensus, although it is not limited to them. Shared reality may be expressed or tacit, newly negotiated or long assumed. Shared reality processes may operate either consciously or unconsciously, achieved through effortful deliberation or automatic information processing. Shared reality is assumed to be a perceptual state, that may correspond to objectively accurate mutual understandings to a greater or lesser degree. Defined as such, shared reality is closely related to notions of perspective taking, intersubjectivity,

< previous page

page_9

next page >

< previous page

page_10

next page >

Page 10 and common ground, thereby fitting well within the social-psychological tradition, which has long observed that people act on the assumption that they inhabit the same world as others who perceive it as they do (Asch, 1952; Heider, 1958; James, 1890). Indeed, virtually all theories of cognitive development assume that children learn to regulate themselves in relation to the desires and demands of the significant others in their lives (e.g., Case, 1985, 1988; Damon & Hart, 1986; Fischer, 1980; Selman, 1980; Sullivan, 1953), leading some to argue that the only means by which the child can establish an understanding of the outside world is as it is objectified through perspective taking (e.g., P. Berger & Luckmann, 1966; Mead, 1982; Vygotsky, 1962, 1978). Yet the capacity to see the world as others see it is not easy. Full-fledged perspective taking is said to not occur until relatively late in human development, perhaps as late as between ages 4 and 6 years (e.g., Case, 1985; Feffer, 1970; Fischer, 1980; Huttenlocher & Higgins, 1978; Piaget, 1965; Shantz, 1983). Moreover, despite the ubiquitous assumption that people are capable of perspective taking, the issue has proved thorny for psychologists and philosophers of mind, because it is no small feat to describe plausible mechanisms by which inherently discrepant subjectivities may be fully shared (e.g., Bakhtin, 1986; Rommetveit, 1974; Sperber & Wilson, 1986; Wittgenstein, 1953). Many of these difficulties are avoided by our conception of shared reality as the mutual perception of intersubjectivity rather than objective intersubjectivity. Hence, from our perspective, the ecological relationship between perceived and objective shared reality is an empirical question. However, like common ground, we conceive shared reality in pragmatic terms, as objectively accurate as necessary to be sustained given the goals, communication status, and knowledge of participants in a given interaction (Clark & Brennan, 1991; Krauss & Fussell, 1996). In short, shared reality is a “working intersubjectivity” that may well dissolve if objective conditions preclude the perception of mutual understanding. The Phenomenology of Shared Reality Shared reality (or lack thereof) is often felt reality. The experience of mutual understanding can be as sweet as the experience of misunderstanding can be sour. Anecdotally, few will deny the frustration elicited when others do not recognize the validity of their beliefs. Nor will many deny the charm of discovering that a new acquaintance experiences a poem or sunset in just the way that they do. Indeed, the delight engendered by the realization of shared reality is evidenced by modal constructions of humor. Jokes often involve the dramatic delay of an already shared understanding of the world, which is realized with a bang in the punch line. From this perspective, it makes sense that so much comedy trades on common stereotypes and prejudices, which provide culturally prefabricated shared realities. Systematic research also suggests that, in general, phenomenology is pleasant when experience is shared (e.g., Newcomb, 1953), and that an inability to achieve shared understandings is disagreeable and unpleasant (e.g., Orive, 1988). One of the most striking discoveries of Asch (1952) was the agitation and visceral discomfort exhibited by participants who found themselves at odds with a unanimous majority who did not share their experience of the length of lines. More recently, Schachinger (1996) found that an unobtrusive measure of the degree to which people believed significant others viewed them the way they viewed themselves was negatively related to anxiety, loneliness, confusion, and anomie. Although the experience of shared reality may be pleasant in many circumstances, it is important to note that shared reality theory does not invoke valence as a parameter in the model. Indeed, little reflection is necessary to think of everyday examples in which distinctly distasteful experiences may be shared. On the one hand, it may be useful or even pleasant to discover that others share the experience of, say, political subjugation, which may explain in part the power and utility of consciousness-raising activities (e.g., MacKinnon, 1987). Alas, however, shared reality processes are not limited to progressive situations designed for healthy self-actualization. They are at least as likely to be operative in the service of everyday interpersonal and institutional subjugation (e.g., Jackman, 1990; Jost & Banaji, 1994; Marx, 1846; Sidanius, 1999). In the case of the shared beliefs forming the ideological justification institutional economic classism, for example, felt shared reality may well be more pleasant for the haves than the have-nots. Moreover, it is likely that situations that allow one to break free of burdensome or otherwise distasteful shared realities about the self engender positive feelings (e.g., Steele & Aronson, 1995). Thus, shared reality theory does not postulate that either shared reality or relational connections are inherently pleasant, but rather that shared reality is required to bind social relationships (whether positive or negative) and objectify cognitions (whether positive or negative). NEW EVIDENCE SUPPORTING SHARED REALITY THEORY As reviewed elsewhere, we believe that evidence gleaned

< previous page

page_10

next page >

< previous page

page_11

next page >

Page 11 from a broad variety of extant literatures provides support for the basic axioms of shared reality theory (Hardin & Higgins, 1996). However, good theory not only provides a parsimonious grasp of extant facts, but also must prove useful in generating new research programs and empirical insights (e.g., McGuire, 1997). Hence, we now turn to a summary of new research generated from the perspective of shared reality theory in which we have focused on the role of relationship management and motivation in social cognition. In particular, we find evidence that relationshipspecific shared realities not only regulate cognition but also social relationships. Shared Reality in Social Cognition Nisbett and Ross (1988) wrote that the single most important lesson of social psychology is the degree to which individual thought and behavior is determined by the immediate situation. Contemporary social-cognition models of representation, including schema theories, connectionist, and construct accessibility theories (e.g., Bargh, 1995; Higgins, 1995; Smith, 1992; Stangor & Lange, 1988), are congruent with demonstrations of malleability and stability in social judgment. Shared reality theory complements these approaches by postulating a social-cognitive mechanism regulating when and for what purposes representations are established, maintained, and changed. Perhaps the most fundamental implication of shared reality theory is that particular cognitions are attached to the particular social relationships in which they are shared. Across a variety of research paradigms, we have found evidence for this in the context of the self-concept and social judgment. As suggested by the axioms illustrated in Fig. 1., evidence for the fundamental postulate of shared reality theory may take two broad forms. On the one hand, cognitions should change as a function of relationship activation along lines that implicate particular shared realities germane to the relationships. On the other hand, relationships should change as a function of belief activation along lines that implicate shared reality processes. Moreover, this implies that cognitive malleability and stability should mirror relationship malleability and stability. That is, to the degree that particular cognitions are attached to the particular relationships in which they are shared, and to the degree that different shared realities are achieved from relationship to relationship, cognitions should vary lawfully as a function of which relationships and concomitant shared realities are activated. We have termed this the relationship-specificity conjecture (Lowery, Hardin, & Sinclair, 2000). Social Identities and Significant Others in Self-Stereotyping Several programs of research have examined various ways in which self- and social judgment are subject to the demands of social relationships along lines laid down by shared reality theory. In one program of research, we investigated how common stereotypes affect the self-concept (Sinclair, Hardin, & Lowery, 1999). To the degree that ethnic and gender stereotypes represent broadly shared cultural beliefs (e.g., Allport, 1954), shared reality theory implies that the self-concept should (a) reflect the stereotypes of one's gender and ethnicity, (b) change along stereotyped dimensions as a function of social identity activation, and (c) reflect stereotypes to the degree that one believes that significant others stereotype the self. It is worth noting that although the first two hypotheses can also be derived from self-categorization theory, the third hypothesis cannot. To investigate these hypotheses, participants evaluated their math and verbal abilities after either their ethnic or gender identity had been made salient across several experiments. To explore relationships among self-evaluation and reflected social appraisal, participants also indicated their perceptions of how people in general as well as important others viewed their math and verbal abilities. Pilot research indicated that the perceived views of people in general reflected participants' sense of broadly held stereotypes, and that the perceived views of important others reflected participants' sense of the views of significant others, including family, close friends, and favorite teachers. The social identity manipulation was simple and unobtrusive. On the top, right-hand corner of the questionnaire, participants indicated their age and either their gender or ethnicity. Experiments involving participants from three different ethnic groups demonstrated that the social identity salience manipulation affected self-evaluations and the perceived evaluations of others along stereotype-consistent lines. Moreover, evidence suggested that participants were vulnerable to self-stereotyping to the degree that they believed that significant others endorsed the stereotypes as applicable to them. For example, Asian American women's selfevaluations of math ability were greater when their ethnicity was salient than when their gender was salient, but their self-evaluations of verbal ability were greater when their gender was salient than when their ethnicity was salient, consistent with stereotypes about Asians and women, respectively. Overall, self-evaluations were highly correlated with perceived evaluations of others, and effects of the identity salience manipulation on the perceived evaluations of others were

< previous page

page_11

next page >

< previous page

page_12

next page >

Page 12 identical to those found on self-evaluations. However, congruent with the mechanism specified by shared reality theory, results indicate that the relationship between self-evaluations and the perceived evaluations of people in general (i.e., stereotyped expectancies) was mediated by the perceived evaluations of significant others. In particular, although the relationship between self-evaluations and perceived evaluations of people in general was eliminated after statistically controlling for the perceived evaluations of significant others, the relationship between self-evaluations and perceptions of significant others remained strong after controlling for the perceived evaluations of people in general. An analogous, stereotype-consistent pattern was found in an experiment utilizing a sample of European Americans, for whom the particular gender and ethnic stereotypes are different. Here again correlations among self-evaluations and perceived evaluations of others were strong and affected identically by the social identity manipulation. Moreover, as with Asian American women, results indicate that the relationship between stereotypes and selfevaluation were mediated by the perceived evaluations of significant others. Although yielding a different pattern of results, an experiment utilizing African-American participants provided converging evidence of the role of shared reality negotiated in relationships with significant others in selfstereotyping. Given evidence that African Americans are particularly likely to cultivate close relationships in which racist stereotypes are collaboratively challenged (e.g., Ogbu, 1986), we anticipated that, although African Americans would be aware of stereotyped social expectancies, this knowledge would not translate into corresponding selfstereotyping effects. This is what we found. In particular, we found that African Americans perceived the evaluations of people in general (i.e., stereotyped expectancies) as more negative about their academic abilities when their ethnicity was salient than when their gender was salient, consistent with prevailing stereotypes. Yet no such effects of social identity salience were found on either the perceived evaluations of significant others or self-evaluations. However, the mediational analysis revealed that the pattern of judgments of social expectancies and self-evaluation observed among African Americans was explained by the same mechanism characterizing the self-stereotyping we observed among Asian and European Americans. Correlations among self-evaluations and perceived evaluations of others were high, and the relationship between self-evaluation and perceived evaluations of people in general was mediated by perceived evaluations of significant others. Hence, the mechanism postulated by shared reality theory integrates both the findings of substantial self-stereotyping among Asian Americans and European Americans as well as the lack of self-stereotyping among African Americans. Relationship Motivation in Self-Stereotyping Shared reality theory postulates that it is the activity of establishing and maintaining social relationships that gives life and resonance to beliefs, including stereotyped beliefs about the self. In particular, to the degree that shared realities serve relational functions as shared reality theory postulates, then effects of social appraisal on the self should occur to the degree that one is motivated to achieve or maintain the relationship. We have pursued this issue in a research program that directly investigated the role of relationship motivation in self-stereotyping by examining the degree to which participants assimilate the likely views of others into the self-concept (Sinclair & Hardin, 2000). In a series of experiments, women completed self-concept measures in the context of imagined or actual interactions with others they knew to value traditional versus nontraditional gender roles. For example, in one experiment, we found that women judged themselves as more feminine after imagining a conversation with Barbara Bush than Hillary Clinton. In another experiment, women completed self-concept measures after imagining either a positive or negative interaction with a famous woman characterized as either gender traditional (Barbara Bush, Martha Stewart) or nontraditional (Hillary Clinton, Madelaine Albright). Although the social self-tuning effect was replicated under conditions in which participants imagined a positive interaction, the effect was eliminated under conditions in which participants imagined a negative interaction, demonstrating the role of relationship motivation in the achievement of shared reality concerning the self. In follow-up experiments, social self-tuning was examined as a function of actual interactions with male confederates who ostensibly held gender traditional or nontraditional attitudes about women. Results suggest that women assimilated their self-views toward the confederate under conditions of high relationship motivation, but if anything contrasted their self-views away from the confederate under conditions of low relationship motivation. For example, women judged themselves as more feminine after interacting with a confederate with traditional (versus nontraditional) views under instructions to attempt to “get along as much as possible,” but judged themselves as less feminine after interacting with a confederate with traditional views under instructions to

< previous page

page_12

next page >

< previous page

page_13

next page >

Page 13 attempt to “evaluate his personality as accurately as possible.’’ These experiments not only demonstrate that shared reality about the self is achieved on highly valued dimensions when relationship motivation is high, but also that shared reality may be actively subverted when relationship motivation is low. Hence, this research extends previous demonstrations of social tuning effects on attitudes about others (e.g., Higgins & Rholes, 1978) to attitudes about the self. More important, it directly demonstrates the role of relationship motivation in the negotiation of shared reality about the self, converging with demonstrations that individual differences in relationship motivation moderate the degree to which shared reality may be achieved about judgments of others (Higgins & McCann, 1984; McCann & Hancock, 1983). Finally, this research is consistent with shared reality theory's postulate that shared reality functions in part to establish and maintain the social relationships in which it is achieved. Relationship Relevance and Automatic Prejudice The communication literature documents the ubiquity with which people seek common ground with others in everyday conversation (e.g., Clark & Brennan, 1991). However, shared reality theory goes an important step further by postulating that common relationship demands change more than the mere expression of beliefs. To further test the power of everyday relationship demands in information processing, we investigated social tuning in the context of automatic prejudice (Lowery, Hardin, & Sinclair, 2000). We examined the hypothesis that even automatic cognitions would assimilate toward the likely beliefs of others and would do so along relationship-relevant dimensions, as implied by shared reality theory. In particular, we assessed automatic prejudice under conditions in which the experimenter was more or less likely to be prejudiced. Across three experiments utilizing two different measures of automatic association, we found social tuning effects on automatic anti-Black prejudice, but only for participants whose ethnicity makes prejudice relevant to the new relationship. In particular, we found that the anti-Black prejudice exhibited by European Americans was greater when the experimenter was White than when the experimenter was Black. Pilot research indicated that participants in the experiments assumed that the experimenter was much less likely to be prejudiced against Blacks when he or she was Black than White. Furthermore, consistent with shared reality theory's prediction that social tuning effects should occur on relationship-relevant dimensions alone, we found that, although anti-Black attitudes of European Americans were affected by experimenter race, the anti-Black attitudes of Asian Americans – whose relationship with blacks is not characterized nearly as much by concern about prejudice – were not affected by experimenter race. This research not only provides support for the relationship specificity conjecture of shared reality theory, but demonstrates that even automatic cognition is dynamically regulated according to prevailing social relationship contingencies. Notably, these experiments directly pit automatic social tuning against simple stereotype priming, which would be indicated by the opposite pattern of results. In summary, our research suggests that broadly shared cultural stereotypes may be dynamically utilized in self- and social perception, but that this occurs according to the parameters described by shared reality theory. Research on social identity-related self-stereotyping suggests that stereotypes are incorporated into the self-concept to the degree that they are shared with significant others (Sinclair, Hardin, & Lowery, 2000). Research directly examining relationship motivation in self-stereotyping demonstrates that the shared realities achieved in even new relationships may have pronounced effects on the self-concept, but do so only to the degree that one is highly motivated to establish the relationship (Sinclair & Hardin, 2000). Research on automatic prejudice demonstrates that automatic social tuning occurs along dimensions chronically relevant to particular intergroup relationships. Shared Reality in Self-Regulation Shared reality theory not only postulates a role for the collaborative construction of self-views and social attitudes, but implies that these shared realities, in turn, affect the course of subsequent social relationships. In one research program, we have found that self-verification behavior is moderated by the degree to which self-beliefs are perceived to be shared with significant others (Hardin & Higgins, 1996). Self-verification is the set of processes in which people go to great lengths to elicit confirmation for the beliefs they have of themselves (e.g., Swann, 1990). This literature demonstrates that, regardless of whether self-beliefs are positive or negative, people privilege selfconsistent versus self-inconsistent as assessed by (a) time (Swann & Read, 1981a), (b) diagnosticity ratings (Swann & Read, 1981b), (c) confidence in feedback (Swann, Griffin, Predmore, & Gaines, 1987), and (d) solicitation of feedback (Swann & Read, 1981b; Swann, Wenzlaff, Krull, & Pelham, 1992). Remarkably, the pattern extends to people's social interaction choices. People choose temporary interaction

< previous page

page_13

next page >

< previous page

page_14

next page >

Page 14 partners whose evaluations are consistent with their self-views, whether positive or negative (Swann, Pelham, & Krull, 1989; Swann, Stein-Seroussi, & Giesler, 1992), and they are more committed to their marriages when the view of their spouses matches their self-concept (Swann, Hixon, & De La Ronde, 1992). Self-verification theory explains these findings in terms of the epistemic benefits of cognitive consistency (e.g., Swann, 1990). We have argued that self-verification effects may be explained instead in terms of relationship motivation (Hardin & Higgins, 1996). Shared reality theory postulates that (shared) self-views are not only worth defending on epistemic grounds, but function in part to maintain valued social relationships. To test this hypothesis, we examined selfverification behavior along dimensions of the self-concept perceived to be more or less shared with significant others, assessing self-verification in one experiment with an information-seeking paradigm and in another experiment with a partner choice paradigm. In a pretesting session, participants listed traits they believed characterized themselves as well as traits they believed significant others thought were true of them. These trait listings allowed us to ideographically identify self-views that were perceived to be shared with at least one significant other versus self-views perceived to be unshared with significant others. We chose shared and unshared traits that were matched on certainty (cf. Pelham, 1991). In one experiment, participants were contacted several weeks after pretesting by telephone, ostensibly to be recruited for a computerized evaluation of their personality. Participants were presented with choices of aspects of themselves they would be most interested in learning about. Replicating the basic self-verification finding, participants expressed more interest in learning about aspects that were selfconsistent than self-inconsistent, whether the attributes were positive or negative. Moreover, self-verification was shown to be affected by the degree to which participants perceived the self-views to be shared with significant others. Participants were more interested in learning about self-consistent attributes if pretesting indicated that they were shared with significant others. A second experiment replicated this pattern of results when self-verification was assessed by partner choice. Participants preferred to meet new acquaintances they believed had impressions of them that were self-consistent; moreover, they preferred to meet new acquaintances whose self-consistent impressions were shared with significant others. A follow-up experiment demonstrated that self-concepts shared with significant others regulate self-descriptions shared with new acquaintances as well as the course of social interaction. To do so, we observed the cognitive and social consequences of activating a self-concept shared with a significant other by manipulating the ostensible characteristics of a new acquaintance in a procedure developed by Andersen and Cole (1986). Previously unacquainted women were recruited in pairs for a study on getting to know others from a sample who had completed a pretesting series of questionnaires in which they had listed self-descriptive attributes as well as the attributes they believed significant others thought characteristic of them. Additionally, participants had listed attributes they believed characterized several significant others including their mothers. In the focal experiment conducted several weeks later, each participant first read a brief psychological profile of her partner, which was based on the attributes describing either her own mother or a yoked participant's mother. Immediately before the meeting, each participant wrote a brief self-description and completed several questions about expectations of the upcoming interaction. After a 10 minute unstructured conversation with their partners, participants completed a parallel questionnaire about the interaction. Results were striking in two respects. First, self-descriptions included more self-attributes perceived to be shared with mothers when participants anticipated meeting a partner who had been described like their own mother than a yoked participant's mother. More important, the partner manipulation had no effect on plausible alternatives, including self-attributes in general, attributes about the self perceived to be held by participants' mothers, or attributes about the self perceived to be held by other significant others. Instead, the effect of anticipating an interaction with someone who resembled participants' mothers affected self-descriptions only on the traits that participants perceived were shared between them and their mothers. Second, the manipulation affected both the anticipation of the interaction as well as post-interaction impressions. Participants anticipated liking more, getting along better, and being more interested in their partners when they had been described in terms resembling their own mother than a yoked participant's mother. Although predictably weaker, these effects held after the actual interaction. In another line of research, we have investigated how the dynamics of relationship motivation mediate selfimprovement motivation (Pham & Hardin, 1999). From the perspective of shared reality theory, ambivalence about self-improvement is less about managing hedonistic impulses and social norms and more about managing mutually incompatible shared realities found in different social relationships. To examine this process, we utilized a standard social comparison paradigm to

< previous page

page_14

next page >

< previous page

page_15

next page >

Page 15 investigate motivation to self-improve in the context of academics. According to shared reality theory, the positive self-improvement effects on academic motivation as a function of exposure to a lazy target should be moderated by the degree to which participants disidentify with the lazy target. Two experiments demonstrated that participants viewed themselves as less lazy, intended to work harder in the future, and exhibited increased self-esteem after exposure to a lazy (versus non-lazy) social target. The effect was most pronounced for people who most valued hard work – as assessed by a self-discrepancy measure and thereby likely to be based in shared realities with significant others (e.g., Higgins, 1989a). Most important, however, results indicate that the effect of negative social perception elicited self-improvement motivation to the degree that participants socially distanced themselves from the lazy target, as indicated by expressions of dislike and disinterest in meeting him. In short, the effect of social comparison was modulated by a kind of relationship management, in which the value of hard work had to be sacrificed to the degree participants were interested in connecting with the lazy target. Shared Reality in Interpersonal Interaction In research extending the shared reality analysis to the regulation of interethnic relationships, we have investigated effects of beliefs about the O.J. Simpson trial on interpersonal perceptions and behavior among African and European Americans. Due to the overwhelming amount of publicity surrounding the differences in white and black attitudes about the trial, we assumed that the Simpson trial represents different interethnic shared realities, and hence hypothesized that making the Simpson case cognitively salient would affect interpersonal relationships differently as a function of their ethnic composition. In one experiment, mixed-ethnicity and same-ethnicity participant pairs completed a cooperative task after subliminal exposure to images of O.J. Simpson, Bill Cosby, or the participant university icons. After completing the task, participants independently rated the quality of consensus achieved in the interaction as well as the extent to which they liked and felt similar to their partners. Results indicate that exposure to O.J. Simpson had opposite effects depending on the ethnic composition of the pairs in a manner consistent with the predictions of shared reality theory. As predicted by several theories, participants in the mixed-ethnicity pairs liked their partners less, felt less similar to their partners, and judged the consensus achieved in the interaction as poorer after exposure to Simpson's face than Cosby's face. In contrast, however, as predicted by shared reality theory alone, participants in the sameethnicity pairs like their partners more, felt more similar to their partners, and judged the consensus achieved in the interaction as better after exposure to Simpson's face than Cosby's face. In both cases, liking, similarity, and consensus judgments were intermediate after exposure to university icons. This pattern of results was replicated in two additional experiments that extended the effects on interpersonal perception to cooperative performance in a game of Pictionary, in which participants take turns attempting to communicate a target word to their partners through nonverbal drawings. Although participants in mixed-ethnicity pairs performed substantially worse after thinking about the Simpson trial than after thinking about the ClintonLewinsky scandal, participants in the same-ethnicity pairs performed substantially better. In summary, this research demonstrates the role of achieving shared reality about the self with significant others in regulating both the dynamics of self-understanding as well as relationship motivation. Self-verification appears to be predicated in part on the degree to which aspects of the self-concept are perceived to be shared with significant others. Moreover, the self is presented to new acquaintances along lines already laid down in significant relationships, thereby affecting the course of the new relationship. Moreover, beliefs assumed to be differentially shared among different ethnic groups can regulate the course of interpersonal perception and behavior along lines directly implied by shared reality theory. CONCLUSION We have argued that the study of social cognition should involve more than the identification of the cognitive foundations of social behavior, but instead should focus on the interdependence of cognition and social relationships. We believe that renewed attention to social-cognitive interdependence will not only reconnect the two great thematic strands represented in the history of social psychology, but promise substantive advances in our understanding of how humans work. We have argued that shared reality theory provides one such synthesis. Although simple, the formulation is generative. In addition to the predictions directly implied by the axioms of shared reality theory, the postulate that epistemic and relational functions are linked through the shared reality mechanism affords the derivation of several empirical corollaries worth future investigation. One concerns the specificity of experience implied by shared reality theory. To the degree that shared realities

< previous page

page_15

next page >

< previous page

page_16

next page >

Page 16 are bound to the particular relationships in which they are maintained, and to the degree that individuals have different clusters of personal experience, individual experience should vary to greater and lesser extent from relationship to relationship and situation to situation. For example, one's experience as an academic, shared and realized at school with students and colleagues, may operate relatively independently of one's experience as a member of the Sierra Club or PTA. Following the same logic, relationship motivation should also vary situationally as a function of the salience of relationship-relevant experiences. A further implication of the relationship-specificity conjecture is that human cognition has the capacity to be quite plastic, at least to the degree that one's social world is complex and compartmentalized. Hence, shared reality theory suggests a mechanism well suited to the repeated demonstrations in the social-psychological literature of cognitive malleability. By the same token, however, shared reality theory also postulates a mechanism for cognitive stability. Just as cognitive malleability may be explained in terms of relationship structure and motivation, so might cognitive stability. Stable social institutions create stable interpersonal relationship dynamics, which, in turn, produce stable cognitions. In short, shared reality theory offers a single mechanism compatible with observations of both stability and malleability in human cognition. The assumption that shared reality is relationship specific also implies that the proximal mode of self-categorization is interpersonal, although this does not in principle preclude self-categorization according to abstract group identities. However, shared reality theory's implication that social identification has psychological resonance because it is modulated by specific interpersonal relationships provides an empirically useful point of departure from selfcategorization theory (cf. Turner, 1984; Turner et al., 1986). Several other implications follow from shared reality theory, particularly as they involve relationship affirmation motives. For example, to the degree that shared reality binds social relationships, the motivation to achieve shared reality should be especially high under conditions of relationship threat or social anxiety. For example, not only do jilted partners in intimate relationships seek shared reality among friends, but it is not uncommon for them to attempt to get even former partners to understand their plight, sometimes even at the cost of personal dignity. However, to the degree that shared reality objectifies experience, motivation to achieve shared reality should be especially high under conditions of uncertainty. This may explain in part why graduate school classmates often become lifetime friends. To the degree that shared realities are realized in the establishment and maintenance of social relationships, they will be defended because a threat to a given shared reality represents a threat to the relationship on which the shared reality is grounded. For example, the realization that a new acquaintance does not appreciate one's political view is not problematic for its inconsistency per se, but rather because it represents a tacit threat to the relationships in which the political view is shared. Hence, from the perspective of shared reality theory, the psychological conflict borne of both intra- and interpersonal attitude discrepancies is viewed in relational terms as relationship-specific conflicting shared realities. When potential shared realities are incompatible, which one prevails? As implied by shared reality theory, the half-life of a given shared reality is positively related to the degree that it is grounded in multiple relationships and to the degree that the relationships on which it is grounded are stable, either through institutional imperatives or conditions in which relational motivation is high. For example, the relative permanence of family relationships makes them particularly potent social foundations of self-understanding, lending shared realities maintained in them more strength in the face of attack than a competing shared reality achieved, for example, with some guy once met at a party. Alternatively, asymmetrical relationship status may render one participant more motivated to affirm the relationship than another participant. In such cases, this corollary implies that movement necessary to achieve shared reality will occur more for the person more motivated to affirm or maintain the relationship. ACKNOWLEDGMENT This research was supported by NIH grant MH-10544 and a UCLA Academic Senate COR grant awarded to Curtis D. Hardin. Terri Conley was supported by a National Science Foundation graduate fellowship. For helpful comments on a draft of this chapter, we thank Brian Lowery, Yesi Pena, and M. Park. ENDNOTES 1. Interestingly, for both Mead (1934) and Freud (1933/1965), the impetus for their theories was the problem of self-consciousness (i.e., how the self makes itself an object of its own perception or, put another way, which part of the self is observing the self). Although their formulations were different, the root mechanism was the same: perspective taking. For Mead, self-consciousness is the product of a repository of others'

< previous page

page_16

next page >

< previous page

page_17

next page >

Page 17 perspectives on the self, motivated by epistemic interests (i.e., the generalized other). For Freud, self-consciousness is the product of significant others' perspectives on the self, motivated by social identification (i.e., the super-ego).

< previous page

page_17

next page >

< previous page

page_18

next page >

page_18

next page >

Page 18 This page intentionally left blank.

< previous page

< previous page

page_19

next page >

Page 19 2 The Personal Need for Structure and Personal Fear of Invalidity Measures: Historical Perspectives, Current Applications, and Future Directions Megan M. Thompson Defence and Civil Institute of Environmental Medicine, Toronto Michael E. Naccarato Kevin C. H. Parker Backbone Media, Inc., Toronto Kingston General Hospital, Kingston Gordon B. Moskowitz Princeton University Without too much absurdity from a functionalistic point of view, a biogenic need for structure or meaning could be postulated and hostility to alien events be accounted for in terms of threat to the structure or meaning attached to the situation. It is not necessary to posit the tendency to structure as a biological imperative, because, native or not, the making of some kind of sense out of a situation appears invariable and unquestionably essential to the prediction of recurrences and differences in the surrounding world. To cope adequately with the environment, even to survive in it, seems to necessitate the ability to read it or define it in some degree of veridicality at least. – Harvey (1963) Attaining meaning has been characterized as a primary human goal. Each person, each stimulus that meets our senses changes our environment, at times challenging our sense of knowing what to expect from the environment. Without structuring this array into coherent units that provide meaning, the world would be experienced as chaos. This is a disturbing and unsettling state that people are driven to avoid/reduce (Peirce, 1877/1957; Dewey, 1929). Why? First, for survival, an organism needs to know both whether immediate danger exists and whether stimuli that satisfy need states are present. Second, aside from promoting survival, the experience of knowing serves the drive to approach pleasant/beneficial and avoid unpleasant/harmful experiences. Without meaning being attached to the stimuli around us we would have no way to determine what stimuli in our environment are pleasurable (and perhaps species-promoting) versus painful (and perhaps species-hindering). Perception, categorization, inference, and memory are the cognitive tools humans are equipped with to provide the meaning that will shape subsequent action toward stimuli. Bruner (1957a) referred to this as categories serving a predictive function (noting that the predictions that are derived from one's inferences are of varying degrees of veridicality). James (1907/1992) stated these cognitive processes produce the meaning that allows us to anticipate, expect, act, and react, thus linking cognition with pragmatism. Third, attaining meaning is also a desired end state in and of itself. Doubt is experienced as unpleasant, and attaining meaning (knowing) allows that unpleasant state to be removed. This fact led Heider (1944) to conclude that humans possess a causal drive – a drive to establish a set of causal explanations that account for the stimuli in the environment. Heider assumed this was “a third basic drive beside the drives for self conservation and for the conservation of the species” (p. 359). Finally, Heider (1958) described this causal drive as not only reducing the unpleasantness of doubt, but also providing a sense of mastery or control over the environment. This sense of determinism allows one to feel as if the universe is not random, which is essential for both avoiding feelings of helplessness and initiating appropriate action. Given the central role that the attainment of knowledge plays in human functioning, it is not surprising that the processes through which and the manner in which meaning is sought represent fundamental pursuits in many areas of psychology and philosophy. How meaning about the social world is attained is an essential concern of social cognition and the dominant theoretical concern of this chapter. Our focus is on individual differences in how people tolerate the existence of uncertainty. Although it is believed that imposing meaning on the

< previous page

page_19

next page >

< previous page

page_20

next page >

Page 20 environment is a pervasive goal for all individuals, the decision as to how to expend both time and cognitive effort toward accomplishing this end is determined in part by chronic tendencies we have called the personal need for structure and the personal fear of invalidity. The Phenomenology of Personal Need for Structure (PNS) An individual possessing a high chronic need for structure prefers structure and clarity in most situations, with ambiguity and grey areas proving troublesome and annoying. Characterized by decisiveness and confidence, such people experience discomfort if they perceive structure and clarity to be missing from situations. They should dislike or be disturbed by people who vacillate or by opinions and situations that lack clarity and order. Although the clearminded, decisive individual is often lauded by society, this style may also lead to rigid, inflexible thinking and an unquestioned acceptance of the validity of one's beliefs (e.g., a reliance on stereotypes). It may prove to be an ineffective or unsuccessful style in highly creative environments that demand the generation of many hypotheses and/or the ability to adopt many perspectives on an issue or problem. Similarly, this style may be less useful in situations requiring the rapid review and reconsideration of beliefs in light of new evidence. The Phenomenology of Personal Fear of Invalidity (PFI) Although some individuals are driven by needs for clarity and structure, others may be more generally concerned with the cost of committing errors and be considered to have a high Personal Fear of Invalidity (PFI). Individuals high in PFI would be preoccupied with the consequence or perceived risk of an undertaking. To avoid potential mistakes, they would be more likely to see alternatives, vacillate between options, and show discomfort with feedback that indicated an error had been made. A heightened concern with error might be manifested through behavioral and cognitive hesitancy and a resistance to committing to situations and ideas. This characteristic manner of reevaluating options is logically linked to an agitation around decisions, as well as vacillation, ambivalence, and procrastination regarding important decisions. But in some circumstances a concern with error may have positive implications. For instance, this flexibility of thought suggests these individuals would be more data-driven than theory-driven and, as such, may be less likely to employ stereotypes or be quick to reach conclusions (see Kruglanski & Freund, 1983). HISTORICAL PERSPECTIVES A Brief Look at Two Philosophical Views of Epistemology The philosophical debate about how people pursue meaning revolved around the pragmatist belief that all knowledge is constructed and subjectively determined (with what is deemed to be “truth” at the whimsy of the perceiver) and the associationist perspective that knowledge is revealed to the perceiver from the truth that is inherent in objects. These beliefs led to opposing positions regarding the manner in which the individual conducts the quest for knowledge, with one camp endorsing the position that people seek truth (the best possible knowledge) and the other asserting that people merely seek an end to doubt (any sufficient knowledge that can structure the situation in a meaningful way that allows for prediction). Seeking truth implies that people prefer a thorough appraisal of the data; the mere removal of doubt suggests people are willing to rely on theories about the data rather than close inspection of the data itself. The pragmatist school believed that knowledge is functional – it allows one to prepare action (e.g., James, 1890). This suggests that if adequate knowledge (that allowing for useful inference and appropriate action) is capable of being produced by expending less effort, then the individual will be satisfied to stop the epistemic process. Peirce (1877/1957) referred to the struggle to remove doubt as inquiry, stating that “the sole object of inquiry is the settlement of opinion. We may fancy that this is not enough for us, and that we seek not merely an opinion, but a true opinion. But…as soon as a firm belief is reached we are entirely satisfied, whether the belief be false or true’’ (pp. 12-13). This position asserted that perceivers seek fast, structured knowledge that is satisfactory to allow for successful interaction with a stimulus rather than seeking to analyze the stimulus further. In opposition to this view that humans freeze their search for knowledge on the first reasonable explanation that comes to mind was the belief that the pursuit of meaning involves the quest for a more absolute truth. Bertrand Russell (as cited in Meyer, 1985, pp. 36-38) summarized this position well in criticizing the notion of inquiry as put forth by John Dewey (a disciple of Peirce). The first criticism is that “data, in the sense in which many empiricists believe in them, are rejected by Dr. Dewey as the starting point of knowledge.” Instead of the search for the meaning inherent in the data, the pragmatists place their “emphasis upon inquiry, as opposed to truth, or knowledge. Inquiry is not for [them], as it is for

< previous page

page_20

next page >

< previous page

page_21

next page >

Page 21 most philosophers, a search for truth; it is an independent activity.” Finally, this conception of humans as seeking closure mostly through effortless processing led Russell to state “knowledge, if Dr. Dewey is right, cannot be any part of the ends of life; it is merely a means to other satisfactions. This view, to those who have been much engaged in the pursuit of knowledge, is distasteful.” What is evident from this debate is that both camps are discussing ways in which perceivers typically attain knowledge by providing closure. The central issue is simply focused on the different strategies for providing such closure. Even the more deliberate process of closely analyzing the data to distill the true essence from a stimulus, with knowledge as an end in and of itself, is an attempt by the individual to gain closure. In this instance, the individual is just thought to be willing to tolerate greater amounts of uncertainty before bringing the epistemic process – the search for knowledge – to a halt. This fundamental difference in how people are presumed to pursue meaning is derived from differing assumptions about the nature of knowledge and meaning. For pragmatists, meaning is imposed on a stimulus by accepting any sufficient explanation; for associationists meaning is distilled from qualities inherent in the stimulus through accurately and effortfully attempting to analyze the data. A Psychological Theory of Epistemology Dual process models that have dominated the field in the last 20 years (see Chaiken & Trope, 1999) reconcile these antithetic philosophical approaches by suggesting two broad strategies people use to understand social stimuli. The opposing endpoints of a theoretical information-processing continuum mirror the distinct strategies debated by philosophers at the turn of the last century. First, people may systematically and effortfully examine the existing data and construct theories and meaning from the ground up. Second, they can heuristically and effortlessly (sometimes implicitly/unconsciously) arrive at a sufficient sense of knowing by starting with existing theories and expectancies and operating from the top down. Our work on individual differences focuses on what end of this processing continuum perceivers typically inhabit, and it has its roots in the theory of Kruglanski and Freund (1983) labeled lay epistemics . The theory of lay epistemology (e.g., Kruglanski, 1988; 1989) details the cognitive and motivational components of knowledge seeking. Although the cognitive component of knowledge seeking refers to the specific content of the concepts, thoughts, and beliefs generated, the motivational component refers to how motives determine the initiation, course, and cessation of the knowledge-seeking process. The cognitive and motivational elements operate concurrently in a twostage process of hypothesis generation and validation. The hypothesis generation stage is chiefly dependent on cognitive capability. This capability is, in turn, governed by the availability (e.g., the requisite knowledge stored in long-term memory) and the accessibility (the momentary ease of retrieval of a portion of what is stored in long-term memory) of particular thoughts or hypotheses. Hypothesis validation is the process through which an individual collects the evidence that serves to support or refute the generated hypotheses. Thus, in attempting to reduce doubt, individuals perceive a stimulus, generate hypotheses about what the appropriate category to capture that stimulus might be, test (validate) the hypotheses, freeze the hypothesis testing when an appropriate solution has been found, and then make inferences and plan action accordingly. The particular course (i.e., the evidence attended to and the weight given a piece of evidence) and termination of the validation process is directed by how quickly the person desires an answer (his or her willingness to live with uncertainty) and whether a specific type of answer is desired. Kruglanski and Freund's (1983) model focused on the testing and cessation of knowledge-seeking processes, labeling these as being affected by two orthogonal motives: (a) needs for structure and validity, and (b) needs for specific versus nonspecific closure. A need for structure, typically operationalized by Kruglanski and Freund as time pressure,1 occurs when an individual is compelled to impose structure to dispel uncertainty, doubt, and ambiguity. Thus, the need for structure is equivalent to a need to quickly remove doubt and arrive at a meaning – any meaning that adequately captures the stimulus situation. This is associated with cognitive consequences such as generating fewer hypotheses, a less thorough examination of relevant information, and a greater degree of confidence in one's judgments (Freund, Kruglanski, & Schpitzajzen, 1985; Kruglanski & Freund, 1983; Mayseless & Kruglanski, 1987). A need for validity (or the fear of invalidity), frequently operationalized as evaluation apprehension (e.g., Kruglanski & Freund, 1983), is related to the perceived costs of error as a result of some decision. When costs are substantial (such as concluding a stimulus is harmless when it is harmful), individuals are more cautious in their judgments. Thus, the need for validity is equivalent to a need to be cautious about inaccuracy. Fear of invalidity is associated with cognitive consequences such as generating more hypotheses, conducting a more careful information search, delaying reaching closure too quickly (lest

< previous page

page_21

next page >

< previous page

page_22

next page >

Page 22 one arrive at knowledge that is incorrect), and having less confidence in one's beliefs (Kruglanski & Freund, 1983; Mayseless & Kruglanski, 1987b). The Development of an Individual Difference Approach to Lay Epistemology The first two authors of this chapter became aware of the need for structure and fear of invalidity in 1984 when we were first-year graduate students, and Arie Kruglanski was a visiting professor, at the University of Waterloo. While taking his seminar detailing the then situationally based theory of lay epistemology, it occurred to us that it was likely that needs for structure and validity characterized typical response styles of certain individuals that would be evident across situations. Certainly individual differences in dogmatism (Rokeach, 1960), authoritarianism (Adorno, Frenkel-Brunswick, Levinson, & Sanford, 1950), and intolerance of ambiguity (Budner, 1962) had been long identified by clinical and motivational psychologists. These individual differences were described as “hot” motives that served basic emotional desires. However, they also contained components of structure seeking and avoiding that suggested to us that such differences might exist independent of the “hot” emotional components. Further, consistent with an interactionist approach (see Blass, 1984), we assumed such styles interact with situational demands, leading people to seek out and remain in situations that best meet these needs. We felt that recasting the situationally based need for structure and fear of invalidity as chronic tendencies had the potential to add to our understanding of how individuals understand and interact with their worlds. With Kevin Parker, we proceeded with the development of these individual difference measures on our own after Kruglanski left Waterloo. By October 1985 we had finalized versions of a 12-item PNS and a 14-item PFI scale. These measures went on to play central roles in our masters' theses (Naccarato, 1988; M. Thompson 1989). Moreover, the scales, particularly the Personal Need for Structure, were soon being used by other researchers examining social-cognitive processes (e.g., Kaplan, Wanshula, & Zanna, 1991; Moskowitz, 1993; Neuberg & Newsom, 1993). Beyond extending the situational theory of lay epistemology into the realm of individual differences, we also postulated an underlying conceptualization in which PNS and PFI represent two separate constructs that are moderately positively related to each other as they occur intrapersonally. For instance, high PNS and PFI might be expected to work in tandem in cases where one seeks out structure in order to clarify what is required in a situation, thereby reducing anxiety and the potential for error. After all, it is easier to avoid making potentially costly mistakes when one knows and sticks to the rules of the game. Second, at its highest levels, greater concerns with error might become so overwhelming as to effectively halt further hypothesis generation, thus appearing to lead to greater rigidity (Effler, 1984). In 1989 Kruglanski published his book describing the theory of Lay Epistemology, and applied it to a variety of social psychological phenomena. Kruglanski articulated a model of knowledge seeking governed by two orthogonal dimensions of closure and specificity, resulting in four distinct motivational orientations. One notable amendment at this time involved changing the names of two of the central constructs of the theory. Presumably because the needs for structure and validity described different situational pressures that affected how quickly people reach closure, these constructs were renamed. “Functionally opposite to the need for closure is the need to avoid closure. Those two needs are conceptualized as ends of a continuum ranging from strong strivings for closure to strong resistance of closure (Kruglanski, 1989)” (c.i. Webster & Kruglanski, 1994, p 1049). But what are the implications of this theoretical reformulation? Need for closure versus need for structure: What's in a name? If the names of these constructs are changed, the metaphorical processing continuum specified by dual-process models would be anchored by needs for closure and needs to avoid closure. Treating closure-seeking and closure-avoidance as opposing processes seems to imply that the philosophical debate between pragmatists and associationists over epistemology was exclusively between ending doubt (need for closure) and prolonging doubt (need to avoid closure). It was not. It was solely about ending doubt, with conflicting strategies with which it is assumed humans prefer to accomplish this task. It is for this reason that the labels need for structure and need for validity would appear to be more appropriate, because those labels suggest different routes toward the common end of attaining closure. They also describe opposing levels of ability to tolerate lacking closure without simultaneously implying that delaying attainment of closure means that closure is something that is always undesired (to be avoided). Indeed, often times closure is desired but is just arrived at more cautiously and slowly. Thus, because we felt the original labels of structure and validity better capture both the philosophical distinction between methods for arriving at knowledge and the psychological constructs that account for how meaning is produced,2 we retained the names need for structure and fear of invalidity in our work.

< previous page

page_22

next page >

< previous page

page_23

next page >

Page 23 We were especially pleased to see that by 1989 Kruglanski also had extended his own theorizing to acknowledge the potential influence of chronic tendencies: ‘‘[S]table individual differences in [need for structure] may also exist. Thus, people may reliably vary in their disposition to apprehend the world in clear-cut terms” (Kruglanski, 1989, p. 6). Webster and Kruglanski (1994) subsequently published the Need for Closure Scale (NFCS), a 42 item scale consistent with their unidimensional vision of closure. Webster and Kruglanski (1994) presented a confirmatory factor analytic (CFA) model and results of CFA analyses illustrating their conceptualization of the NFCS in terms of a second order factor structure with five factors loading onto a single closure construct. The NFCS included 11 of the 26 items in our PNS and PFI scales. Kruglanski and Webster (1994, see also Kruglanski, Atash, DeGrada, Manetti, Pierro, & Webster, 1997) maintained that the scales (PNS/PFI and NFCS) tapped orthogonal constructs, reporting low correlations between the measures. Neuberg and colleagues disputed these results (Neuberg, Judice, & West, 1997; see also Neuberg, West, Judice, & M. Thompson, 1997). They provided convincing evidence that the three scales are not orthogonal, obtaining an average correlation of .7 across numerous samples. Neuberg et al. (1997) also persuasively challenged the psychometric quality of the Webster and Kruglanski (1994) NFCS. They concluded that the NFCS is a multidimensional instrument measuring an ostensibly unidimensional closure construct – indeed they discussed the dangers inherent in adopting such an approach. The dimensionality of these respective constructs is an issue we shall return to later. Moreover, Neuberg et al (1997) determined that the most potent NFCS items are those 11 taken from our original PNS and PFI scales. Given the importance that pursuing knowledge serves in human functioning, and given the central concern of social cognition with delineating the mechanisms involved in how knowledge is produced, it seems critical to have the best vehicle possible to measure such phenomenon. The measures we present here will hopefully be of some assistance in this endeavor, illuminating important individual differences in structuring and making sense of the social world. THE PSYCHOMETRIC DEVELOPMENT OF THE PNS AND PFI SCALES Psychometrics: Phase I - Initial Item Generation From the inception this work the first three authors attempted to adhere to the strict lay epistemic definitions of the situational need for structure and fear of invalidity constructs, focusing on the face validity of each item. Each PNS item was clearly linked to a desire for structure, certainty, and decisiveness, and every PFI item reflected a concern with error and decisional consequences. Efforts were made to keep each scale item content free (i.e., not bound to particular content domains such as political ideology) and to tap the cognitive, affective, and behavioral manifestations of the PNS and PFI, thereby promoting the application of the constructs to a variety of psychological phenomena. This process yielded a pool of 25 PNS and 30 PFI statements.3 Subjects and Procedure Two hundred fifty-eight introductory psychology students (104 males and 154 females) at the University of Waterloo completed the 25-item PNS(A) and 30-item PFI(A)4 scales in an “Attitude, Belief, and Experience Survey” administered as part of a mass testing procedure. Responses were recorded on a 6-point Likert rating scale ( strongly disagree - strongly agree ). Results for PNS (A) Factor Structure of the Initial Version of the PNS(A). Table 2.1 includes the factor analytic results of this iteration of the PNS scale. Principal components factor analysis (PCA),5 employing oblique and varimax rotations, led to the retention of three factors, accounting for 47.9% of the variance. We labeled the factors (a) Preference for Orderliness (12 items; 28.8% variance; e.g., “I like to see to it that my work is carefully planned and organized;” (b) Discomfort with Unpredictability (10 items; 11.7% variance; e.g., “It bothers me when something unexpected interrupts my daily routine”); and (c) Disdain for Ambiguity (3 items; 7.9% variance; e.g., “It's annoying to listen to someone who cannot seem to make up his or her mind”). Results of the PCA using an oblique rotation indicate that Factor I, Preference for Orderliness, was moderately correlated with Factor II, Discomfort with Unpredictability ( r = .28) and unrelated to Factor III, Disdain for Ambiguity ( r = .08). Discomfort with Unpredictability yielded a moderate correlation with Disdain for Ambiguity ( r = .21). One criterion for inclusion in the next version of the PNS scale was a factor loading of above .40 on any of the first three factors. When items were considered to be identical in meaning, only higher loading items were retained. Using these criteria, 15 items were retained for the PNS(B) scale. Factor analyses on retained items

< previous page

page_23

next page >

< previous page

page_24

next page >

Page 24 yielded a similar factor structure, with the three factors accounting for 57.9% of the explained variance in this subset of PNS items (Factor I: 7 items, 34.1%; Factor II: 5 items, 12.4%; Factor III: 3 items, 11.4%). Factor loadings ranged between .01 and .86. Internal Consistency. Reliability analyses conducted on the PNS(A) scale yielded a Cronbach's alpha of .88, with itemtotal correlations between .08 and .68. The Cronbach's alpha for the 15 retained items was .85, with item-total correlations ranging between. 13 and .64. Results for PFI(A) Factor Structure of the Initial Version of the PFI(A) . Principal components analysis yielded identical unifactorial solutions for both varimax and oblique analyses accounting for 22.7% of the variance in PFI scores. Factor loadings ranged from .01 to .76. The 15 items retained in the next version of the PFI scale had loadings of .40 and above. Poorly worded or ambiguous items were deleted from further analyses. Principal component analyses conducted on retained items revealed a one-factor solution. Factor loadings ranged from .01 to .76, accounting for 23.4% of the variance in PFI scores. Internal Consistency. The internal consistency of the PFI(A) scale was more than adequate, with a Cronbach's alpha of .84 and item-total correlations ranging from .004 to .69. A second reliability analysis of the retained PFI items yielded a Cronbach's alpha of .87 with item-total correlations ranging between .35 and .70. Psychometrics: Phase II Three issues were of importance here. First, the three-factor structure that emerged from the original PNS items suggested the scale might not reflect as pure a personal need for structure as might be wished. Second, we hoped the refinement of the scale would increase the amount of variance accounted for by scale items. Third, earlier analyses eliminated almost all reversals on both scales, thus several contrait items were developed that were mirror image reversals of items retained for this iteration of the scales. This second item generation phase resulted in 27 items in the PNS(B) scale and 30 in the PFI(B). Subjects and Procedure Participants were 210 (77 males and 133 females) introductory psychology students who completed the PNS(B) and PFI(B) measures as part of a mass testing session – a procedure identical to that of the first study. Results for PNS(B) Principal Components Analysis . A principal components analysis of the PNS(B) scale, presented in Table 2.1, again revealed a three-factor solution, which accounted for 42.6% of the variance in PNS scores (Factor I: 9 items, 26.8%; Factor II: 10 items, 9.3%; Factor III: 8 items, 6.5%). Factor loadings ranged between .004 and .80. The factor intercorrelations were r = .31 (Factor I & Factor II), r = .28 (Factor I and Factor III), and r = .02 (Factors II and III). Items with factor loadings above .40 were retained. Again, ambiguous items were eliminated, and an attempt was made to retain as many reversals as possible. Similar factor analyses performed on the 12 retained PNS scale items revealed a one-factor solution, accounting for 36.5% of the variance in PNS scores. This unifactorial solution indicated that the item selection process reduced the noise in PNS scores without a substantial reduction in the amount of variance accounted for. Factor loadings ranged from .50 to .73. A review of a three-factor solution imposed on the retained items failed to produce three conceptually meaningful factors. Internal Consistency. As Table 2.1 also shows, the Cronbach's alpha for the 27-item PNS(B) scale was .85, with itemtotal correlation ranging from .01 to .67. The reliability analyses conducted on the 12-item final version PNS scale produced a Cronbach's alpha of .84, with item-total correlations between .40 and .62. Results for PFI(B) Principal Components Analysis . Table 2.2 includes the principal components factor analysis of the PFI(B) scale. A one-factor solution accounted for 31.2% of the variance in PFI scores. Factor loadings ranged from .02 to .74. Items with factor loadings above .50 were retained. Factor analyses run on the retained 14 items indicated that a onefactor solution provided the best fit with the data, accounting for 40.0% of the variance in PFI scores. Factor loadings ranged from .44 to .72. Internal Consistency. Table 2.2 also indicates the results of reliability analyses conducted on the revised PFI(B) scale, which yielded a very strong Cronbach's alpha of .91. Item-total correlations on these items ranged from .02 to 74. Similar analyses were performed on the 14 items that were retained for the final PFI scale. Here the alpha was .88, and item-total correlations ranged from .44 to .72. Together these results suggest that the PFI scale has an excellent internal consistency.

< previous page

page_24

next page >

< previous page

page_25

next page >

Page 25 Table 2.1: Psychometric Summary of the PNS Scale (PNS(B)/PNS) ItemFactor LoadingsItemTotalCorrelationsItemFactor LoadingsItems-Total CorrelationsPNS(B) 27PNS12PNS(B) 27PNS12PNS(B) 27PNS12PNS(B) 27PNS12123123PNS1 *.15.60.25.55.47.46PNS14 *.31.73.02.59.57.49PNS2 *.16.34.62.57.47.48PNS15-.08.45.05.22PNS3 *.64.04.42.70.55.60PNS16 *.70.08.33.73.58.62PNS4-.04-.02.43.15PNS17-.26.26-.43-.16PNS5.79.09.17.55PNS18 *.57.28-.08.52.44.41PNS6 *.09.42.40.50.41.40PNS19.31.64-.11.47PNS7.22.32.70.57PNS20 *.28.16.57.55.46.46PNS8.56.03.43.48PNS21.03-.01.59.25PNS9.24.55.27.51PNS22-.004.28-.34.01PNS10.19.32.13.22PNS23 *.80.18.06.71.61.61PNS11.04.56-.22.22PNS24 *.31.004.48.51.34.41PNS12 *.64.18-.07.53.42.42PNS25 *.38.55.32.71.63.62PNS13.31.64.34.67PNS26.24.07.07.22PNS27.51.24.14.49EigenvaluesTotal %Chronbach's alphaPNS(B)277.242.511.7742.6.85PNS124.3836.5.84Note: * Items retained in PNS scale. Highest loading factors on the retained items from PNS27 are indicated in boldface type. Psychometrics: Phase III We administered our 12-item PNS and 14-item PFI scales to a separate sample of 157 introductory psychology students (29 males, 128 females) at the University of Waterloo as part of a mass testing session. Our aim was to determine whether the psychometric qualities of the items would yield values consistent with earlier samples. Results for PNS Principal Components Analysis and Reliability Analysis . Table 2.3 includes the results of psychometric analyses of the 12-item PNS scale. The PCA yielded a one factor solution, with factor loadings between .54 and .69, accounting for 37.8% of the variance PNS scores. Reliability analysis yielded a Cronbach's alpha of .84, with item-total correlations ranging from .58 to .60. Results for PFI Principal Components Analysis and Reliability Analysis . As indicated in Table 2.4, the PCA of the 14-item PFI scale produced a one factor solution which accounted for 31.4% of the variance in PFI. Factor loadings ranged from .22 to .73. The results of the reliability analysis of the PFI scale, also presented in Table 2.4, produced a Cronbach's alpha of .82, with item-total correlations ranging between .18 and .62. Therefore, the internal attributes of the final version of each measure remain good, each yielding scale characteristics comparable to that of earlier samples. Questions concerning the convergent and divergent validity of our measures remain to be answered, however, and are the focus of the next section. The Convergent and Discriminant Validity of the PNS and PFI Measures We have hypothesized that PNS and PFI are ubiquitous constructs that underlie important phenomena in a variety of psychological domains. As well, we suspect that there exists a moderate and positive relation between PNS and PFI. To explore these ideas, we selected a variety of scales measuring constructs reflecting a variety of human experience: Rigidity About Personal Habits (RAPH; Meresko, Rubin, Shantz, & Morrow, 1954), Right-Wing

< previous page

page_25

next page >

< previous page

page_26

next page >

Page 26 Table 2.2: Psychometric Analyses of the Personal Fear of Invalidity Scale (PFI(B)/PFI) ItemFactor LoadingsItemTotalCorrelationsItemFactor LoadingsItems-Total CorrelationsPFI(B) 30PFI14PFI(B) 30PFI14PFI(B) 30PFI14PFI(B) 30PFI 14PFI1 *.55.55.52.47PFI16.34.50PFI2.24.20PFI17.56.71PFI3.43.38PFI18 *.76.44.44PFI4.58.53PFI19 *.50.52.56.57PFI5.01.02PFI20 *.62.64.49.44PFI6.45.40PFI21.51.51.29PFI7.76.73PFI22.30.62PFI8 *.53.51.50.44PFI23.66.69PFI9 *.63.66.58.60PFI24 *.74.74.72PFI10.52.48PFI25 *.78.78.53.54PFI11 *.65.66.60.58PFI26.58.62.27PFI12 *.68.68.64.60PFI27 *.30.56.57PFI13.13.11PFI28 *.62.66.74.71PFI14 *.64.67.59.60PFI29 *.78.77.44.47PFI15.34.31PFI30.51.56.37PFI(B) 30PFI14Eigenvalues9.355.61Total % Variance31.240.0Cronbach's Alpha.91.88Note: PFI30(b) indicates secondversion of PFI scale. * Items retained in PFI scale. Factor loadings for retained PFI items are in bold face type. Table 2.3: Factor Loadings and Reliability Analyses of the PNS Scale ItemFactor LoadingItem-Total Correlation1.It upsets me to go into a situation without knowing what I can expect from it..56452*I'm not bothered by things that upset my daily routine..59.483.I enjoy having a clear and structured mode of life..66.574.I like a place for everything and everything in its place..61.515.*I like being spontaneous..54.446.*I find that a well ordered life with regular hours makes my life tedious..61.507.I don't like situations that are uncertain..58.488.I hate to change my plans at the last minute..68.589.I hate to be with people that are unpredictable..57.4710.I find that a consistent routine enables me to enjoy life more..69.6011.*I enjoy the exhilaration of being put in unpredictable situations..64.5312.I become uncomfortable when the rules in a situation are not clear..55.43Eigenvalue 4.45Total % Variance37.8Cronbach's Alpha .84Note *.Note: Contrait items

< previous page

page_26

next page >

< previous page

page_27

next page >

Page 27 Table 2.4: Factor Loadings and Reliability Analyses of the PFI Scale ItemFactor LoadingItem/Total Correlation1.*I may struggle with a few decisions but not very often..58.482.*I never put off making important decisions..50.423.Sometimes I become impatient over my indecisiveness..57.454.Sometimes I see so many options to a situation that it is really confusing..70.595.I can be reluctant to commit myself to something because of the possibility that I might be wrong..68.586.I tend to struggle with most decisions..73.627.Even after making an important decision I continue to think about thepros and cons to make sure that I am not wrong..62.498.*Regardless of whether others see an event as Positive or negative I don'tmind committing myself to it..22.189.I prefer situations where I do not have to decide immediately..37.3010.*I rarely doubt that the course of action I have selected will be correct..47.3911.I tend to continue to evaluate recently made decisions..61.4912.I wish I did not worry so much about making errors..68.5613*Decisions rarely weigh heavily on my shoulders..43.3614.I find myself reluctant to commit to new ideas but find little comfort inremaining with the tried and true..36.27Eigenvalue:4.35Total Percent Variance:31.40Cronbach's Alpha:.82Note:* Contrait items Authoritarianism (RWA; Altemeyer, 1988), the Beck Depression Inventory (BDI; Beck, Rush, Shaw, & Emery, 1979), and Self-Consciousness (Fenigstein, Scheier, & Buss, 1975). Personal Need for Structure Authoritarianism . The notion that people require more or less structure and clarity in their lives is inherent in the work on the authoritarian personality (e.g., Adorno et al., 1950; Altemeyer, 1988). We expect that people high in PNS will also evidence a predilection for clarity and certainty in thought and action – a major aspect of authoritarianism. However, the authoritarian syndrome also includes a preoccupation with power and strength and right-wing political attitudes that need not relate to the PNS construct, which taps a more general idea of structure. This suggests that the magnitude of the correlation between the PNS and F scales should be moderate rather than high. Rigidity About Personal Habits (RAPH) . The RAPH scale measures the degree of rigidity evident in one's personal habits and includes traditionalism, rule boundedness, and discomfort and opposition to change. People scoring high on the RAPH are less likely to modify their behaviors and personal habits even when it may be appropriate to do so (Meresko et al., 1954). Because both the PNS and RAPH constructs emphasize rigidity and discomfort with change, a positive relation is expected to exist between high PNS and RAPH scores. Depression . Stable, negative, global attributions and rigid self-schemata that disrupt the depressive's ability to utilize positive information typify a depressed state (Beck, 1963; Beck et al., 1979). This primitive thinking style reduces “the complexity, variability and diversity of…experience and behavior” (Beck et al., 1979, p. 15). We suspect this rigidity would be reflected in a moderately positive correlation between PNS and scores on the BDI. Self-Consciousness. Self-consciousness is the extent to which an individual is in touch with one's inner thoughts and feelings, and concerns about oneself as a social object (Fenigstein et al., 1975). Theoretically, there is little reason to expect any relation between the PNS and self-consciousness scales. Personal Fear of Invalidity Authoritarianism . We expect a slight positive relation between PFI and RWA scores because high

< previous page

page_27

next page >

< previous page

page_28

next page >

Page 28 scores on measures of authoritarianism reflect a tendency to engage in rigid disciplinary behavior (Adorno et al., 1950), conceptually similar to punishment for committing errors. Nonetheless, we also expect the relation between PFI and authoritarianism to be much less than that of PNS and RWA responses. RAPH. Inspection of the RAPH indicates that there are few items that overlap with the identification of authority and harsh injunctions that we believe account for the slight relation between the PFI and RWA sales. Thus, we expect there to be little relation between scores on the PFI and RAPH. Depression . Beck and colleagues maintained that an emphasis on the negative aspects of experience contributes to depression such that depressives appear to need “absolute certainty of the correctness of a decision” (Beck et al., 1979, p. 168). It could be argued that it is the concern with error that leads an individual to focus on negativity and failures. This focus, in turn, leads to dysphoria and other associated symptoms of depression. Self-Consciousness. One of the general dimensions of self-consciousness is the tendency to attend to inner thoughts; this is referred to as private self-consciousness. The other dimension involves an awareness and concern about the self as a social object and is tapped by two factors: a greater sensitivity to rejection by a peer group (public self-consciousness) and discomfort in the presence of others (social anxiety; Fenigstein, 1979). Concern about how one may be viewed by others might be viewed as anticipatory apprehension linked to the possibility of embarrassment due to the commission of some social error. Social anxiety is similar to public self-consciousness, but here the emphasis is on concerns with active censure from others and is likely tied to some history of social faux pas or perhaps the threat of punishment for errors. Thus, we expect a positive relation to exist between PFI and Self-Consciousness. Subjects and Procedure Subjects were the 157 male and female first-year university students who participated in the third psychometric study described earlier. Subjects completed the validity scales along with several others during class time and received course credit for their participation. Results A multitrait-monomethod analysis was conducted to establish the validity of our scales. Two patterns of results are important. First, each scale should be positively related to other scales that tap related constructs, Table 5: Multitrait-Monomethod Matrix for University of Waterloo Sample PNSRWARAPHPFIBDIS.C.PNS(.84)RWA.37(.73)RAPH.64.46(.82)PFI.24.22.15(.81)BDI.29.20.25.47(.86)S.C..26.10.21.64.48.(75)Note: Scale reliabilities are presented along the diagonal. SC=Self-Consciousness Scale establishing the convergent validity of the PNS and PFI measures (Selltiz, Wrightsman, & Cook, 1976). In this case, PNS should be positively and more highly correlated with RWA and RAPH than should the PFI scale, whereas PFI should evidence a greater relationship to Self-Consciousness and Depression than should PNS. Second, evidence of the discriminant validity of the scales would be indicated by low correlations between the PNS and PFI measures and other scales that are expected to tap less related phenomenon (Selltiz et al., 1976). Thus, the PNS measure should yield moderate to small correlations with the Depression and Self-Consciousness scales. Similarly, the discriminant validity of the PFI scale would be evident by yielding correlations of a small magnitude with both right-wing authoritarianism and rigidity about personal habits. The Interrelationship of PNS and PFI. Results of Pearson correlations, presented in Table 2.5, confirmed our thinking about the conceptual interrelation of PNS and PFI being moderately and positively related to one another ( r = .24, p = .001) in this sample.6 Personal Need for Structure. Table 2.5 results substantiated our prediction concerning a moderate and positive relations between Authoritarianism and Personal Need for Structure ( r = .37, p < .001). As expected, we also found a positive relation between high PNS and RAPH scores ( r = .64, p

Page 29 correlation between PNS and Self-Consciousness, controlling for the effects of PFI. The results of this correlation indicate little association between PNS and Self-Consciousness ( r =. 14, p = .04) after the effects of PFI were partialed out. Personal Fear of Invalidity. Substantiating our hypothesis, the correlation between the PFI and Authoritarianism scales was r = .22, p = .005. Results of the z test also tend to substantiate our predictions, in that the correlation between PFI and Authoritarianism was of a smaller magnitude than that obtained for the PNS and RWA scales [ z (158) = 1.61, p < .10, one tailed]. As we had also anticipated, there was little relation evident between scores on the PFI and RAPH ( r =. 15, p = .005), providing evidence of the discriminant validity of the PFI scale. Results were even more striking when we performed a partial correlation between PFI and RAPH scores, controlling for the effects of PNS ( r = .001, ns.). Also as expected, high PFI was strongly associated with a tendency to experience depression ( r = .47, p

Page 35 Fig. 2.5. Confirmatory factor analysis model for the Arizona State sample: 11-item PNS scale, Neuberg & Newsom two-factor model

goodness of fit indexes [ X 2 (21) = 129.91, p < .001, GFI: .84; AGFI: .66; RMR: .07; CFI: .71; NFI: .69]. Direct comparisons of the three models indicate that both two-factor models provide a better account of the data [Thompson et al., vs. Kruglanski: X 2 cha(8) = 85.79, p < .001; Neuberg vs. Kruglanski: X 2cha(2) = 86.81, p < .001]. In this instance, the Thompson et al. model did not provide a significantly better fit to the data than the Neuberg model [ X 2cha(6) = 1.02, ns.]. FUTURE DIRECTIONS The literature reviewed on the PNS and PFI measures suggests their significance in terms of important processes in social-cognitive phenomena, attitudes, and, on a more limited basis, behaviors. From the outset of our work, we conceptualized these measures as assessing pervasive cognitive styles. To the extent that this assumption is true, the measures should be readily applicable to an even wider variety of psychological phenomena. In this section, we address the future theoretical and applied research questions concerning the measures that remain to be explored. A central goal for future research should be to continue to disentangle the underlying theoretical relation between the Structure, Validity, and Closure constructs. The results of our own analyses here, as well as the prior work by Neuberg, suggests that a multidimensional model provides a better account of the data in each sample than does a unidimensional model. Accordingly, the question remains as to whether there is a place for the unidimensional need for closure construct. We believe so. According to Heider (1944), every human interaction introduces at least a modicum of doubt. We see a need for closure as the arousal state caused by this doubt. Structure seeking is one process by which such arousal – which is ubiquitous – is reduced. Thus, it seems probable that people may have developed separate mechanisms for how they tolerate doubt (e.g., how much and for how long) and how they work toward the removal of doubt, the reduction of arousal caused by doubt, and the attainment of meaning. Despite his reticence about the Need for Closure scale, Neuberg too has similarly concluded that Closure is a viable theoretical construct. Indeed, he utilized Webster and Kruglanski's depiction of a two-stage process of seizing and freezing as a way in which to understand this relation. Seizing “represents the person's urgent desire to obtain any closure as quickly as possible” (Neuberg et al., 1997, p. 1405), whereas freezing “is organized around protecting the answer just obtained, the existing structure” (Neuberg et al., 1997, p. 1405). Closure here involves the somewhat nonreflective cessation of a cognitive search (i.e., the seizing), and structure is a mechanism invoked to maintain closure. Future research should be directed toward the determination of the particular antecedents, correlates, and consequences of needs for structure, validity and closure. Perhaps in its most unadulterated form, a need for closure is associated with novel environments, extreme

< previous page

page_35

next page >

< previous page

page_36

next page >

Page 36 Fig. 2.6. Confirmatory factor analysis model for the Arizona State sample: Kruglanski et al. one-factor model

levels of time pressure, and/or less important or personally relevant knowledge domains. Whatever theoretical relation between closure, validity, and structure prevails, we believe that each construct has a place in the understanding of knowledge-acquisition and validation processes. Efforts also should be directed toward the further elucidation of the interrelations among PNS, its subfactors, and PFI. The CFA evidence presented here clearly only begins this task. The ideal construct validity study would be one that included both the PNS and PFI scales and demonstrated that the two constructs have similar knowledgeacquisition effects under some conditions while having distinct effects under other conditions. For instance, people high in PNS should demonstrate a preference for consistency and structure in virtually all situations. However, persons high in PFI should only wish to invoke structure under conditions that increase their concerns with error. It would also be useful to determine how these measures may be related to specific phases in the judgment and decision making process. For instance, it would be interesting to determine which differences in cognitive styles emerge in the hypothesis-generation stage, the validation stage, or both. Preliminary work on this issue suggests that, once attributes are generated, individuals who are high in PNS sort them in a more simplistic manner (Neuberg & Newsom, 1993). Although highly suggestive, future research might more directly address the hypothesisgeneration sequence. Similarly, research might assess the hypothesis-validation sequence by directly tracking how respondents go about weighing, promoting, and discarding alternative hypotheses. The link between PNS and the processes related to stereotyping is one of the most robust findings associated with the PNS scale (e.g., Kaplan et al., 1991; Moskowitz, 1993; Naccarato, 1988; Neuberg & Newsom, 1993; Pilkington & Lydon, 1997; Schaller et al., 1995). The next important step would be to demonstrate that individuals possessing high PNS also engage in punitive behavior toward certain groups or individuals. If this relation is established, research efforts should be directed toward lessening the likelihood that those high in PNS will form, maintain, and act on negative stereotypes. Schaller's (Schaller, Boyd, Johannes, & O'Brien, 1995) work has already provided a potential start point for such efforts. As the above research illustrates, much of the past work concerning PNS has highlighted its potential negative consequences. However, our original aim was to create measures of cognitive styles that would be free of some of the evaluative aspects present in other scales such as Intolerance of Ambiguity. Thus, it would be interesting to show that a high-PNS style might be as readily associated with positive outcomes, at least in some instances. For example, those high in PNS might be more resistant to incorporating new negative information into an existing positive schema. Would the need for consistency persevere even when a behavior has some cost for the individual? In these instances, perhaps it would be individuals who are high in PNS who would be more likely to “show the courage of their convictions”. These ideas naturally lead to a consideration of the broader issue of how cognitive styles affect the attitude-behavior relation. For instance, the desire for structure and consistency between past and present beliefs should dictate that individuals high in PNS might be more likely to act in ways that are consistent with their beliefs. Conversely, the hesitation, ambivalence, and doubt so

< previous page

page_36

next page >

< previous page

page_37

next page >

Page 37 characteristic of high PFI indicates that it should be much easier to disrupt the attitude-behavior links of individuals high in PFI. Moreover, it would be interesting to explore the reactions of those people who are high in these cognitive styles in dissonance-arousing situations. It would be particularly interesting to identify the dissonance reduction techniques associated with each cognitive style. Presumably, individuals high in PFI experience and thus endure the affective and cognitive struggles inherent in dissonance more often. Although dissonant states are unpleasant for all individuals, they may prove especially distressing for people high in PNS. Thus, those high in PNS may be motivated to reduce dissonance more quickly. Indeed, might there be instances in which high-PNS individuals demonstrate greater attitude change to justify and reconcile their apparently inconsistent actions? Similarly, the doubt, reevaluation of options, and revisiting of decisions once made that is typical of high PFI should mean that there would be evidence of a variety of conflicts at each point in the epistemic process (e.g., anticipatory regret, postdecision regret; see Janis & Mann, 1968) for those high in PFI. These measures could also have a potential application in the context of interpersonal relationships. For instance, one of the most salient characteristics of the transition between the infatuation and accommodation phases of a relationship is the discovery (and hopefully acceptance) of the more negative aspects associated with a partner (Braiker & Kelley, 1979; Brickman, 1987; Levinger, 1983; Murray & Holmes, 1993). For whom would confronting the negative qualities of one's partner (see Holmes & Rempel, 1989) be most damaging? Would the drive to consistency of those high in PNS simply overwhelm the acknowledging of negativity or lead them to see virtues in faults (see Murray & Holmes, 1993)? Would these basic cognitive styles therefore affect level of relationship commitment or at least the latency with which a commitment is made? Perhaps those high in PNS might be more likely to make a commitment early on, whereas the doubt and hesitation associated with high PFI may create greater reluctance to commit to a love relationship. An applied domain that may benefit from these measures is health. A major problem among health practitioners is that patients often fail to follow prescribed drug/medical regimes (see Meichenbaum & Turk, 1987). Perhaps individuals high in PNS would be more likely to keep appointments and follow medication instructions. More interestingly, would high-PNS patients only be likely to do so if the medical and drug instructions were simple in nature? One additional benefit of this study would be to indicate an instance when it would be advantageous for an individual to possess a high Personal Need for Structure. At the Defence and Civil Institute of Environmental Medicine (DCIEM), we are exploring the relation of PNS and PFI to successful leadership styles. It may be that a high-PNS leader may be seen as too domineering and resistant to alternative ideas. In contrast, a team leader who is high in PFI may be seen as too indecisive to be effective. The more sophisticated hypothesis would be that there are certain circumstances that would be conducive to each style. For instance, in a crisis situation, team members might be most likely to follow and have respect for the archetypical decisive and confident leader. A related issue is whether team decision making will benefit or suffer from the cognitive styles of members or from particular combinations of cognitive styles among team members. One could conceive of situations in which the decision-making process of a two-person crew consisting of individuals high in PNS would exacerbate the quick closure and high confidence that mark this cognitive style. Should a team be comprised of individuals high in PFI, the hesitation and doubt characteristic of this style might also be magnified. Depending on the parameters of the decision-making situation (e.g., in an aviation setting where timely, correct decisions are imperative), premature closure or too long a decision latency could have disastrous results. Indeed, we have preliminary data suggesting that individuals' typical decision-making styles affect both the quality and latency of team-level judgments. We found that the judgments of military reservists performing a team naval assessment simulation were more accurate and timely when both leader and subordinates were low in PNS (Thompson & Baranski, 1999). We are also interested in exploring other real-world effects of these cognitive styles. This ability or inability to deal with ambiguity may also be manifested in problems dealing with ambiguous issues in certain occupations or situations. One such ambiguous situation is military rules of engagement during peace support operations (Litz, Orsillo, Friedman, Ehlich, & Batres, 1997). Thus, a further real-world application is determining whether those soldiers possessing high PNS might have more problems dealing with often ambiguous rules of engagement or with other cultures while on deployment. CONCLUDING THOUGHTS Over 15 years ago, Blass (1984) documented a trend toward the convergence of social and personality psy-

< previous page

page_37

next page >

< previous page

page_38

next page >

Page 38 chology. He noted 11 social-psychological constructs that had undergone reformulation from an individual differences perspective (e.g., self-consciousness;, Duval & Wicklund, 1972; Wicklund, 1975; the need for uniqueness; Fromkin, 1972; Snyder & Fromkin, 1977; the belief in a just world; Lerner & Simmons, 1966; Rubin & Peplau, 1975). This interactionist perspective was thought to reduce the variance left unaccounted for in traditional social-psychological paradigms, and thus to better approximate real-world phenomena (Blass, 1984; Bowers, 1973; Ekehammer, 1974; Endler & Magnusson, 1976). The PNS and PFI measures continue this tradition by addressing individual difference factors that may affect and interact with knowledge-seeking and decisionmaking processes in important ways. In the introduction of this chapter, we summarized two philosophical positions concerning the seeking of knowledge. The associationist school emphasized the seeking of some ultimate objective truth, whereas the pragmatist view suggested that people would be content applying a satisficing rule (see Payne, Bettman, & Johnson, 1993; Simon, 1982) of knowledge seeking. The perspective we present here corroborates the approach of dual processing models in reconciling the philosophical distinctions of the associationist and pragmatist schools. In the present case, we would characterize these distinctions as reflecting stable knowledge-seeking orientations captured by the personal need for structure and the personal fear of invalidity scales – orientations that are both implicated in a more comprehensive overarching epistemological process. These measures also speak to the broader issue of whether humans are essentially cognitive or motivational beings. Advocates of synergism propose a dynamic interreliance between motivation and cognition: Attempts to ascribe primacy to one or the other system fail to do justice to this intricate relationship (see Sorrentino & Higgins, 1986). In this vein, Tetlock and Levi (1982) suggested that determining “how people reconcile their desires with their beliefs about the world’’ (p. 84) reflects a truer, pluralistic view of human nature. This is precisely what our measures attempt to do by (a) detailing cognitive activities in their focus on knowledge structures (i.e., cognitions), and (b) by delineating important motivational moderators of this process. Thus, ultimately we offer the PNS and PFI scales as useful tools in undertakings that seek to reveal the ways in which people come to understand many aspects of their world, and detailing the mechanisms by which tentative hypotheses become transformed into unquestioned beliefs and facts. ACKNOWLEDGMENT This research was supported by Social Sciences and Humanities Research Council of Canada doctoral fellowships to Megan M. Thompson and Michael E. Naccarato, and a Natural Science and Engineering Research Council Visiting Scientist in Canadian Government Laboratory Postdoctoral fellowship to Megan M. Thompson. We thank Joseph Baranski and Luigi Pasto for the helpful comments on earlier drafts of this chapter. We are indebted to Steve Neuberg for his guidance and encouragement in the completion of this work. ENDNOTES 1. Need for structure/closure has also been operationalized via the soliciting global judgments (Freund, Kruglanski, & Schpitzajzen, 1985), and environmental noise (Kruglanski & Webster, 1991). 2. A decision bolstered by one of us having had the pleasure of hearing Stanley Schachter eloquently articulate this same logic during a colloquium discussion at Columbia University in 1990. 3. An initial psychometric iteration of 62 PNS and 54 PFI items was completed and formed the basis of these 25- and 30-item scales. The results of this initial iteration are available from the first author. 4. We use the identifiers PNS(A) and PFI(A) to refer to the first iteration of the scales and PNS(B) and PFI(B) to refer to the second iteration of the scales. 5. Factor analyses were conducted using both pairwise and listwise deletion of missing data. Both sets of analyses yielded similar factor structures. Due to concerns regarding missing data in the earlier psychometric studies, the factor analyses results reported are based on pairwise deletion of missing data. A similar strategy was used regarding the correlational results. Again the pattern of findings was similar with listwise or pairwise deletion of missing data. 6. The positive correlation between PNS and PFI has been found to be robust across several studies and samples. Correlations between the PNS and PFI have included: r = .32, p = .0001 (n = 258) in the first psychometric study reported here, and r = .34, p = .0001 (n = 210) in the second psychometric study included in this chapter.

< previous page

page_38

next page >

< previous page

page_39

next page >

Page 39 7. All sentences were pretested (see Winter & Uleman, 1984) so that a priori links between traits and sentences were unlikely. 8. Identical effects were obtained when Pilkington and Lydon replaced their measure of prejudice with a more direct measure of psychological threat. 9. We are indebted to Steve Neuberg for the collection of the Arizona State University data on our behalf. 10 In the CFA study analyses, we used 18 items from the Intolerance of Ambiguity scale and 9 items from Rigidity About Personal Habits scale because they improved the reliabilities of each measure. 11. Because the pattern of correlational results was similar for all Self-Consciousness subscales, only the overall score was included in this model. 12. To be as consistent as possible, we retained the correlated errors between PNSS1 and PNSS2 and between PNSS2 and PFI across the Neuberg and Kruglanski models. Note that this model provided the strongest results for the one-factor closure model.

< previous page

page_39

next page >

< previous page

page_40

next page >

page_40

next page >

Page 40 This page intentionally left blank.

< previous page

< previous page

page_41

next page >

Page 41 3 Behavioral Discrepancies and the Role of Construal Processes in Cognitive Dissonance Jeff Stone University of Arizona, Tucson Over 40 years ago, Leon Festinger (1957) introduced the theory of cognitive dissonance, a rich and fascinating theoretical framework about the interplay among behavior, cognition, and motivation. Festinger's original views and initial research on cognitive dissonance predate what many seem to recognize as the formal introduction of social cognition into the field during the 1970s (e.g., Markus & Zajonc, 1985; Wyer & Srull, 1989; Fiske & Taylor, 1991). Nevertheless, Festinger made several assumptions about the cognitive processes that underlie dissonance phenomena. For example, in his discussion of dissonance arousal, he made assumptions concerning the cognitive representation of information in memory, the accessibility of cognitions about behavior and belief, the perceived fit between these cognitions, the computation of a ratio of consistent to inconsistent cognitions, and the assessment of the relevance (i.e., applicability) and importance of the cognitions in the ratio. Festinger (1957) predicted that when these processes led to the perception of more dissonant than consonant cognitions, a negative drive state would be aroused, and people would become motivated to reduce the arousal. Dissonance reduction, in turn, could be accomplished by altering the cognitions in the dissonance ratio so that there were more consonant than dissonant relations among the elements. Thus, Festinger relied heavily on assumptions about representation and process to shape dissonance theory's novel and important predictions about social behavior. Contemporary researchers continue to investigate dissonance phenomena, in part, because many of the original assumptions about the process of dissonance arousal and reduction are the focus of considerable theoretical and empirical controversy in the field. The decades of research have produced a number of revisions to Festinger's (1957) emphasis on psychological consistency, each of which makes specific assumptions about the cognitive mechanisms that arouse and reduce cognitive dissonance (e.g., see Harmon-Jones & Mills, 1999). The various perspectives disagree over several theoretical issues, including whether aversive behavioral consequences are necessary for dissonance to be aroused (see Harmon-Jones, Brehm, Greenberg, Simon, & Nelson, 1996; Scher & Cooper, 1989; Thibodeau & Aronson, 1992) and how individual differences such as self-esteem moderate dissonance motivation (Spencer, Josephs, & Steele, 1993; Thibodeau & Aronson, 1992). The debate among the various perspectives has inspired new research (e.g., hypocrisy, see Stone, Aronson, Crain, Winslow, & Fried, 1994; trivialization, see Simon, Greenberg, & Brehm, 1995), but there currently is little agreement about the cognitive processes that underlie dissonance motivation. This chapter reviews the debate over the proper interpretation of dissonance phenomena with an eye toward how the various theoretical perspectives have treated the role of social cognition in dissonance arousal and dissonance reduction. It then presents a new process model of dissonance that was designed to synthesize the various perspectives. The goal of the new model is to highlight a cognitive process that plays an important role in how dissonance is aroused and subsequently reduced. Specifically, it argues that dissonance begins when people commit a behavior and then interpret and evaluate the meaning of what they have done. The key to understanding which motivational state follows from the assessment of behavior lies in the type of attributes and standards people rely on to evaluate the quality or appropriateness of their behavior. By focusing on the way in which people construe action, it becomes possible to predict the conditions under which behavioral discrepancies create the different motivational states that have been specified by the various perspectives on cognitive dissonance.

< previous page

page_41

next page >

< previous page

page_42

next page >

Page 42 LOOKING BACK: THE EARLY HISTORY OF SOCIAL COGNITION IN COGNITIVE DISSONANCE THEORY Few ideas in social psychology have generated as much theoretical and empirical interest as the theory of cognitive dissonance (Festinger, 1957). The premise of the original theory was elegantly simple: Inconsistency between two cognitions causes an aversive drive state similar to hunger or thirst. Festinger (1957) theorized that, like other drive states, people are motivated to reduce the discomfort associated with dissonance. However, in the case of dissonance, Festinger posited that people are motivated to restore consistency among the discrepant cognitions to reduce their discomfort. At the time dissonance theory was published, psychological consistency was a ubiquitous theme in social psychology (Heider, 1946; Osgood & Tannenbaum, 1955). Festinger conceptualized the motivation for psychological consistency as stemming from the need for accurate knowledge about what he called “reality.” In a previously unpublished early draft of dissonance theory (in Harmon-Jones & Mills, 1999), Festinger proposed that people form cognitions in two ways: directly from their own experience and indirectly from communication with others. The impact of direct and indirect experience would exert pressure on the cognitions to conform to the experience: In other words, there will be forces acting on the person to have his cognition correspond to reality as he experiences it. The result of this will be that, in general, persons will have a correct picture of the world around them in which they live. This is, of course, not surprising since the organism would have a hard time surviving if this were not the case. (Festinger, 1999, p. 356) Festinger (1957, 1999) proposed that elements of cognition – such as attitudes, beliefs, values, and feelings – typically reflect what a person actually does or feels, or they reflect what actually exists in the environment. In some situations, a person's reality may be what other people think or do; in other circumstances, reality may be what is experientially encountered or it may be what a person is told by others. Thus, Festinger appeared to assume that the motive for consistency among cognitions stemmed from a basic desire for a coherent and meaningful set of connections among oneself, other people, and the events in one's experience. Of course, Festinger was not interested in the equilibrium (as he called it) that people experience when in a state of consonance or consistency among cognitions. He was more interested in what happens when there is inconsistency between cognitions. He argued that people often have cognitive elements that deviate markedly from reality, such as when people conduct a novel behavior, learn new information, or are misled by others. When this is the case, he proposed that “the reality which impinges on a person will exert pressures in the direction of bringing the appropriate cognitive elements into correspondence with that reality” (1957, p. 11). He called this pressure cognitive dissonance. In discussing the reduction of cognitive dissonance, Festinger (1957) talked in depth about the process of cognitive or behavior change that was necessary to restore psychological consistency. A complete discussion of his views on dissonance reduction is beyond the scope of this chapter. Nevertheless, according to Festinger, the foremost determinant of how people choose to reduce dissonance is the responsiveness of these elements to reality. Although Festinger believed people were capable of self-serving distortions of reality (e.g., the smoker), he stressed that cognitive or behavior change had to mirror social and physical reality for either to reduce dissonance effectively. The overall picture painted by Festinger of the dissonance-plagued individual was that of a vigilant information processor, one who must seek a careful balance between physical and social reality on the one hand and the need to reduce an unpleasant state of arousal on the other. The original theory of cognitive dissonance was broad in scope, encompassing just about any type of cognitive inconsistency, but history suggests that Festinger was very interested in inconsistency between behavior and cognition. He noted in his early draft (1999) that, whereas cognition certainly dictates behavior because there is a “general tendency to make one's cognition and one's behavior consonant,” behavior often “steers’’ cognition (p. 358). His interest in how behavior might influence cognition is reflected in the fact that most of the original experimental research on cognitive dissonance examined an inconsistency between behavior and a specific attitude or belief (e.g., Aronson & Mills, 1959; Festinger & Carlsmith, 1959). Take, for example, one of the original experimental tests of dissonance theory – the Festinger and Carlsmith (1959) forced compliance experiment. In that experiment, participants first completed a very boring experimental task. Then, through an ingenious ruse, the experimenter asked participants if they would be willing to tell a waiting accomplice that the task was actually fun and interesting. It presented those who agreed with an inconsistency between their perception of reality (that the task was very boring) and their behavior (telling someone

< previous page

page_42

next page >

< previous page

page_43

next page >

Page 43 that the task was fun and interesting). The key manipulation in this study, however, was the amount of external incentive participants were offered for telling the lie about the task. Festinger and Carlsmith predicted that when participants were offered $20 to tell the waiting accomplice that the task was enjoyable, no dissonance would be aroused because saying something that was not true would be consistent with being paid handsomely for it. In contrast, when participants were offered only $1 to stretch the truth about the task, Festinger and Carlsmith predicted that dissonance would be aroused because, in this case, there is no clear external justification or explanation for their distortion of reality. As a result, participants would be motivated to restore consistency between their behavior and beliefs. The data show that, when the incentive for being untruthful to the waiting confederate was low, participants changed their attitudes about the task to bring them into line with their behavior. Consistent with what they had told the person who was waiting, participants came to believe that the task was fun and enjoyable. Also as predicted, those paid $20 did not alter their perception of how boring the task had been; high incentive or reward did not produce liking for the task. Thus, as predicted by dissonance theory, people would restore cognitive consistency when they acted in a way that was discrepant from their perceptions of reality. The Festinger and Carlsmith (1959) experiment proved to be more than just an intriguing finding; it inspired hundreds of experiments that helped distinguish social psychology from previous schools of thought about human behavior. At the time dissonance theory was published, the most dominant view of human behavior in psychology derived from learning theory, and psychologists were attempting to explain most human actions in terms of reinforcement contingencies. Dissonance theory made an impact because it was able to account for human behavior in situations where the relevant contingencies were supposedly not capable of eliciting behavior. Other early research designed to investigate the unique predictions made by dissonance theory revealed that people can bolster their religious beliefs in the face of disconfirming evidence (Festinger, Schacter, & Rieken, 1956), increase the value of items because they were difficult to select (Brehm, 1956), dislike others more after causing them harm (Davis & Jones, 1960), and enhance their attraction to other people for whom they have suffered (Aronson & Mills, 1959). The framework provided by dissonance theory was able to pull together a number of seemingly unrelated and perplexing social phenomena. At the same time, it showed that human beings are thinking, active participants in their environments and not simply the passive recipients of contingencies from the social context. However, at the same time dissonance theory was carving out new territory for the field of social psychology, researchers began to raise important questions about the basic assumptions of the original theory. One serious ambiguity in the theory concerned the precise nature of the cognitions underlying cognitive dissonance effects. Festinger (1957) focused primarily on inconsistencies that involved behavior without making a specific statement about the types of attitudes or beliefs that are relevant or important in a given situation. Aronson (1992) mused that this was an important problem with the theory from the outset, because even Festinger's students often had to ask Leon which cognitions would be inconsistent when applying the theory's predictions to a novel phenomenon. It was not long before the ambiguity about the cognitions responsible for dissonance led to several attempts to revise the original theory. One of the first revisions was offered by Aronson (1968; Aronson & Carlsmith, 1962), who argued that dissonance invariably involves a discrepancy between behavior and cognitions about the self. Specifically, Aronson argued that dissonance is a function of discrepancies between behavior and self-expectancies for competent and moral behavior. He predicted that people who hold positive expectancies for competent and moral behavior (e.g., people with high self-esteem) are likely to experience dissonance following behavior such as lying or poor decision making. However, people who do not expect to behave in a competent or moral fashion (e.g., people with low self-esteem) would not experience dissonance following an incompetent or immoral behavior (e.g., Aronson & Carlsmith, 1962). Aronson's focus on cognitions about the self redefined dissonance motivation as stemming from a specific need for self- consistency. Following this and other revisions (e.g., Brehm & Cohen, 1962), much of the research on dissonance in the 1960s focused on uncovering the situational antecedents that were necessary for dissonance to motivate attitude change (e.g., perceptions of choice; Linder, Cooper, & Jones, 1967), as well as expanding the theory into new domains (e.g., attraction; Darley & Berscheid, 1967) and wrestling with alternative accounts of the motivational basis of dissonance (e.g., self-perception theory; Bern, 1965). When social psychology began to turn toward questions about cognitive representation and process in the early 1970s, cognitive dissonance researchers followed suit. One important advance was the introduction of attribution principles into the process of dissonance arousal and reduction. The discovery that people could

< previous page

page_43

next page >

< previous page

page_44

next page >

Page 44 misattribute dissonance arousal to another source (Zanna & Cooper, 1974) and that attributions about personal responsibility for behavior moderate attitude change (Cooper, 1971; Wicklund & Brehm, 1976) represent two such applications. Another major advance made during the 1970s was the demonstration of the affective component of cognitive dissonance, including evidence that dissonance operates like an arousal state (Pallak & Pittman, 1972) that is necessary for dissonance-induced attitude change to occur (Cooper, Zanna, & Taves, 1978). As with other areas of theory and research in social psychology, the rise of social cognition had an important impact on the evolution of cognitive dissonance theory. According to some in the field, interest in cognitive dissonance theory lost its steam by the early 1980s (e.g., Ableson, 1983). Explanations range from the hypothesis that the “cold,” information-processing emphasis of social cognition had so dominated the field that few young researchers were interested in “hot” motivational theories (Berkowitz & Devine, 1989; Markus & Zajonc, 1985), to the problems associated with the use of deception and highimpact experimentation common in dissonance experiments (Aronson, 1992). But the idea that dissonance theory lost favor in the field overlooks some very important advances made during the 1980s, two of which involved the application of social cognition to understand the processes underlying dissonance arousal and reduction. If interest in dissonance had waned, Cooper and Fazio's (1984) “New Look” at dissonance and Steele's (1988) theory of selfaffirmation certainly revitalized theoretical and empirical intrigue in dissonance phenomena. COGNITIVE PROCESSES IN DISSONANCE AROUSAL: THE AVERSIVE CONSEQUENCES MODEL In their review of the literature, Cooper and Fazio (1984) concluded that dissonance did not appear to have anything to do with psychological consistency. The evidence gathered over the previous 20 years indicated that attitude change did not always occur when behavior and belief were inconsistent. They also proposed that the evidence supporting a role for cognitions about the self was equivocal. Cooper and Fazio argued that whereas the findings concerning self-consistency (Aronson & Carlsmith, 1962) and self-esteem effects were provocative, ‘‘each (perspective) probably has not addressed sufficient data to be a complete theory of the causes of cognitive dissonance” (p. 232). Cooper and Fazio concluded that the relevant data, based mostly on research in the forced compliance (or counterattitudinal) paradigm, indicated that dissonance occurs when people take personal responsibility for having committed a behavior that produced an aversive outcome. The motivation to justify behavior through attitude change only operates when people perceive that their behavior has created an unwanted or aversive outcome or product. In detailing the cognitive processes that underlie dissonance, Cooper and Fazio (1984) proposed a sequential model by which dissonance is aroused and subsequently reduced. In the model, dissonance arousal begins when people engage in a behavior and then immediately assess its consequences. The information used to assess the consequences of behavior includes where the outcome falls in relation to latitudes of acceptance and whether the consequences are revocable. Only when the behavioral consequences are perceived to fall outside of the latitudes of acceptance and are perceived to be irrevocable do people conclude that the behavioral outcome is aversive or unwanted. At this point, people proceed to the next step, which involves the attributional search for responsibility. Responsibility for the aversive behavioral outcome is determined by evaluating two pieces of information: choice and forseeability. Responsibility for the aversive consequence is high if people perceive that they acted under their own volition; if volition is perceived to be low, then they conclude that they had no responsibility for the act despite having committed it. In addition, people determine whether they could have foreseen the outcome either at the time they conducted the act or when they reflect back on what they have done. If they conclude that they could not have foreseen the negative consequences of their behavior, they do not perceive responsibility for the outcome. If both choice and foreseeability for the behavioral outcome are perceived to be high, then people accept responsibility for the behavioral outcome. According to the model, the acceptance of responsibility causes dissonance arousal. How people interpret the arousal, however, is a critical determinant of the nature of dissonance motivation. Cooper and Fazio (1984) suggested that the initial arousal that arises from acceptance of responsibility for an aversive outcome is general and undifferentiated. Consistent with the two-factor theory of emotion (Schacter & Singer, 1962), interpretation of the arousal is dependent on cues in the context. For dissonance arousal to motivate cognitive or behavior change, it must be interpreted as a negative psychological state (e.g., psychological discomfort). If people label the arousal as positive or negative but attribute it to another source, the arousal will not cause the motivation that precipitates

< previous page

page_44

next page >

< previous page

page_45

next page >

Page 45 cognitive or behavior change. Evidence for these suppositions comes from several lines of research, including the use of the misattribution paradigm, which has demonstrated that people are capable of misattributing the arousal caused by their behavior to other internal sources, such as to a pill they believe is arousing (Zanna & Cooper, 1974), or to external sources, such as a booth said to create feelings of claustrophobia (Fazio, Zanna, & Cooper, 1977). It has also been shown that people can interpret the arousal state as positive when humor or other cues associated with pleasant states are present (e.g., Cooper, Fazio, & Rhodewalt, 1978; Rhodewalt & Comer, 1979). According to the Cooper and Fazio (1984) model, once labeled negatively and attributed to the actor, dissonance arousal becomes dissonance motivation – the pressure to alter one's perception of the behavioral outcome. If the consequences of a counterattitudinal behavior are irrevocable, Cooper and Fazio proposed that people are motivated to change their attitude about the outcome to convince themselves that it was not as aversive as initially perceived. Thus, the motivation to reduce dissonance lies squarely in the service of rendering the consequences of behavior as nonaversive. Limitations to the Processes Specified by the Aversive Consequences Model Several researchers criticized the aversive consequences attributional model for presenting an overly narrow picture of cognitive dissonance phenomena (e.g., Berkowitz & Devine, 1989; Thibodeau & Aronson, 1992; see also Aronson, 1992). The model also inspired new empirical work on dissonance, some of which was designed specifically to test the necessity of aversive consequences for arousing cognitive dissonance. Two findings, in particular, raise questions about the aversive consequences process model. There is evidence suggesting that cognitions about the self play an important role in dissonance processes, and there is evidence that aversive consequences may not be necessary for dissonance arousal to occur. First, a number of studies show that cognitions about the self can moderate dissonance processes. For example, Aronson and Carlsmith (1962) showed that people with negative self-expectancies would not experience dissonance following an unwanted test performance, which was later shown to be moderated by how certain people are of a negative self-view (Maracek & Mettee, 1972). More recently, Steele, Spencer, and Lynch (1993) reported that when primed to think about self-attributes before making a difficult decision, people with high self-esteem did not change their attitudes to justify their decision, whereas those with low self-esteem showed significant post-decision justification. Similarly, Prislin and Poole (1996) showed that people with high and moderate self-esteem reacted differently to the negative consequences of making a counterattitudinal advocacy. Although self-concept effects have been difficult to replicate (e.g., Cooper & Duncan, 1971; Ward & Sandvold, 1963), there is enough past (e.g., Glass, 1964) and present (e.g., Stone, 1999) evidence to suggest that there may be conditions under which idiosyncratic cognitions about the self operate in dissonance processes. Second, two contemporary lines of research suggest an aversive negative product for behavior is not necessary for dissonance to be aroused. One challenge comes from the work on hypocrisy (e.g., Aronson, Fried, & Stone, 1991; see also Aronson, 1992). Hypocrisy was operationalized as a situation in which people make a proattitudinal statement about the importance of performing a specific prosocial behavior, such as the use of condoms to prevent AIDS (Stone et al., 1994), conserving water during a drought (Dickerson et al., 1992) or recycling (Fried & Aronson, 1995). By itself, the proattitudinal statement does not arouse dissonance because it is neither inconsistent with one's beliefs nor capable of producing an aversive outcome. However, dissonance can occur when participants are then made mindful of the fact that they do not perform the behavior they have advocated to others. The discrepancy between behavior and belief, as in the classic dissonance paradigms, is predicted to arouse dissonance. As a result, participants become motivated to “practice what they preached” (e.g., take shorter showers, Dickerson et al., 1992; purchase condoms, Stone et al., 1994) even when other strategies for dissonance reduction are available (e.g., Stone et al., 1997). The hypocrisy paradigm raises questions about the necessity of aversive consequences for dissonance arousal because participants persuade someone else to perform a positive, prosocial behavior, only to be made aware later that they do not take their own good advice. It suggests that it is not necessary to produce a negative outcome to another individual for dissonance to be aroused. A more direct challenge to the necessity of aversive consequences was introduced in an article by Harmon-Jones et al. (1996), in which aversive outcomes were manipulated independently of attitude-discrepant behavior. For example, in one representative study, participants wrote an essay under conditions of high or low choice, in which they proposed that a foul-tasting beverage – Kool-Aid mixed with vinegar – was enjoyable and refreshing. They were then told to discard their essay in the trash,

< previous page

page_45

next page >

< previous page

page_46

next page >

Page 46 eliminating any consequence for having written the essay. The results show that, despite the absence of an aversive consequence for having written the essay (i.e., there was no product for their behavior), participants in the highchoice condition showed significantly more favorable attitudes toward the foul-tasting beverage compared with those who wrote the essay under conditions of low-choice. Harmon-Jones et al. (1996) concluded that, although aversive consequences are sufficient to arouse dissonance, they are not necessary; the only necessary condition for the arousal of cognitive dissonance is psychological inconsistency (Festinger, 1957). Together, the evidence for self-concept differences and that dissonance may be present in the absence of a negative behavioral outcome suggest that the aversive consequences model does not provide a comprehensive understanding of the dissonance arousal process. It does not help us understand how dissonance can be aroused in the absence of an aversive outcome for behavior or how dissonance can be aroused for some people but not others when all have performed the same discrepant act. COGNITIVE PROCESSES IN DISSONANCE REDUCTION: SELF-AFFIRMATION EFFECTS One important advance in theory about dissonance-reduction processes emerged from research on the theory of self-affirmation (Steele, 1988). Steele proposed that dissonance experiments, like the Festinger and Carlsmith (1959) procedure, typically induce research participants to engage in actions that pose a threat to the integrity of their selfbelief system. One way to restore the integrity of the self-system is to eliminate the discrepancy by changing relevant attitudes or beliefs. According to self-affirmation theory, dissonance reduction through attitude or behavior change is just one way people go about the business of maintaining the integrity of their globally positive selfconcept. Under some conditions, justification or rationalization may not be the most efficient or desirable route for reducing dissonance. According to the theory (Steele, 1988), the primary goal of dissonance reduction is to restore the positive integrity of the entire self-system. As a result, any thought or action that restores the integrity of the self is sufficient for dissonance reduction. The novel prediction made by self-affirmation is that if a person can call on other positive aspects of his or her self-concept when threatened, dissonance will be reduced without having to confront the issue that caused the threat in the first place. To use the famous example of the smoker first used by Festinger (1957), the person who smokes despite his or her knowledge of the dangers of smoking experiences an attack on his or her positive self-integrity. The smoker's dilemma can be resolved by recalling or focusing on other aspects of the self that are highly regarded: “I may smoke,” the smoker may reason, “but I am one heck of a parent, athlete, and lover!” By bringing to mind other cherished aspects of the self, the smoker can reduce dissonance without changing the cognitions relevant to his or her discrepant behavior. A number of studies show that the accessibility of positive self-attributes can attenuate changing one of the discrepant cognitions following the arousal of dissonance. As one example, Steele and Lui (1983) induced dissonance through counterattitudinal behavior and then had half the participants complete a scale measuring socio-political values prior to completing a measure of their attitudes toward their discrepant behavior. The data show that dissonance-induced attitude change was eliminated when participants with strong sociopolitical values were allowed to reaffirm those values by completing the sociopolitical survey before their attitudes were assessed. Participants who were not value-oriented, or who did not complete the sociopolitical value measure after writing the essay, reduced dissonance by changing their attitudes. The attenuating effect on dissonance reduction of priming selfattributes indicates that people do not necessarily need to confront the discrepant cognitions directly following a discrepant behavior. If other positive self-attributes are somehow brought to mind following a discrepant act, dissonance is reduced without justification of the act. A further tenet of self-affirmation theory concerns the dispositional availability of positive self-attributes. For selfrelevant thought to attenuate attitude or behavior change, people must think about more positive than negative selfattributes following a discrepant act. People with high self-esteem presumably have more positive attributes available for affirmation compared with people with low self-esteem. Therefore, affirmation of the self should be more available and efficient for people with high compared with low self-esteem. Steele and his colleagues (Steele et al., 1993) tested this prediction in an experiment in which participants with high or low self-esteem participated in the free-choice paradigm originally reported by Brehm (1956). In Brehm's study, female participants were allowed to make a choice between two household appliances they had rated previously as equally attractive. After they made their difficult decision, when asked to re-rate the attractiveness of the items, participants rated the chosen alternative as more desirable, whereas the unchosen alternative was

< previous page

page_46

next page >

< previous page

page_47

next page >

Page 47 rated as less desirable than before the decision. According to Brehm (1956), the potential inconsistency introduced by choosing one desirable alternative over another aroused dissonance, which was then reduced by altering perceptions of the desirability of the two alternatives after the difficult choice. In the contemporary version of the free-choice paradigm by Steele et al. (1993, Experiment 2), participants were asked to choose between two desirable compact disks. To test the use of self-attributes for dissonance reduction, for some of the participants, self-attributes were primed when they completed the Rosenberg (1979) self-esteem scale before making their decision; the other participants made their decision without having their self-attributes primed. The results show that when self-attributes were primed before the dissonant act, participants with high self-esteem did not change their ratings of the alternatives, whereas participants with low self-esteem showed significant decision-justifying change in their ratings of the alternatives. In the no-prime control, both self-esteem groups showed similar levels of significant postdecision justification. Note that the Steele et al. (1993) pattern of self-esteem differences in dissonance reduction are exactly opposite what is predicted by the self-consistency revision of dissonance theory (i.e., people with high self-esteem should show more dissonance reduction that people with low self-esteem; see Thibodeau & Aronson, 1992). In summary, self-affirmation theory (Steele, 1988) proposes that the reason people change their discrepant attitudes or behavior is because these options for dissonance reduction are the only ones provided by the experimenter. If alternate strategies for affirming the self are provided, self-affirmation theory predicts that participants will use the strategy that most fully restores the integrity of the self even when it has little to do with the behavior that caused the dissonance in the first place. By showing people may have flexibility about how dissonance is reduced, the selfaffirmation perspective seriously questions whether dissonance induces a motive to restore consistency or reduces perceptions of aversiveness following a discrepant behavior. Limitations to Self-Affirmation Processes Other research, however, indicates that there are important limitations to how and when people can rely on the accessibility of positive self-attributes to avoid changing cognitions associated with a discrepant act. One qualification was published by J. Aronson, Blanton, and Cooper (1995) and Blanton, Cooper, Skurnik, and Aronson (1997). In the J. Aronson et al. (1995) research, participants completed an attitude-discrepant essay writing task in which they argued to reduce funding for handicapped services. Participants were then allowed to read the results of a personality test that they had completed prior to their participation. Specifically, they could read paragraphs that described their high standing on a number of positive self-attributes, some of which were attributes related directly to the discrepant act (e.g., compassion) and some that were unrelated to the essay (e.g., creative). The results show that in the high-dissonance conditions, participants avoided reading information about positive self-attributes that were relevant to the dissonant behavior. DesPite the fact these attributes were positive, participants preferred to read paragraphs that told them how wonderful they were on the irrelevant attributes, such as how creative they were. In a follow-up study by Blanton et al. (1997), participants were provided with positive feedback about their standing on positive traits that were either relevant (i.e., compassion) or irrelevant (i.e., creative) to the discrepant essay about handicap funding. When the positive feedback was relevant to the counterattitudinal act, participants showed significantly more attitude change compared with participants who wrote the essay but received no feedback. This suggests that the positive feedback increased dissonance arousal beyond what was induced by just writing the essay under conditions of high choice. In contrast, when the positive feedback was irrelevant to the behavior, participants showed significantly less attitude change compared with the relevant feedback and no feedback control conditions. Together, these studies indicate that people will avoid thinking about positive selfattributes that relate to the discrepancy (e.g., J. Aronson et al., 1995); when they cannot avoid exposure to related self-attributes, thinking about positive attributes will increase the need for dissonance reduction (Blanton et al., 1997). This suggests that the accessibility of positive self-attributes will only attenuate attitude or behavior change when people can bring to mind attributes that distract them from thinking about the discrepancy. In another line of research, Simon, Greenberg, and Brehm (1995) proposed that when people bring to mind other positive aspects of their self-concept, the salience of such important cognitions may cause people to trivialize or reduce the importance of the behavioral discrepancy. In a series of experiments, Simon et al. (1995) found that when people made an attitude discrepant advocacy, they were more likely to reduce the importance of what they had done when provided the opportunity to think about positive self-attributes before they were asked to indicate their attitude toward the essay topic. Thus, based on one of Festinger's (1957) original

< previous page

page_47

next page >

< previous page

page_48

next page >

Page 48 assertions about dissonance reduction, Simon et al. (1995) argued that the accessibility of positive self-attributes attenuates attitude change by reducing the importance of the behavioral cognition. Affirmation of the self may only occur if people can somehow downplay the discrepancy. A third challenge to self-affirmation theory concerns how people prefer to reduce dissonance when more than one strategy is available. Stone, Weigand, Cooper, and Aronson (1997) distinguished between two general strategies for dissonance reduction: direct, which is defined as changing one of the discrepant cognitions, and indirect, which involves strategies like self-affirmation whereby people reduce dissonance without altering directly cognitions associated with the discrepancy. Stone et al. (1997) observed that, in many self-affirmation studies, participants are provided with the opportunity to self-affirm (i.e., use an indirect strategy such as focus on positive self-attributes) before they know they will have an opportunity to change their attitudes or behavior (i.e., use a direct strategy). Finding that people will use an indirect strategy for dissonance reduction before they are offered the use of a direct strategy does not tell us much about which strategy they prefer to use for dissonance reduction. It may be that people usually want to resolve the discrepancy directly through attitude or behavior change. However, if they perceive that the indirect affirmation option is the only option available, they may use it to reduce their discomfort, but they may then feel regret when the second option presents itself. To investigate preferences among different reduction strategies, Stone et al. (1997) offered more than one strategy at a time and allowed participants to choose following an induction of hypocrisy. In the first experiment, participants were made to feel hypocritical about their practice of safer sexual behavior to prevent AIDS (e.g., Stone et al., 1994). In one condition, participants were then provided with an indirect, self-affirming option for dissonance reduction: They were offered an opportunity to donate their subject payment to a homeless shelter. In another condition, participants were offered the opportunity to donate to the homeless shelter, but before they used this option, they were provided with a direct strategy for dissonance reduction: They were offered the opportunity to use their subject payment to purchase condoms. The results show that when offered only the indirect option, fully 83% of those in the hypocrisy condition donated to the homeless shelter. This indicates that when it is the only option available, people will use an indirect self-affirmation strategy for dissonance reduction. However, when the indirect strategy was offered alongside the direct strategy, fully 78% chose to purchase condoms, whereas only 13% chose to donate to the homeless shelter. A follow-up study showed that participants preferred to use a direct reduction strategy even when they could self-affirm on a behavioral dimension that held more importance for their selfconcept than the direct strategy – a serious challenge to the tenets of self-affirmation theory. These studies show that, under some conditions of dissonance arousal (e.g., hypocrisy), people have a strong preference for directly resolving the discrepancy that caused their dissonance. Indirect strategies for dissonance reduction, such as selfaffirmation, are useful when they are the only means available. However, if given a choice, people sometimes prefer to reduce dissonance by altering the cognitions that contributed to their discomfort. Together, these studies suggest that self-affirmation theory does not provide a comprehensive understanding of the processes underlying the reduction of cognitive dissonance. People may avoid thinking about positive aspects of the self if they remind them of the behavioral discrepancy. When they do think about positive attributes that are unrelated to the discrepancy, such thinking may reduce dissonance by shifting attention away from the discrepancy (Cooper, 1999) or by providing a frame by which people can reduce the importance of the behavioral discrepancy (e.g., Simon et al., 1995). Even when other positive aspects of the self are available for use as a reduction route, people sometimes choose to confront the discrepancy head on and change the relevant cognitions or behavior, thereby erasing the discrepancy altogether (e.g., Stone et al., 1997). CONSTRUAL PROCESSES IN DISSONANCE: THE INTERPRETATION AND EVALUATION OF ACTION When put all together, the 40 years of research and deliberation over the proper interpretation of dissonance phenomena resembles a giant, unfinished jigsaw puzzle. Several researchers have constructed some pieces or even large sections of the puzzle, but the debate over the necessary conditions for dissonance to operate has obscured the overall picture. As Berkowitz and Devine (1989) noted, the focus on analysis over synthesis has failed to produce an integrated portrait of the entire dissonance arousal and reduction process. In a recent article, Stone and Cooper (2000) argued that one way to integrate all the pieces and begin to see the big picture is to assume that each of the contemporary perspectives describes an important part of the cognitive dissonance process. For example, it is quite reasonable to suppose that, under some conditions, people are moti-

< previous page

page_48

next page >

< previous page

page_49

next page >

Page 49 vated by a desire to uphold important attitudes (e.g., Harmon-Jones et al., 1996) or beliefs they have about themselves (i.e., self-consistency). It is equally tenable that, under some conditions, people are motivated to reduce the perceived aversive consequences of their unwanted behavior (i.e., the “New Look”). It is also the case that, under some conditions, people would prefer to think about other positive aspects of themselves rather than face the implications of their behavior (i.e., self-affirmation). If we assume that behavioral discrepancies are capable of inducing different motivational states, then we can see that the different perspectives are neither synonymous nor simple linguistic translations of one another. Rather, they each describe a distinct and important piece of the overall dissonance process and, in doing so, make a unique contribution to our understanding of cognitive dissonance arousal and reduction. Nevertheless, to bring these various perspectives under one roof requires looking for what they have in common. Indeed, all of the contemporary perspectives on dissonance theory have at least one common bond: Each makes an important assumption about how people assess the psychological meaning of their behavior. Specifically, every major perspective on cognitive dissonance, starting with Festinger (1957), assumes that dissonance begins when people commit an act and then try to make sense of it, which they accomplish by determining how their behavior fits with other relevant cognitions. However, because each makes a specific assumption about the cognitions people use to make sense of their behavior, each perspective on dissonance offers its own interpretation of dissonance motivation. Thus, the various models share the assumption that dissonance begins with the detection of a discrepancy between behavior and some preexisting set of cognitions. Where the models appear to differ most is in their assumptions about the content of the preexisting cognitions, especially those related to the self, that people use when deciphering their behavior. Stone and Cooper (2000) proposed a new model to account for the processes by which people determine that their behavior is discrepant from other relevant cognitions – the first step in many of the models of dissonance. According to the self-standards model , people may determine the fit between cognitions by comparing the outcomes of their behavior against cognitive representations for how they should have or could have conducted themselves. That is, dissonance processes start with a comparison between behavior and a guide for behavior, which can include cognitive representations of various attributes (e.g., attitudes and beliefs) or self-standards (e.g., Higgins, 1989a; 1990). The basic assumption of the self-standards model is that the motivational basis of dissonance – why and how people reduce dissonance – depends in part upon the attributes or standards people use to interpret and evaluate what they have done. An important assumption of the self-standards model is that any given behavior may be construed in a multitude of ways, for example, as relevant to one's attitudes, self-concept, or the rules of a given society or culture. The model predicts that how people construe the meaning of a given behavior, and the specific discrepancy that they perceive as a result, determines the nature of the arousal and motivation that follows (see Blanton, chap. 5, this volume; Vallacher, 1992). There are two contemporary areas of research and theory in social cognition that offer support for the assumptions made by the present model. The first has to do with how people interpret or define action; the second has to do with the way in which people evaluate their behavior. The Interpretation of Behavior Multiple lines of research in social cognition have made the observation that the interpretation and evaluation of behavior is a malleable state of affairs. When people act, the meaning they ascribe to what they have done can be influenced by the context in which it was conducted. Vallacher and Wegner (1985; see Vallacher & Kaufman, 1996) proposed a hierarchical model of action identification that specifies that any given action can be identified at various levels of interpretation, ranging from detailed, mechanistic descriptions of behavior that explain how an act is completed, to more comprehensive and abstract identities that explain why a behavior was conducted and with what effect (i.e., consequences for behavior). To illustrate, imagine a researcher using a computer to analyze data or write a paper. The same activity on the computer can be thought of as “punching the number pad’’ and “typing letters on the keyboard” (low-level identifications), or as “entering the data” or “writing a paper” (intermediate-level identifications), or as “conducting research” and “advancing my career” (high-level identification). Like other models of self-regulation (e.g., Pyszczynski & Greenberg, 1992; Scheier & Carver, 1988), Vallacher and Wegner's model of action identification assumes low-level identities (e.g., typing another manuscript) provide the information necessary for reaching the higher-level identities (i.e.,. advancing my career), and higher-level identifications (i.e., advancing my career) serve as goals to direct the lower-level activities (writing another manuscript). Higher-level action identities often define the actor as well as the action. Vallacher and Wegner suggested

< previous page

page_49

next page >

< previous page

page_50

next page >

Page 50 that as action identities become more comprehensive and abstract, “doing” becomes “being”; behavior identified at the highest levels becomes self-defining. ‘‘Playing the piano” becomes “expressing myself” and then “being a musician.” The experience of successfully maintaining lower-level identities (e.g., “playing the piano”) facilitates movement up the identification hierarchy (e.g., “expressing myself”) until the behavior represents an important selfconcept dimension (e.g., “I'm a musician”). Within the hierarchy of possible identities, people tend to focus on only one identity for their behavior at a given time. The level of identification that becomes “prepotent” in thinking about behavior is determined by the difficulty of the behavior (Vallacher & Wegner, 1985). Behaviors that are unfamiliar, difficult, require effort, or are timeconsuming tend to be identified at a lower level in the hierarchy. The logic of this claim is largely functional; thinking about the mechanics of an unfamiliar or difficult action facilitates smooth and successful performance of the act. For example, when dancing a waltz for the first time, the novice bride and groom have to think in terms of the mechanics of the dance to successfully execute each step. Despite the functional value of thinking about new or difficult acts in terms of the lower-level mechanics, action identification theory (Vallacher & Wegner, 1985) maintains that people are not very content with mechanistic representations of behavior. Even when a behavior is personally difficult, people prefer to think about it at a level that provides comprehensive psychological meaning. The interesting observation made here is that, because people prefer to think about their behavior at a more comprehensive and abstract level, when they are conducting an act that requires low-level identification they are susceptible to new ways of conceptualizing what they are doing. They become acutely sensitive to the context surrounding the act, and they embrace whatever higher-level identity cues are provided by the situation (Vallacher, 1992). Consequently, the emergence of more abstract and comprehensive interpretations for behavior are most likely to occur when behavior is personally difficult and a situational cue makes accessible a higher-level identity for understanding what they are doing. What theory and research on how people interpret their actions offers to our understanding of dissonance processes is the observation that the construal of one's own behavior is malleable and dependent on the context in which the behavior occurs. This may especially be the case when people engage in acts that are novel, difficult, or require extensive cognitive effort. To the degree that a given behavior requires low-level identification for successful execution, interpretation of what the act means to and for the individual is particularly susceptible to cues provided in the behavioral context. The implication is that any of the classic dissonance-arousing acts, such as making a difficult decision (Brehm, 1956), attempting to persuade a confederate about the nature of a task (Festinger & Carlsmith, 1959), or having to construct a counterattitudinal essay (e.g., Scher & Cooper, 1989) could be prone to the emergence process. Consequently, the interpretation and evaluation of these acts could be influenced by whatever identity cues are operating in the situation (Vallacher, 1992). The construal of a particular behavior as representing a significant discrepancy should therefore be influenced by the salience and accessibility of cognitions relevant to the evaluation of behavior. Salience and Accessibility of Attributes and Standards for Behavior Part and parcel of the debate among the various perspectives on dissonance is the disagreement over the information people use to interpret and evaluate the psychological meaning of their behavior. Once people have acted, dissonance theorists generally assume that the behavior or its outcome are evaluated against some relevant criteria. When the behavior is judged not to fit the relevant criteria, a discrepancy is detected and the dissonance process is engaged. What determines the criteria for evaluation? According to the self-standards model of dissonance (Stone & Cooper, 2000), the context in which the act occurs can make some criteria especially more salient or accessible, and therefore invoke a specific construal of the act. The specific content of the discrepancy, and the nature of the dissonance motivation that follows, may be determined by what criteria are made salient in the context in which behavior is performed. For example, the psychological consistency perspective (e.g., Festinger, 1957; Harmon-Jones et al., 1996) predicts that dissonance can occur for discrepancies between behavior and a specific attitude or belief. If in fact the construal of behavior is dependent on what cues are salient in the context, then it seems possible for dissonance to be aroused if the situation cues the use of relevant attitudes or beliefs to make sense of the behavior. When this is the case, people should identify their behavior at a relatively intermediate level of abstraction – one that does not focus on the mechanics of writing nor on the higher level identity implied by such acts (e.g., competent). For example, if asked to describe what they have done, people might say they just “advocated a position on a topic opposite my own” or “decided

< previous page

page_50

next page >

< previous page

page_51

next page >

Page 51 between two attractive alternatives.” Such interpretations of behavior may focus people on resolving the discrepancy between behavior and a specific attitude or belief relevant to the act. The motive for psychological consistency may then operate when people make an intermediate-level identification of their behavior possibly because the situation makes salient specific attitudes or beliefs and the context either does not make salient a higher-level identification or, as in the Harmon-Jones et al. (1996) procedure, eliminates any high-level meaning (i.e., by eliminating the consequences for one's act). Thus, psychological consistency may be a central motivating factor in dissonance even in the absence of an aversive outcome or product for behavior, but perhaps only when something in the context cues the use of specific attitudes or beliefs to interpret the behavior. However, there is no reason to believe that people routinely use specific attitudes or beliefs to interpret and evaluate their behavior. According to the self-standards model (Stone & Cooper, 2000), any behavior that can be perceived as discrepant from a specific attitude or belief can also be perceived as discrepant from cognitions at a higher level of abstraction, such as cognitions about the self or normative rules for conduct. When people interpret and evaluate their behavior at higher levels of meaning, the discrepancy they perceive and the dissonance motivation that arises may take on qualitatively different characteristics. From the perspective of the aversive consequences model (Cooper & Fazio, 1984), dissonance processes turn on the perception of a negative behavioral outcome. How do people determine that their behavior has reaped an aversive consequence? According to the analysis offered by Cooper and Fazio (1984), they do so by comparing the outcome of the behavior to standards for what they should have or could have achieved from the perspective of others – what might be called normative standards for behavior (e.g., Higgins, 1990). The aversive consequences model assumes that people interpret and evaluate their behavior at a higher, more abstract level of meaning than what is suggested by the psychological consistency perspective (Harmon-Jones et al., 1996). For example, when norms are salient and people are asked to identify the meaning of a counter-attitudinal act or difficult decision, they might say, “I have violated norms for social responsibility” or “I have made a socially undesirable choice.” According to the selfstandards model, when normative standards for behavior are salient, most people, regardless of individual differences in the content of their self-concept, should perceive a discrepancy between their behavior and the relevant normative standard. Consequently, dissonance should be aroused for most people (assuming that they subscribe to the salient norms), and reduction will be targeted toward reducing the negative perception of the behavioral outcome. Alternatively, a different high-level construal of behavior is assumed by the self-consistency revision, which predicts that cognitions about the self play a role in the process of dissonance arousal. Specifically, according to the selfconsistency revision (e.g., Aronson, 1968; Thibodeau & Aronson, 1992), dissonance involves a discrepancy between behavior and specific beliefs about the self-attributes of competence and morality. Moreover, because people vary in how competent and moral they expect themselves to be, the self-consistency perspective predicts that self-esteem differences should moderate for whom dissonance is aroused. Dissonance from the self-consistency viewpoint is predicated on a perceived discrepancy between behavior and idiosyncratic self-expectancies for competent and prudent behavior. A central assumption to this process concerns the cognitive elements that underlie self-expectancies. According to Thibodeau and Aronson (1992), self-expectancies for behavior are based on personal standards for morality and competence that are ‘‘culturally derived, and largely shared, by most people within a given society or subculture” (p. 596). In other words, the standards that people rely on to assess their behavior are normative in nature. If in fact the standards used to assess behavior are always derived culturally and shared by all, how can people with different self-concepts draw different conclusions about the meaning of what they have done? One possibility is that, for self-consistency processes to operate in dissonance, in addition to the relevant normative standards, people must also take into account their actual self-concept when assessing their behavior. Stone and Cooper (2000) proposed that when people commit a dissonant act, two chronic aspects of the self can become accessible in memory: a) the self-concept, represented by an actual self-attribute for competence or morality; and b) the normative self-standards for competence or morality. Together, these two cognitions can form a self-expectancy, in that people expect their behavior to confirm or verify the chronic relationship – the degree of discrepancy – between the actual self and normative standard on the dimensions of morality and competence. So, once induced to commit a dissonant act, a person might assess the competence or morality of his or her behavior by comparing the contextual self inferred from the behavior (e.g., “I just said something stupid and immoral”) against a chronic selfexpectancy (“I am

< previous page

page_51

next page >

< previous page

page_52

next page >

Page 52 usually a smart and decent person – relative to the norm for competence and morality”). As in this example, when a discrepancy is detected between the behavior and chronic self-expectancy, dissonance is aroused. However, another person with a different expectancy might not perceive a discrepancy in this case. For example, another person might infer the same contextual self from behavior (“I just said something stupid and immoral”), but conclude that it does not represent a discrepancy when compared against a less positive self-expectancy (e.g., “I am NOT ALWAYS a smart and decent person – relative to the norm for competence and morality’’). In this case, the use of selfexpectancies – the chronic degree of discrepancy between actual and normarive standards – does not lead to dissonance arousal. Thus, idiosyncratic self-expectancies can be represented in memory as a chronic discrepancy between an actual self-attribute (e.g., competence or morality) and its relevant normative standard. Dissonance is aroused when a given behavior fails to confirm the chronic relationship between the actual self-concept and its normative self-guide. What distinguishes people with positive versus negative expectancies for their behavior is the size of the discrepancy between their chronic actual-self, referred to as the self-concept , and the normative standards for competence and morality. These differences in how self-expectancies are represented (i.e., the size of the discrepancy between the chronic actual self and self-standard) seem to operate as the definition of self-esteem in the self-consistency perspective (cf. James, 1890; Rosenberg, 1979; see also Moretti & Higgins, 1990). Thus, people with high selfesteem, who have a smaller chronic discrepancy between their actual self and norms for behavior, should experience dissonance when “lying, advocating a position contrary to their own beliefs, or otherwise acting against one's principles – the stock in trade of countless dissonance experiments” (Thibodeau & Aronson, 1992, p. 592). Alternatively, those with negative expectancies, who have a larger chronic discrepancy between their actual self and normative standards, are simply not surprised when they act in a way that deviates from the norm. In the self-standards model, the activation of self-expectancies also depends on what information is made salient in the context of a given behavior. Specifically, for chronic self-expectancies to moderate dissonance, the normative standard cannot be the only salient criterion by which people evaluate their behavior. In addition to the norms, an idiographically based, unique conception of the self-concept (e.g., actual self-attributes) must also become accessible during the construal of behavior. For example, as supposed by the self-consistency perspective (Thibodeau & Aronson, 1992), if people were to use their own personal standards to assess their behavior, they may activate their chronic self-expectancies and then self-esteem differences in dissonance reduction would emerge. Otherwise, if idiosyncratic actual self-attributes are not made accessible when assessing behavior, then most people, regardless of their self-expectancies, should perceive the behavior as discrepant from the norm, and dissonance should manifest itself without self-concept moderation of the arousal process(Cooper & Fazio, 1984). Finally, unlike the other views of dissonance, self-affirmation theory (e.g., Spencer, Josephs, & Steele, 1993; Steele, 1988) does not define how people determine a threat to global self-integrity, nor does it define the cognitive representation of a self-threat. It focuses primarily on how people cope with dissonance once it is aroused. As a result, it is conceivable that the affirmation process could follow the detection of discrepancies that involve attitudes, self-expectancies, or normative standards, although as noted earlier, there are important limitations to how the accessibility of positive self-attributes can reduce dissonance (e.g., Blanton et al., 1997; for full discussion of affirmation processes in the self-standards model, see Stone & Cooper, 2000). In summary, each of the perspectives assumes that dissonance is aroused when people detect a discrepancy between their behavior and specific attributes or standards for behavior. The malleability with which people interpret and evaluate their behavior, however, suggests that the construal process may not always rely on any one source of information, but instead may be dependent on what criteria are salient or become accessible in the context in which the behavior occurs. Consequently, the nature of the dissonance motivation that follows from a behavioral discrepancy should be a function of the attributes and standards people rely on to interpret and evaluate their behavior. THE CONSTRUAL OF BEHAVIORAL DISCREPANCIES: EMPIRICAL EVIDENCE The approach to testing the predictions made by the self-standards model has thus far involved making specific attributes and standards accessible in the context of one of the classic dissonance-arousing behaviors. If the interpretation of behavior is subject to whatever attributes or standards are brought to mind in the situation, it should be possible to manipulate independently the motivation for psychological consistency, self-consistency, selfaffirmation, or the reduction of aversive consequences by priming the specific criteria each

< previous page

page_52

next page >

< previous page

page_53

next page >

Page 53 perspective assumes people use to understand their behavior. Evidence that different motives were engaged by priming various attributes and standards is shown in part by the pattern of dissonance reduction (e.g., attitude change) across groups of participants with different levels of self-esteem. In one of the first experiments designed to examine how the accessibility of self-attributes and standards in the context of a classic dissonance-arousing behavior might influence the processes, Stone (1999) had participants with high or low self-esteem (as measured in a pretest session) conduct a free-choice paradigm based on the procedures reported by Steele et al. (1993). The accessibility of self-attributes was manipulated for some participants when they completed the Rosenberg (1979) self-esteem scale just before they made their choice; other participants were primed for self-attributes in the same way just after they made their decision. The results presented in Table 3.1 show that both self-esteem groups reported significant change in their perceptions of the alternatives in the no prime, decision-only control condition. This suggests that when selfattributes were not primed, both groups appeared to be focused on the inconsistency inherent in their decision and were motivated to change their attitudes about the alternatives (e.g., Festinger, 1957). However, when primed for their self-attributes before the choice, participants with high self-esteem showed significantly less justification of their decision compared with participants with low self-esteem. When primed after they made their decision, participants with high self-esteem showed significant post-decision justification, whereas participants with low self-esteem did not. The results in the priming conditions suggest that when self-attributes were primed before the decision, they operated as self-affirming resources (Steele et al., 1993) and reduced the need for decision justification among those participants with high (but not low) self-esteem. In contrast, when self-attributes were primed after the difficult decision, they appeared to engage self-expectancies and the motivation TABLE 3.1: The Interaction Between Self-Esteem and the Timing of the Self-Attribute Prime on the Spread of Alternatives (Reported in Stone, 1999). Self-EsteemTiming Of Self-Attribute PrimeNoneBeforeDecisionAfterDecisionHigh1.350.551.35Low1.522.480.52Note. Higher numbers indicate more dissonance reduction. for self-consistency (e.g., Aronson, 1968). As a result, those with high self-esteem showed more dissonance reduction than those with low self-esteem. This study shows that the nature of dissonance motivation can be a function of when people think about themselves in the context of making a difficult decision. In another series of studies, Stone, Galinsky, and Cooper (1999) examined how the accessibility of different selfstandards for behavior would influence the arousal and reduction of dissonance. Specifically, in one experiment based on the free-choice paradigm described earlier, participants with high or low self-esteem were asked to make a difficult choice between two desirable alternatives and then complete a measure of “self-understanding.” The task was designed to prime different standards for behavior by having participants write a short description of a fictitious target person who exemplified two positive self-attributes such as competent and rational. To manipulate the type of standards made accessible by this task, some participants wrote about the target person using their own personal standards. Specifically, they were instructed to think about the standards for competence and rationality from their own perspective and then describe the target person using their own personal standards for each trait. Other participants wrote about the target person using normative standards. They were directed to think about the same standards from a societal perspective and describe the target using “the standards for what most people think” represent each attribute. Once they completed the self-standard priming task, the experimenter had them rate the psychology studies for desirability again. As seen in Table 3.2, participants with high and low self-esteem in the decision-only control condition showed significant levels of post-decision justification. As noted previously, the free-choice paradigm, which was created to investigate psychological consistency in the early days of dissonance research, appears to focus people on the discrepancy between the choice and specific beliefs about the alternatives (e.g., Festinger, 1957). However, when primed for their own personal standards, participants with high self-esteem reported significantly more post-decision justification of their choice compared with the justification reported by participants with low self-esteem. In contrast, when normative standards were made salient, both self-esteem groups reported relatively equalivalent levels of postdecision justification. Thus, as predicted by the self-standards model (Stone & Cooper, 2000), making personal standards for behavior salient engaged self-expectancies and caused participants with high self-esteem to seek more dissonance reduction than partici-

< previous page

page_53

next page >

< previous page

page_54

next page >

Page 54 pants with low self-esteem. In contrast, making salient the norms for behavior caused the same level of decision justification for both self-esteem groups. Stone et al. (1999) conducted a second experiment to examine the effects of self-standards primes in another classic dissonance paradigm – one involving counterattitudinal advocacy. In addition, several measures were included in an attempt to capture mediational evidence of the hypothesized processes underlying the various predictions made by the standards model. For example, measures were included to tap the valence of the self-relevant thought induced by the standards primes. According to the self-standards model (e.g., Stone & Cooper, 2000), when personal standards are salient in the context of a discrepant act, people with low self-esteem, who presumably hold less positive expectancies for their behavior, should show evidence of more negative self-relevant thought, which should correspond with less self-justification after a discrepant act. In contrast, people with high self-esteem, who presumably hold more positive expectancies for their behavior, should show evidence of more positive self-relevant thought, and this should correlate with more justification of the discrepant act. Furthermore, the model predicts that when normative standards are salient, most people, regardless of their prevailing self-esteem, should be focused on the positive normative standards for behavior. Consequently, the content of self-relevant thought should be positive for those with high or low self-esteem when normative standards are made salient, and this should correlate with the level of justification following the discrepant act. These predictions were tested by having participants circle traits that applied to each type of attribute when they were primed for personal versus normative standards following the induced compliance task. Self-report measures of negative and positive affect were also collected. Elliot and Devine (1994) presented evidence that self-reports of psychological discomfort were enhanced by writing a counter-attitudinal essay TABLE 3.2: The Self-Esteem X Standards Prime Interaction on the Spread of Alternatives (in Experiment 1 of Stone, Galinsky, & Cooper, 1999). Self-EsteemSelf-Standards PrimeNonePersonalStandardsNormativeStandardsHigh0.881.770.55Low1.080.920.86Note. Higher numbers indicate more dissonance reduction. under high-choice conditions, and Galinsky, Stone, and Cooper (2000) found that psychological discomfort was reduced by focus on irrelevant positive attributes following dissonance arousal. According to the proposed selfstandards model of dissonance, when personal standards are salient, people with high self-esteem should experience more psychological discomfort and more negative affect directed at the self compared with people with low selfesteem. In contrast, when normative standards are salient, both those with high or low self-esteem should experience psychological discomfort and negative self-affect associated with dissonance and the salience of the normative prime. Thus, to investigate the potential mediating role of negative and positive affect, after participants completed the induced compliance task (and in some conditions, the standards primes), they completed a measure of their current affective state. In Experiment 2 (Stone et al., 1999), participants with high or low self-esteem made a counterattitudinal advocacy under conditions of high or low choice. Specifically, participants were asked to generate arguments in favor of a tuition increase at the University of Arizona (e.g., Simon, Greenberg, & Brehm, 1995). Once they had completed the advocacy task and choice manipulation, those randomly assigned to the standards prime condition were asked to examine a list of 30 trait words that ranged from negative to positive (Anderson, 1968). The traits were chosen for their relevance to the counterattitudinal argument task (e.g., Blanton et al., 1997). Those randomly assigned to the Normarive Standards Prime condition were told the list of traits represented various normative standards for behavior. They were told to examine the list and think about which traits represent the type of person they should be from the perspective of others. They were then instructed to circle the traits on the list that best represent the normative standards guiding their behavior. Those randomly assigned to the Personal Standards Prime condition were told the list of traits represented various personal standards for behavior. They were told to examine the list and think about which traits represent the type of person they personally wanted to be. They were then asked to circle the traits that best represented the personal standards guiding their behavior. After the standards prime manipulations, participants received a questionnaire designed to measure three affective states: psychological discomfort, negative self-affect, and positive affect (Elliot & Devine, 1994; Galinsky, et al., 2000). After participants completed the affect measures, the experimenter had them complete the attitude questionnaire ostensibly because he forgot to have them complete it earlier. The primary measure

< previous page

page_54

next page >

< previous page

page_55

next page >

Page 55 asked participants to indicate their agreement with the statement “I believe that tuition should be raised for the next academic year.”1 Analysis of the attitude change scores shown in Table 3.3 reveals a significant interaction between self-esteem and the self-standards priming conditions. Planned contrasts showed that participants in the high-choice control condition tended to justify their advocacy more ( M = 1.80) than low-choice control participants ( M = 1.06), and this was not moderated by self-esteem. Further contrasts revealed that attitude change was significantly moderated by selfesteem when self-standards were primed. Those with low self-esteem showed significantly less attitude change compared with participants with high self-esteem when personal standards were primed, and they showed less attitude change compared with those with high or low self-esteem when normative standards were made salient. There were no differences found between the other standards prime conditions. Analyses of the mediational measures of affect and trait listings did not reveal reliable differences across the experimental conditions. However, within-cell correlations showed that, among participants with high self-esteem and personal standard primes, attitude change was positively correlated with both feelings of psychological discomfort ( r=0.35, p

Page 58 that follows from a specific interpretation and evaluation of action, dissonance theory can move forward in new directions that continue to present important insights into human social behavior. ENDNOTES 1. Also included was a question designed to measure perceptions of choice to write the essay. Analysis of this item indicates that those in the low-choice condition perceived significantly less choice to write the essay compared with participants in the high choice conditions.

< previous page

page_58

next page >

< previous page

page_59

next page >

Page 59 4 Self-Evaluation: The Development of Sex Differences Eva M. Pomerantz Jill L. Saxon Gwen A. Kenney University of Illinois, Urbana-Champaign How people evaluate themselves has a number of important implications for their psychological functioning. Selfevaluation appears to influence the manner in which people deal with challenge (see Bandura, 1994; Eccles, Wigfield, & Schiefele, 1998, for reviews), as well as people's emotional experiences (see Abramson, Metalsky, & Alloy, 1989; Higgins, 1987, for reviews). Hence, a great deal of attention has been directed toward understanding how people evaluate their competencies and expectancies for future performance. In addition to the question of the consequences of self-evaluation, several other questions have guided work on the topic. What motives drive the manner in which people evaluate themselves? What aspects of the situation influence how people evaluate themselves? What characteristics of the person determine the form that self-evaluation takes? The effort to answer questions such as these in social psychology began with James (1890), continued on with Festinger (1954a), and is evident in much current social-cognitive work (see Eccles et al., 1998; Suls & Wills, 1991; Swann, 1990; Taylor & Brown, 1998; Trope, 1986; Wood, 1989, for reviews). Although these questions are important, another question of import is whether there are sex differences in selfevaluation. This question has taken on particular significance for investigators attempting to elucidate sex differences in psychological functioning (e.g., Dweck, 1986; Eccles, 1984; Nolen-Hoeksema & Girgus, 1994; Ruble, Greulich, Pomerantz, & Gochberg, 1993). The issue of sex differences in self-evaluation is at the heart of the present chapter. To situate the work on sex differences in self-evaluation, the chapter begins with a review of self-evaluation as it has been studied more generally. This is followed by a review of work on sex differences in self-evaluation. Subsequently, an approach that may be used to guide future research on sex differences in self-evaluation and its consequences for psychological functioning is presented. We end by briefly highlighting some of the key differences in work on selfevaluation in general and work on sex differences in self-evaluation. We also suggest how the two might be better integrated. A GENERAL OVERVIEW OF SELF-EVALUATION Motives Underlying Self-Evaluation One of the most central questions guiding work concerned with self-evaluation is that of what motivates selfevaluation. That is, why do people evaluate themselves? What purpose does it serve? Until recently, this has been a hotly contested issue. In his seminal theory of social comparison, Festinger (1954a) initially proposed that people have a need to accurately evaluate themselves to reduce any uncertainty they have about themselves. He argued that, in an effort to meet this self-assessment need, people seek to compare themselves with similar others. This concern with self-assessment is also evident in Carver and Scheier's (1998) model of self-regulation. Carver and Scheier proposed that people desire to assess their progress toward their goals in an effort to ensure that they eventually meet their goals. Only by accurately assessing their progress can people change ineffective strategies and adopt effective ones (see also Ruble & Frey, 1991). Although Festinger viewed the self-assessment motive as the central force guiding self-evaluation, he also suggested that people are motivated by self-improvement. As a consequence, according to Festinger, people compare themselves not only with similar but also with superior others. Such upward comparison has been well documented (see Taylor & Lobel, 1989, for a review) and actually appears to foster self-improvement (e.g., Blanton, Buunk, Gibbons, Kuyper, & Hans, 1999). As investigators further explored the motives that drive self-evaluation, two other motives came to be viewed as central. Although research subsequent to Festinger's theory on social comparison provided evidence for both the self-assessment and self-improvement motives, it also suggested that these two motives were in no way the only motives guiding self-evaluation (see

< previous page

page_59

next page >

< previous page

page_60

next page >

Page 60 Wood, 1989, for a review). Indeed, a number of investigators have argued that people are motivated to evaluate themselves by a need to feel good about themselves (e.g., Epstein, 1973; Taylor & Brown, 1988; Tesser, 1988). Evidence for this self-enhancement motive has been provided in a number of studies showing that people go out of their way to ensure that they view themselves in a positive light, often comparing themselves to inferior others (see Taylor & Brown, 1988, for a review). Swann (1990) introduced another motive playing a role in self-evaluation. He proposed that the central goal guiding self-evaluation was the desire to verify one's self-views (see also Secord & Backman, 1965). People seek to preserve their views of themselves in an effort to predict and control their environment. Swann has provided a great deal of evidence for self-verification by showing that people prefer positive feedback in areas in which they possess positive views of themselves, but negative feedback in areas in which they possess negative views of themselves (e.g., Swann, Pelham, & Krull, 1989). Two other motives have also been identified, although they have received less emphasis in the work on selfevaluation than the other motives. Self-evaluation may sometimes be guided by an effort to solidify one's identity (e.g., Tajfel, 1978; Wicklund & Gollwitzer, 1982). For example, females may be willing to evaluate themselves negatively in male stereotypical domains, such as math, to heighten their feelings of femininity. When self-evaluation is public, people may be guided by the desire to make a good impression (e.g., Baumeister, 1982; Paulhus, 1982). Such a desire may often be driven by the need to obtain the approval of others (Baumeister, 1982). The selfpresentation motive may often manifest itself as self-enhancement, as people attempt to present themselves in a positive manner. However, people may also want to show that they are concerned with accurately evaluating themselves. Moreover, the appearance of concern with self-improvement may be of importance. People may also want to be viewed as correct in their evaluations, and thus their self-evaluation may show evidence of selfverification (see Tice, Butler, Muraven, & Stillwell, 1995). Although there was initially a good deal of debate over which of the motives was dominant, it has recently been suggested that the motives underlying self-evaluation vary as a function of characteristics of the situation and individual. Indeed, Taylor, Neter, and Wayment (1995) provided evidence that, at times, self-assessment, selfimprovement, self-enhancement, and self-verification motives may all drive self-evaluation. In the face of much work suggesting that self-enhancement and self-verification were the most predominant motives, Trope demonstrated in several studies that the self-assessment motive guided self-evaluation when people felt uncertain about their abilities (e.g., Trope & Bassok, 1982). Subsequently, Trope (1986) suggested that the motive guiding self-evaluation is determined by a larger self-regulatory system in which people attempt to maintain their self-worth while also seeking to meet their goals (see also Taylor et al., 1995). In such a system, short-term gains in self-worth are often sacrificed for long-term gains in self-worth. According to Trope, people are likely to engage in self-assessment, seeking out negative, but useful, information when they feel good about themselves. However, when people do not feel good about themselves, they may avoid such information, focusing instead on positive, but not very useful, information. In this vein, Trope and Neter (1994) provided evidence that self-assessment motives predominate when people feel good about themselves. People receiving positive feedback about their competencies in one area were more willing to seek out negative, but useful, feedback about their competencies in another area. In contrast, people feeling poorly about themselves appeared to be guided by self-enhancement motives, preferring positive, but not very useful, feedback about their competencies. Trope and Pomerantz (1998) elaborated on this finding by showing that, when people place a good deal of importance on being competent in a particular area, they are likely to seek out both negative and positive feedback. The proclivity for negative feedback in areas of importance, however, decreases when negative feedback is received in another area, but increases when positive feedback is received in another area. Hence, self-assessment and self-enhancement motives wax and wane as a consequence of how people feel about themselves. Ruble and Frey (1991) also provided a framework incorporating the multiple motives that underlie self-evaluation. These investigators have proposed that such motives vary as a function of skill acquisition. In short, initial skill acquisition is characterized by concerns with learning how to do the task at hand. Hence, when one is just learning a skill, self-evaluation is oriented toward self-improvement. Once a skill is acquired, the main motive is to assess one's competence accurately. At this point, uncertainty is high. Consequently, people engage in self-evaluation for purposes of self-assessment. Once conclusions are reached about competence, the concern is with maintaining adaptive strategies. According to Ruble and Frey, self-evaluation may be quite infrequent at this time. When people do engage in self-evaluation, they may do so to enhance their self-image in an effort to remain involved in the task. The maintenance phase continues until people are given a chance to discontinue

< previous page

page_60

next page >

< previous page

page_61

next page >

Page 61 the task. At this point, people will again be concerned with self-assessment in an effort to make an informed decision about whether to continue. Research examining changes in the manner in which children evaluate themselves as they progress through elementary school is quite consistent with Ruble and Frey's framework (e.g., Butler, 1989; Pomerantz, Ruble, Frey, & Greulich, 1995). In a somewhat different vein, Swann has sought to determine the situational contingencies of the motives guiding self-evaluation. His aim has been to distinguish between the conditions in which the self-enhancement motive is dominant and those in which the self-verification motive is dominant. Drawing from Shraguer (1975), Swann and colleagues (Swann, Griffin, Predmore, & Gaines, 1987) showed that the self-enhancement motive is more likely to guide how people feel about evaluative feedback, whereas the self-verification motive is more likely to drive how people think about evaluative feedback. Relatedly, Swann, Hixon, Stein-Seroussi, and Gilbert (1990) demonstrated that people's initial preferences for feedback are guided by the self-enhancement motive, whereas their later preferences are guided by the self-verification motive. Swann and colleagues viewed this timing difference as reflecting a difference in the complexity of cognitive operations necessary to meet each motive. Because selfenhancement requires fewer cognitive operations than self-verification, when cognitively constrained, people will engage in self-evaluation reflecting self-enhancement; otherwise, self-verification will dominate. Hence, selfverification may predominate over self-enhancement when cognitive resources are limited. Investigators have implicated a number of other situational factors influencing the motives driving self-evaluation. These include, but are not limited to, past or future threat (Taylor et al., 1995), one's stage in the decision-making process (Taylor & Gollwitzer, 1995), and perceptions of the modifiability of traits (Dunning, 1995). Individual differences, particularly self-esteem (e.g., Wayment & Taylor, 1995; Wood, Giordano-Beech, Taylor, Michela, & Gaus, 1994), also appear to play an important role in the motives driving self-evaluation. Processes Involved in Self-Evaluation A variety of social-cognitive processes are involved in self-evaluation. It is through such processes that people are able to meet the various motives just highlighted. Information seeking is one of the key processes involved in selfevaluation. The type of information people seek has been assumed to influence the conclusions they make about themselves. A good deal of theory and research has simply focused on the valence of the information people seek, comparing the preference for positive versus negative information (e.g., Swann et al., 1989). However, in the tradition of social comparison, there has also been a focus on the type of comparisons people make. Of particular interest has been the extent to which people choose to compare themselves with inferior others, similar others, or superior others (e.g., Buunk, Collins, Taylor, VanYperen, & Dakof, 1990). Notably, although information seeking may influence how people evaluate themselves, how people evaluate themselves may also influence the sort of information people seek. When people have self-verification motives, their decision about what type of information to seek is guided by their self-views. The types of attributions people make for their performance have also been viewed as a key process involved in selfevaluation (see Weiner, 1985, for a review). When people view their performance as due to ability, their performance is likely to influence how they evaluate themselves. Attributions of failure to ability are likely to result in negative self-evaluation, whereas attributions of success to ability are likely to result in positive self-evaluation. Attributions to other causes, such as effort or luck, are likely to mitigate the effect of performance on self-evaluation because such attributions carry the implication that performance will change in the future. Of course, self-evaluation can also influence the type of attribution made. People with positive views of themselves are probably less likely than those with negative views of themselves to see their failure as due to a lack of ability. Several other social-cognitive processes are also involved in self-evaluation. First, within-person comparisons have been the focus of much work concerned with self-regulation. One form of such a comparison is contrasting the actual self with the self that is being pursued (e.g., Higgins, Bond, Klein, & Strauman, 1986). In this vein, people compare their current accomplishments with what they aim to accomplish. Such comparisons result in an evaluation of one's goal-related progress, with implications for how the self is viewed. People may also make temporal comparisons, contrasting their present performance with their past performance to arrive at a self-evaluation (e.g., Ruble & Frey, 1991). Second, people's evaluations of others have been assumed to contribute to self-evaluation (e.g., Wagner, Wicklund, & Shaigan, 1980). By denigrating others' abilities, one may come to view one's own abilities in a positive light. Similarly, by giving others' more credit for their accomplishments than they deserve, one may come

< previous page

page_61

next page >

< previous page

page_62

next page >

Page 62 to evaluate oneself in a negative light. In a somewhat different vein, Tesser (1988) suggested that at times people's self-evaluation is enhanced when they view close others as possessing positive attributes. In Tesser's terms, people bask in the reflected glory of close others. Consequences of Self-Evaluation One of the reasons that self-evaluation has received such a great deal of attention is because of its consequences for psychological functioning. Self-evaluation, as well as the social-cognitive processes involved in self-evaluation, has been suggested to have two major consequences for psychological functioning. First, self-evaluation appears to play a large role in motivation, thereby influencing achievement. Second, it has a number of emotional consequences that have implications for the experience of internalizing symptoms, particularly anxiety and depressive symptoms. Motivational Implications . The manner in which people evaluate themselves has been key in models of motivation. For example, a number of investigators concerned with the extent to which people persist in a particular area have relied on expectancy-value frameworks (e.g., Atkinson, 1964; Eccles, 1987; Feather, 1982). In such frameworks, persistence in an area is assumed to be a function of people's expectancies for their future performance in the area and the value they place on performing well in the area. A good deal of evidence has accrued in support of such models, with expectancies being a particularly strong predictor of persistence (see Eccles et al., 1998, for a review). In a slightly different vein, Deci and Ryan (1985) suggested that self-evaluation influences the quality of motivation. These investigators argued that feelings of self-determination lead people to be intrinsically motivated. Part of feeling self-determined, according to Deci and Ryan, is feeling competent. Hence, positive self-evaluation is likely to foster intrinsic motivation, whereas negative self-evaluation is likely to foster extrinsic motivation. Dweck (1986; Dweck & Leggett, 1988) proposed that how people account for their failure influences their motivation. In this vein, Diener and Dweck (1978, 1980) documented that children attributing their failure to effort show a mastery-oriented reaction to failure. In the face of failure, these children engage in self-monitoring and selfinstructions in an effort to change their strategies to make them more effective. In contrast, children who do not attribute failure to effort display a helpless reaction to failure. These children search for explanations for their failure instead of focusing on how to improve their strategies. In addition, they evaluate themselves negatively when confronted with failure, often displaying negative affect. In the end, these children often give up. Emotional Implications . The manner in which people account for their failure has also been argued to be a central influence in the onset of depressive symptoms. Abramson and colleagues (1989) proposed that depressive symptoms may emerge as a consequence of how people respond to failure. In brief, people with a tendency to blame failure on global and stable causes may be more likely to evaluate themselves negatively in the face of failure. This negative self-evaluation is expected to foster hopelessness, which in turn is expected to lead to depressive symptoms. Empirical evidence is generally consistent with this model (e.g., Lewinsohn, Roberts, Seeley, Rohde, Gotlib, & Hops, 1994; Nolen-Hoeksema, Girgus, & Seligman, 1986), although there is also some evidence that depressive symptoms may cause negative self-evaluation (e.g., Cole, Martin, & Powers, 1997; Nolen-Hoeksema et al., 1994). Self-evaluation has also been linked to anxiety symptoms (e.g., Leitenberg, Yost, & Caroll-Wilson, 1986), although the role of self-evaluation in the onset of anxiety symptoms has received relatively little attention. Taylor and Brown (1988) have taken a somewhat different approach to the issue of the emotional consequences of selfevaluation. These investigators argued that there is a great deal of evidence to suggest that overly positive selfevaluation – positive illusions – promotes mental health (but see Colvin & Block, 1994). Higgins (1987) attempted to distinguish the type of self-evaluation that leads to depressive symptoms from that leading to anxiety symptoms. In his self-discrepancy theory, Higgins suggested that people hold guides for what they would ideally like to be (the ideal self) and for what they feel they ought to be (the ought guide). To the extent that one's perceptions of what one is actually like (the actual self) differ from these guides, people will experience negative emotions. However, according to Higgins, the type of negative emotion is dependent on the type of discrepancy (but see Tangney, Niedenthal, Covert, & Barlow, 1998). He has shown that when people feel that their actual self does not meet their ideal guide, they experience dejection-related emotions, such as disappointment, sadness, and shame, which put them at risk for depressive symptoms (e.g., Higgins, Bond, Klein, & Strauman, 1986). Yet when people feel that their actual self does not meet their ought guide, they experience agitation-related emotions, such as fear, guilt, and restlessness, putting them at risk for anxiety symptoms. Notably, Higgins has further suggested that the emotional consequences of self-discrepancies may be the most intense for people possessing strong self-guides.

< previous page

page_62

next page >

< previous page

page_63

next page >

Page 63 The strength of a self-guide is defined by its accessibility, coherence, and the commitment felt toward it. Accessibility refers to the likelihood that the guide will be cognitively activated; coherence refers to whether it has clear implications for behavior; and commitment is the extent to which one is motivated to live up to the standards embodied by the self-guide. Operationalizing self-guide strength as accessibility, Higgins, Shah, and Friedman (1997) have found evidence that strong self-guides intensify the emotional impact of discrepancies. Summary In answer to the question of what motivates people's self-evaluation, investigators have emphasized four central motives. People are motivated by concerns with accurate self-assessment as well as with self-improvement. In addition, people evaluate themselves with an eye to enhance their self-image, boosting their self-worth. People also seek to verify their self-views in an effort to control their environment. These motives have been integrated into models in which people are assumed to attempt to regulate their progress toward their goals while simultaneously maintaining their self-worth. The emphasis of these models is that the dominance of self-assessment, selfimprovement, self-enhancement, and self-verification motives depends on a variety of factors. Although these four motives have received a great deal of attention in work examining self-evaluation, there are also two additional motives of importance – solidifying one′s identity and managing the impression one makes on others. It is essential for future models to integrate the occurrence of these latter two motives with the occurrence of the other four motives. Notably, there are multiple social-cognitive processes by which people evaluate themselves. These processes, as well as the final evaluation at which people arrive, have important implications for people′s motivation and emotional experiences. We now turn to the issue of sex differences in self-evaluation. Although some of the issues we have discussed up to this point, such as the consequences of self-evaluation, have been of concern to investigators focusing on sex differences in self-evaluation, other issues, such as the motives guiding self-evaluation, have been given scant attention at most. AN OVERVIEW OF SEX DIFFERENCES IN SELF-EVALUATION Although research on sex differences in self-evaluation has not generally focused on motives, there is some evidence to suggest that the self-evaluation of females is guided by different motives than the self-evaluation of males. In general it appears that the self-evaluation of females is driven by self-assessment, whereas the self-evaluation of males is driven by self-enhancement. Indeed, females appear to be more sensitive than males to failure feedback, adjusting their evaluation of themselves to reflect such feedback (see Roberts, 1991; Ruble et al., 1993, for reviews). Moreover, females are often more likely than males to attribute their failure to a lack of ability (see Deaux, 1976; Ruble et al., 1993, for reviews; but see Sohn, 1982). Although there may be sex differences in the motives guiding self-evaluation immediately following failure, such sex differences may not be as pronounced at other times. Research comparing the self-perceptions of females and males generally finds that males have more positive perceptions of themselves than do females in stereotypically masculine areas, such as math and sports (see Eccles et al., 1998; Ruble et al., 1993, for reviews). Females, however, have more positive self-perceptions than do males in stereotypically feminine areas, such as reading and interpersonal relations. Moreover, during the elementary school years, females also tend to view themselves more positively in terms of their general academic skills. Unfortunately, measures of actual performance have rarely been collected in conjunction with self-perceptions. As a consequence, it is difficult to determine whether there are differences in the extent to which females and males accurately represent their performance. Consistent with the perspective that self-assessment guides the selfevaluation of females, whereas self-enhancement guides the self-evaluation of males, the few studies that have collected measures of performance have indicated that females tend to accurately estimate their performance, sometimes underestimating it, whereas males tend to overestimate their performance (Crandall, 1969; Frey & Ruble, 1987). However, the tendency for females to view themselves negatively in stereotypically masculine areas and males to view themselves negatively in stereotypically feminine areas may be indicative of a self-verification motive. People may internalize the societal expectancies and then seek to confirm them. This tendency, however, may also represent an attempt to solidify one′s gender identity. Moreover, to the extent that such self-evaluations are made in public, it could represent a tendency to make a favorable impression by demonstrating adherence to societal standards (see Lenney, 1977). Unfortunately, sex differences in some of the other social-cognitive processes involved in self-evaluation (e.g., information seeking, within-person comparisons) have not been systematically examined. The interest in sex differences in self-evaluation has

< previous page

page_63

next page >

< previous page

page_64

next page >

Page 64 grown out of an effort to understand the persistent under-representation of women in high-level positions, particularly in advanced quantitative fields (e.g., Eccles, 1989; Hoffman, 1972; Lenney, 1977). Although sociological factors such as institutional barriers are crucial to women's absence from these positions, the sex differences in selfevaluation may also play a key role (see Eccles, 1989, 1994). Females may be less likely than males to pursue highlevel positions, especially in stereotypically male fields, because of the motivational consequences of negative selfevaluation. More recently, investigators have become interested in the sex differences in self-evaluation because the differences may account, in part, for sex differences in internalizing symptoms (e.g., Ruble et al., 1993). As early as elementary school, females experience greater anxiety symptoms than males (see Feingold, 1994, for a review); by adolescence, females are more vulnerable to depressive symptoms than are males (see Nolen-Hoeksema, 1987; Nolen-Hoeksema & Girgus, 1994, for reviews). Because self-evaluation has been linked to internalizing symptoms, it is quite possible that the sex differences in self-evaluation represent precursors to the sex differences in internalizing symptoms (see Ruble et al., 1993). Several models have been proposed to account for the sex differences in self-evaluation. These models generally fall into two categories.1 One set of models has emerged from the area of adult social psychology, with a particular emphasis on social cognition. The models focus primarily on the role of gender stereotypes in sex differences in selfevaluation. The guiding philosophy is that sex differences in self-evaluation occur when gender stereotypes are salient. A second set of models has grown out of the area of developmental psychology. These models focus on gender socialization. The major postulation of these models is that members of society, such as parents and teachers, treat females and males differently during childhood, and this causes females to evaluate themselves more negatively than males in some situations. During the past decade, explanations for the sex differences in selfevaluation have been based primarily on gender stereotype models. This uneven focus has been due in large part to the conclusion that there is little evidence that parents and teachers treat females and males differently (see Eccles & Blumenfeld, 1985; Lytton & Romney, 1991, for reviews). However, such a conclusion may be premature. Thus, the lack of attention to gender socialization models may not be warranted. Gender Stereotype Models The focus of gender stereotype models has been on the role that proximal cues play in sex differences in selfevaluation. In these models, consistent with Deaux and Major's (1987) proposal that gender-related interactions are influenced by the social context in which they occur, females are not expected to evaluate themselves more negatively than males across the board. Instead, sex differences are only expected when cues related to gender stereotypes are salient. These cues are assumed to activate societal expectancies about what females and males are like, which in turn guide self-evaluation. Gender stereotype cues may be made salient by features of the environment in which self-evaluation takes place (e.g., the presence of another individual, a reminder of one's gender) or by the area of self-evaluation (e.g., math skills, interpersonal relations). Lenney (1977) presented the first formal proposal that gender stereotypes play a role in the sex differences in selfevaluation. Lenney noted that there was a good deal of inconsistency in the sex differences in self-evaluation. She suggested that the inconsistency was meaningful and that it reflected the possibility that the sex differences in selfevaluation are sensitive to situational cues. Her analysis centered, albeit not entirely, on the role of cues related to gender stereotypes. When tasks are viewed as masculine, females evaluate themselves negatively. Conversely, when tasks are viewed as feminine, males evaluate themselves negatively. Lenney emphasized that the gender linkage of the task can be reflected both in the culturally held stereotype (e.g., males are better at sports than females, females are better at reading than males) as well as in the simple labeling of the task (e.g., this is a task in which females usually outperform males, this is a skill that males are more likely to possess than females). In addition, Lenney proposed that when one's gender is made salient (e.g., being the only female in the room, repeated references to gender), gender stereotypes are likely to be activated. She expected that, as a consequence, sex differences in self-evaluation would be evident in such situations. Lenney's proposal was followed by Eccles’ (1984, 1989) application of an expectancy-value approach to understanding sex differences in self-evaluation. The central concern of this model is with what underlies sex differences in achievement-related choices. Key to such choices are people's expectancies for their future performance in a particular area and the value they place on success in the area. Consistent with other expectancyvalue models, when people possess high expectancies for performance in a particular area and place a great deal of value on success in that area, they choose to pursue that area even when they are faced with failure. Eccles posited that culturally held stereotypes about

< previous page

page_64

next page >

< previous page

page_65

next page >

Page 65 females' and males' competencies and roles foster sex differences in expectancies for performance and the value placed on performance in gender stereotypical areas. First, gender stereotypes provide information about females' and males' skills in these areas, and this leads to sex differences in expectancies for performance. For example, because gender stereotypes indicate that females are less skilled at math than males, females expect to do more poorly in math than do males. Second, gender stereotypes shape the importance of the attributes and goals associated with gender-stereotypical areas for females and males. This, in turn, fosters sex differences in the value placed on success in these areas. For example, because gender stereotypes characterize females as more likely than males to become involved in child care, females may view the attributes associated with being involved in child care as more central to their identity than do males. Consequently, in a desire to solidify their identity, females may place more value than males on attaining such attributes. Eccles suggested that people's expectancies and values may be influenced by proximal factors (i.e., gender cues in the immediate situation) as well as distal factors (i.e., socialization influences) that accrue over time. Most recently, Steele (1997) has put forth a model of stereotype threat. This model was designed to explain sex differences in achievement in math and science, as well as racial differences in general school achievement. Although the major concern is with actual achievement, the model also speaks to self-evaluation. According to Steele, stereotype threat arises when one is doing something for which a negative stereotype about one's group applies. In terms of gender, the negative stereotype is that females are not good at math and science. Significantly, stereotype threat only occurs when the negative stereotype applies to an area central to self-definition. Consequently, stereotype threat is most likely to affect the self-evaluation of females who do well in the areas of math and science. Steele argued that stereotype threat is elicited by cues in the immediate environment that emphasize stereotypes. The cues can be as subtle as reminding people of their gender or as blatant as directing people's attention to the relevance of gender stereotypes to the task at hand. Steele argued that stereotype threat can lead to anxiety over performance through two pathways. First, when taking a test in math or science, females may feel that, despite past success, they may confirm the stereotype that females lack competence in this area. This may foster anxiety that interferes with performance. Second, stereotype threat can lead to disidentification with the area for which the negative stereotype is relevant. Females continually reminded of the stereotype that females are not good at math and science may cease to view being skilled in these areas as an important aspect of their identity. Such disidentification may undermine persistence in math and science, thereby inhibiting their performance. Their poor performance may lead females, often guided by self-assessment motives, to evaluate their skills in math and science in a negative light. Evaluation of the Models . The models focusing on gender stereotypes have a variety of strengths. Perhaps most important, consistent with Deaux and Major's (1987) analysis of gender interactions, these models posit that the sex differences in self-evaluation are not evident in all situations. Specifically, the gender stereotype models suggest that the sex differences in self-evaluation are evident in situations in which cues related to gender stereotypes are salient. This postulation is consistent with the empirical evidence indicating that females evaluate themselves more negatively in stereotypically masculine areas, whereas males evaluate themselves more negatively in stereotypically feminine areas. The gender stereotype models, however, also have some notable weaknesses. A number of findings suggest that the salience of gender stereotype cues in the immediate situation is not the only factor underlying females' tendency to evaluate themselves more negatively than males. First, as noted earlier, research suggests that females are particularly likely to evaluate themselves more negatively than males in the face of failure. It appears that, at least in the face of failure, females are guided by self-assessment, whereas males are guided by self-enhancement. Yet none of the gender stereotype models can make these predictions. Second, although little research has examined selfevaluation in very young children (i.e., children younger than 5 years of age), there appear to be sex differences in reactions to failure among such children. As early as 3 years of age, females are more likely than males to experience shame after failure, suggesting that they may be more likely to draw negative conclusions than boys about their competencies from failure (Stipek, Recchia, & McClintic, 1992). However, because children do not have extensive knowledge of gender stereotypes at this age, particularly in regard to males' and females' psychological attributes (see Ruble & Martin, 1998, for a review), such findings pose a potential problem for models positing that gender stereotypes represent the major cause of the sex differences in self-evaluation. Although gender stereotype cues may maintain and even promote the sex differences in self-evaluation once children become aware of gender stereotypes, such cues are unlikely to be involved in the initial development of the sex differences. Third, despite

< previous page

page_65

next page >

< previous page

page_66

next page >

Page 66 the fact that the gender stereotype models provide an excellent account of why females are less likely than males to occupy high-level positions in math and science, they do not account for the pattern of sex differences in internalizing symptoms described earlier. Gender Socialization Models The general philosophy guiding the gender socialization models is that teachers and parents treat females and males differently during childhood, leading either directly or indirectly to sex differences in self-evaluation under some circumstances. These models focus on differences in how males and females are treated and the processes by which such differences lead to sex differences in self-evaluation. Most of the models assume that gender socialization is due to culturally held gender stereotypes, but give little, if any, attention to the processes by which these stereotypes lead to gender socialization. Although much of the research on gender socialization has proceeded without the guidance of theoretical models, there have been a few impressive attempts to build such models. Teacher Gender Socialization. One of the most influential models of teacher gender socialization is the set of proposals put forth by Dweck (Dweck & Bush, 1976; Dweck, Davidson, Nelson, & Enna, 1978). Dweck was concerned with why females attribute their failure to ability, whereas males attribute their failure to effort or external sources, such as the evaluator. The central hypothesis of her model is that such sex differences are due to teachers' differential use of evaluative feedback with girls and boys during the elementary school years. First, teachers may provide their female students with more positive than negative feedback, but provide their male students with more negative than positive feedback. This may lead females to be more likely than males to attribute positive feedback to external causes, but negative feedback to internal causes. Second, Dweck proposed that teachers give girls negative feedback about intellectual aspects of their work but give boys negative feedback about nonintellectual aspects. As a result, females are more likely than males to perceive negative feedback as relevant to their ability. Third, teachers are less likely to attribute girls' failure to a lack of motivation than they are boys' failure. Consequently, females may be less likely than males to explain their failures as due to a lack of effort. In summary, Dweck suggested that teachers are more sparing and discriminating in their use of negative evaluative feedback with girls than with boys, and this leads girls to be more likely than boys to view such feedback as particularly informative about their ability. Roberts (1991) also argued that females view feedback as more informative about their ability than do males. Fennema and Peterson (1985) identified another dimension of teachers' behavior as playing a role in the sex differences in self-evaluation: control. The central hypothesis of Fennema and Peterson's model is that teachers exert more control over girls than over boys during the elementary school years, leading girls to have less confidence in their abilities than boys and consequently to attribute their failure to a lack of ability more often than boys. Fennema and Peterson outlined a variety of ways in which teachers may be less likely to promote independence among girls than among boys: Teachers encourage less independent thinking in girls than in boys, teachers interact less and at lower cognitive levels with high-achieving girls than with high-achieving boys, and teachers are more likely to impose rules over girls than over boys. This pattern of differential treatment presumably leads females to be more likely than males to view themselves as incompetent and thus to attribute their failure to a lack of ability. The notion that females may be less likely than males to be encouraged to act independently in the classroom has been echoed by investigators concerned with teacher socialization of general sex roles (e.g., Brophy, 1985; Fagot, 1978, 1981). Fennema and Peterson viewed the sex differences in attributions for failure as crucial to elucidating the decline in females' achievement in math following elementary school. Drawing on research by Dweck (e.g., Dweck et al., 1978; Dweck & Repucci, 1973), these investigators argued that females' tendency to attribute failure to ability leads them to avoid challenge and to give up when faced with challenge. According to Fennema and Peterson, this is particularly detrimental to achievement in math. Parental Gender Socialization. Hoffman (1972) suggested two dimensions on which parents differentially socialize girls and boys that are central to the development of the sex differences in self-evaluation. According to Hoffman, this differential socialization begins as early as infancy. First, parents are more likely to encourage independence among their sons than among their daughters. According to Hoffman, parents' encouragement of independence may manifest itself in a variety of practices, including, but not limited to, allowing children to explore and manipulate their environment, permitting children to attempt moderately risky behaviors (e.g., using scissors, climbing on a stool) on their own, allowing children to play away from home, and praising independent behaviors (e.g., taking one's first step, putting a puzzle together on one's own). Second, parents are more likely to encourage their sons than their daugh-

< previous page

page_66

next page >

< previous page

page_67

next page >

Page 67 ters to establish an identity distinct from the desires and wishes of their parents. Hoffman suggested that this is because mothers are often the primary care-givers, and consequently girls identify more strongly than boys with the central authority figure in their lives. Biologically based sex differences in activity level may also lead boys to create more conflict with their parents than do girls as early as infancy, which aids boys in creating a distinct identity. Hoffman argued that this pattern of differential treatment hinders females from developing skills to cope with challenges, thereby inhibiting their confidence to do so. This, in turn, causes females to be dependent on others to cope with such challenges. Consequently, females are more likely than males to be concerned with pleasing others in an effort to maintain the guidance of others. Hence, females more than males are driven to achieve by a desire to please others. Hoffman proposed that such an orientation has both positive and negative consequences. On the one hand, when achievement is valued by others, it heightens motivation to perform well, increasing effort and thereby achievement. On the other hand, when others do not value achievement, it decreases motivation in this area because effort is directed to areas that are valued by others (e.g., being a good friend, having children). Moreover, Hoffman argued that being motivated by a desire to please others can lead to poor achievement as it heightens the emphasis put on performance instead of mastery. Higgins (1991) highlighted several important differences in the way that parents socialize their daughters and sons, many of which are consonant with those emphasized by Hoffman (see also Block, 1983; Huston, 1983). The central theme among these differences is that parents are more controlling with their daughters than with their sons in a number of ways: Parents put greater pressure on girls than on boys to be nurturant, obedient, and socially responsible; parents respond more quickly to the mistakes of their daughters than to those of their sons; parents are more likely to interrupt, intrude, and direct their daughters' activities than those of their sons; and parents maintain closer track of girls than of boys, often keeping their daughters under surveillance. Drawing from his self-discrepancy theory, Higgins argued that this pattern of gender socialization leads females to possess stronger self-guides than males. Specifically, the parenting practices used more with girls than with boys emphasize the importance of meeting standards set by parents. Consequently, females may feel more compelled than males to meet standards valued by society. According to Higgins, possessing strong self-guides may have psychological trade-offs. It may increase motivation to perform well and, to the extent that this leads to success, foster positive self-evaluation, inhibiting internalizing symptoms. However, when such motivation does not lead to success and failure is encountered, possessing strong self-guides may lead to negative self-evaluation, fostering internalizing symptoms (see also Costanzo, Miller-Johnson, & Wencel, 1995). Evaluation of the Models . Similar to the gender stereotype models, the gender socialization models are consistent with the extant research indicating that females do not evaluate themselves more negatively than males across the board. However, unlike the gender stereotype models, the gender socialization models are consistent with research finding that females are particularly likely to evaluate themselves more negatively than males in the face of failure. Moreover, the gender socialization models assume that parental gender socialization begins in the earliest years of children's lives. These models also do not suggest that knowledge of gender stereotypes is necessary for the sex differences in self-evaluation. Thus, the gender socialization models allow for the possibility that sex differences in self-evaluation may be evident even at very young ages. Finally, as we elaborate on later in describing our approach, the gender socialization models may also have the power to explain the pattern of sex differences in internalizing symptoms. Despite these strengths, the gender socialization models have some significant weaknesses. First, these models do not fully account for why females evaluate themselves more negatively than males in stereotypically male areas but more positively than males in stereotypically female areas. Second, the research on whether teachers and parents actually treat children differently is ambiguous at best. For example, after conducting a meta-analysis examining parental gender socialization, Lytton and Romney (1991) concluded that there is little evidence that parents socialize their daughters and sons differently; similar conclusions have been made in reviews of teacher gender socialization (e.g., Eccles & Blumenfeld, 1985). However, there are a number of reasons that such conclusions may be premature (see Block, 1983; Fennema & Peterson, 1985; Keenan & Shaw, 1997; Leaper et al., 1998; Pomerantz & Ruble, 1998). Many of these reasons grow out of the approach we suggest in the next section, and thus they are elaborated on there. Third, and relatedly, the lack of attention in these models to the causes behind gender socialization is problematic. Indeed, as highlighted in the next section, the inattention to this issue may be one of the reasons that evidence for gender socialization is so ambiguous.

< previous page

page_67

next page >

< previous page

page_68

next page >

Page 68 Fig. 4.1. An approach integrating gender stereotypes and gender socialization models of sex differences in selfevaluation.

FUTURE DIRECTIONS In this section, we describe an approach for future research integrating ideas from both the gender stereotype and gender socialization models. This involves incorporating social-cognitive notions, such as construct activation, expectancy-based judgments, and stereotype use, into existing gender socialization models. The major strength of such an approach is that it takes a step toward accounting for the complex pattern of findings on parents' differential treatment of girls and boys as well as on sex differences in self-evaluation and internalizing symptoms. Moreover, it also has the potential to explain the development of these phenomena in general. This approach has generated a model we have begun to test in our research (see Fig. 4.1). Consistent with the gender socialization models, the central hypothesis is that parents are more controlling with their daughters than with their sons. We view control as attempts on parents' part to push children toward meeting standards. As highlighted shortly, there are a number of practices by which parents may do this. Unlike the prior gender socialization models, attention is given to the factors – namely, culturally held gender stereotypes – that lead to parents' differential use of control with their daughters and sons. Parental control is expected to convey to children that meeting standards, particularly those set by parents, is important. As a consequence, females may become more invested than males in meeting standards. This may have consequences for self-evaluation as well as internalizing symptoms. In a departure from prior models of the sex differences in self-evaluation, social-contextual and social-cognitive developmental factors are viewed as important to the emergence of the sex differences in selfevaluation as well as in internalizing symptoms. Gender Stereotypes and Parental Control One of the key hypotheses is that gender socialization is due to culturally held gender stereotypes. (The development of such stereotypes in the culture and how they get transmitted to parents is beyond the scope of this chapter; but see Schaller & Conway, chap. 10, this volume). Culturally held gender stereotypes may lead parents to use control differently with their daughters and sons through two mechanisms. First, parents′ perceptions of their children as possessing stereotypically female characteristics (e.g., helplessness, indecisiveness) or as lacking stereotypically male characteristics (e.g., confidence, self-reliance) may lead parents to be more controlling. For example, when parents view their children as lacking independence, they may make decisions for them. Second, the more parents possess stereotypical feminine socialization goals (e.g., child should be responsible, obedient) and lack stereotypical masculine socialization goals (e.g., child should be independent, assertive), the more controlling parents may be. For

< previous page

page_68

next page >

< previous page

page_69

next page >

Page 69 example, when parents are concerned with ensuring that their children are independent, they may refrain from making decisions for them. Because parents may be more likely to hold stereotypically feminine perceptions and goals for girls than for boys, parents may be more likely to use control with girls than with boys. The proposal that gender stereotypes cause parents to exert control differentially over girls and boys suggests that such differential treatment may be context-specific. Forms of control to which gender-linked perceptions of children or gender socialization goals are relevant may be the only forms of control that parents use differently with girls and boys. For example, perceptions that girls but not boys are helpless may lead parents to monitor the progress of girls but not of boys. These same perceptions, however, may not influence parents' use of discipline with girls versus boys because this form of control may not be governed by perceptions of children as helpless or as possessing other gender stereotypical characteristics. Grouping together multiple forms of parental control may thus obscure gender socialization (Block, 1983; Pomerantz & Ruble, 1998). Yet most prior research has done just this (see Lytton & Romney, 1991, for a review). Research on stereotypes suggests an additional set of factors that may moderate the extent to which parents differentially use control with their daughters and sons. For stereotypes to exert an influence, they must be accessible (see Bargh, 1990; Higgins, 1996, for reviews). Stereotypes become accessible when they are made salient by situational cues, triggered by immediately observable characteristics of the interaction partner, or chronically high in the system guiding people's perceptions of the world (see Deaux & Major, 1987; Higgins & King, 1981; Moskowitz, Wasel, Gollwitzer, & Schaal, 1999, for reviews). Once stereotypes are accessible, even people who do not endorse stereotypes may act on them when they are not motivated to avoid doing so or have depleted cognitive resources (e.g., Devine, 1989; Fiske, 1989; Geis, 1993). Hence, although parents may not endorse gender stereotypes, they may act on these stereotypes in their daily interactions with their children because they have goals other than gender-neutral socialization or are preoccupied with other issues (e.g., being on time, a conflict with a spouse). Consistent with this proposal, correlations between parents′ socialization beliefs and behaviors are quite low (see Holden & Edwards, 1989, for a review). Evidence for the proposal that culturally held gender stereotypes may underlie parents' differential use of control with their daughters and sons is beginning to emerge, although, as of yet, this evidence is somewhat indirect. In a sample of mothers of elementary school children (ages 7 - 10), Pomerantz and Ruble (1998) examined the possibility that parents' differential use of control with their daughters and sons is limited to forms of control to which gender stereotypical characteristics are relevant. These investigators assessed four forms of control to which gender stereotypes were thought to be relevant: helping (e.g., helping child with homework when child does not request it), monitoring (e.g., checking to make sure child has done homework correctly when child does not request it), decision making (e.g., making decisions for child about the topic of a school project), and praising (e.g., giving child a reward for getting a good grade). To test the hypothesis that the differential use of control is dependent on the form of control, mothers' use of control in the form of disciplining (e.g., withdrawing privileges for getting a poor grade) was also assessed. It was expected that gender stereotypical characteristics would not be relevant to this form of control, and thus differential treatment would not be evident. Because it was crucial to obtain assessments of mothers’ use of control that would not be biased by their beliefs, control was assessed with a daily checklist completed over a 1to 3-week period. As anticipated, mothers of girls were more likely than mothers of boys to use control in the forms of helping, monitoring, decision making, and praising, but not disciplining. Because the daily checklist also provided information about the frequency with which children engaged in behavior (e.g., requests for help) that might elicit mothers' use of control, it was possible to account for girls' and boys behavior. However, there were no sex differences in children's behavior. Mothers also completed measures of their beliefs about control (Pomerantz, 1995). Consistent with prior research, mothers' beliefs were only modestly correlated with their behavior. Moreover, there were no differences in the beliefs of mothers of girls and mothers of boys. Although more research is clearly needed, these findings point toward the possibility that culturally held gender stereotypes may underlie parents' differential treatment of girls and boys. Consequences of Parental Control How might parents' differential use of control with their daughters and sons lead to sex differences in selfevaluation? The gender socialization models, as well as other theory and research on control (e.g., Deci & Ryan, 1985; Harackiewicz, Manderlink, & Sansone, 1984; Nolen-Hoeksema, Wolfson, Mumme, & Guskin, 1995), suggest that control may communicate to children that it is essential for them to meet standards, heightening their

< previous page

page_69

next page >

< previous page

page_70

next page >

Page 70 investment in successfully doing so. When parents use control with their children, they provide two types of information that may heighten children's concern with meeting standards (Higgins, 1991). First, through the use of control, parents convey what their standards are. Second, parents' use of control communicates that they feel it is important for children to meet such standards. For example, parents who frequently monitor their children's homework to ensure that children do it correctly indicate that doing well in school is important and that part of doing well involves doing homework correctly. The continual emphasis on the standards parents deem important may lead children to feel compelled to meet these standards, often feeling that their self-worth is dependent on doing so. In Higgins' (1991) terms, parents' frequent use of control may lead children to possess strong self-guides. The proposal that parental control may heighten children's concern with meeting standards has begun to receive empirical support in a study (Kenney & Pomerantz, 2000) employing laboratory observations of mothers' use of control with their elementary school children (ages 7 - 10). In this study, children were presented with a difficult cognitive abilities task, and mothers were told that they could help their children as little or as much as they wanted. These mother-child interactions were then coded for the extent to which mothers used control with their children (e.g., directed how children did the task, did the task for children). As anticipated, children whose mothers used a good deal of control were quite concerned with meeting standards. These children reported that performing well on an intelligence task they were about to complete was very important to them. In addition, they reported feeling a good deal of pressure from others to be perfect (e.g., ‘‘Other people think I have failed if I do not do my very best all the time,” “Other people always expect me to be perfect”). Although this research suggests that parental control fosters investment in meeting standards among children, longitudinal research is essential for determining the direction of influence. Consequences for Self-Evaluation and Internalizing Symptoms The heightened concern with meeting standards resulting from parental control may have psychological trade-offs. Key to this proposal is that when a great deal of investment is put into meeting standards, meeting standards is often central to protecting one's self-worth (see Pelham, 1991; Sedikides & Strube, 1997; Tesser, 1988; Wicklund & Gollwitzer, 1982). In our research with both children and adults (Pomerantz, Saxon, & Oishi, 1999), placing a great deal of importance on meeting standards is associated with the view that one's self-worth is contingent on successfully meeting standards. The contingent self-worth associated with investment in meeting standards may foster psychological trade-offs (Higgins, 1991). On the one hand, people placing importance on meeting standards may fear that failure to meet standards will threaten their self-worth. Hence, failure is viewed as quite upsetting. This view may foster worrying over performance, thereby leading to anxiety symptoms. On the other hand, people placing importance on meeting standards may endeavor to protect their self-worth by exerting a great deal of effort to meet their standards, resulting in a good deal of success. This may lead to positive self-evaluation, consequently promoting positive emotions (Higgins, 1991). However, when failure is encountered, people invested in meeting standards may be particularly vulnerable to negative self-evaluation and its consequences (Higgins, 1987, 1991). Because parents are more likely to treat girls than boys in a manner that may foster investment in standards, females may be more vulnerable to anxiety symptoms than males. Yet females may be less vulnerable to negative self-evaluation and consequent depressive symptoms than males when effort fosters success. When effort does not foster success, however, females may be more vulnerable to depressive symptoms than males. Our recent research has yielded a good deal of evidence suggesting that when children and adults are concerned with meeting standards, they are prone to worrying and anxiety symptoms. We have found that children (ages 7 10) who place importance on doing well at an intelligence task are particularly likely to feel nervous right before the task (Kenney & Pomerantz, 1999). In addition, children (ages 9 - 12) who place importance on doing well in school, possessing social skills, and pleasing others are highly likely to worry about their performance in these areas (Pomerantz, et al., 2000). Even when people (college students) list their own standards, there is a link between being invested in meeting standards and worrying over whether one will be able to meet standards (Pomerantz et al., 2000). Notably, this link is evident in concurrent analyses as well as in longitudinal analyses predicting worrying from concern with meeting standards at an earlier time while adjusting for worrying at the earlier time. Consistent with our proposal that concern with meeting standards fosters worrying because it is accompanied by the view that failure will be very upsetting, the link between concern with meeting standards and worrying is mediated by such a view among both children and college students (Pomer-

< previous page

page_70

next page >

< previous page

page_71

next page >

Page 71 antz et al., 2000). There is also evidence suggesting that concern with meeting standards has positive consequences, particularly the promotion of positive emotions and the inhibition of depressive symptoms. This has been documented by Emmons (1986) in concurrent analyses. In our work (Pomerantz et al., 2000), evidence for such a link is also evident in longitudinal analyses predicting depressive symptoms from concern with meeting standards at an earlier time while adjusting for depressive symptoms at the earlier time. Among both children and college students, the positive selfperceptions of competence associated with concern with meeting standards mediated the relation between such concern and positive emotions. It appears that the link between concern with meeting standards and selfperceptions may be due to the heightened effort fostered by the concern with meeting standards (Pomerantz et al., 2000). Notably, as suggested by Higgins (1987, 1991), when people concerned with meeting standards face failure, they may be particularly prone to depressive symptoms. Indeed, Higgins and colleagues (1997) provided evidence that this is the case. Moreover, Dweck and colleagues (see Burhans & Dweck, 1995, for a review) have shown that children who feel that their self-worth is contingent on their performance react negatively to failure. Social-Contextual and Social-Cognitive Developmental Changes Social-contextual and social-cognitive changes that children experience as they progress through elementary school and into adolescence are postulated to intensify the relations among children's concern with meeting standards, their self-evaluation, and their internalizing symptoms (Rholes, Newman, & Ruble, 1990; Ruble, 1994; Rutter, 1986). This is anticipated to heighten the sex differences. The transition from elementary school to junior high school is a critical social-contextual change that children experience. This transition is often characterized by heightened failure because standards may no longer be met simply through increased effort in junior high school, as was the case in elementary school (see Eccles et al. 1993; Ruble & Seidman, 1996, for reviews). The transition from elementary to junior high school is also characterized by a good deal of uncertainty about how to go about meeting standards (see Eccles et al., 1993; Ruble & Seidman, 1996, for reviews). Consequently, even if standards may be met through effort, children may not know how to channel this effort. These two aspects of the transition from elementary to junior high school may lead children concerned with meeting standards to view themselves negatively because they feel incapable of meeting standards. This negative self-evaluation may foster feelings of disappointment and shame, leading to depressive symptoms among children concerned with meeting standards. Therefore, because parents are more likely to treat their daughters than their sons in a manner fostering the development of concern with meeting standards, the transition may heighten negative self-evaluation and depressive symptoms more for females than for males. Consistent with this proposal, research finds that girls are more likely than boys to experience decrements in self-esteem and increments in depressive symptoms following such transitions (see Nolen-Hoeksema & Girgus, 1994, for a review). In terms of social-cognitive developmental changes, a substantial body of research indicates that, as children progress through elementary school, they increasingly conceive of ability as a stable capacity (see Rholes et al., 1990; Stipek & Mac Iver, 1989, for reviews).2 Conceiving of ability in this manner implies that poor performance cannot simply be attributed to a temporary lack of effort, but has implications for future performance. Consequently, performance-relevant information may be more meaningful to older than to younger children. Negative information may be more likely to lead to feelings of hopelessness in older than in younger children because hopelessness is only possible when one's situation (e.g., a lack of ability) is viewed as stable. Feelings of hopelessness are critical to depressive symptoms (Abramson et al., 1989), and thus conceiving of ability as stable may be essential to the onset of depressive symptoms. Conceiving of ability as a stable capacity may therefore influence children's vulnerability to depressive symptoms more than to anxiety symptoms. Indeed, we have found that when failure is encountered, concern with meeting standards is more likely to be associated with depressive symptoms among children who possess conceptions of ability as a stable capacity than among children who do not possess such conceptions (Saxon, Altermatt, & Pomerantz, 1997). This pattern, however, was not evident for anxiety symptoms. Thus, even in the face of failure, sex differences in depressive symptoms may not be evident before the late elementary school years, when children develop a conception of ability as a stable capacity. The introduction of social-contextual and social-cognitive developmental changes suggests why the sex differences in anxiety symptoms may emerge prior to the sex differences in depressive symptoms. When the criteria for success are not difficult to meet, being concerned with meeting standards may foster anxiety, but

< previous page

page_71

next page >

< previous page

page_72

next page >

Page 72 not depressive, symptoms. The more concerned children are about meeting standards, the more anxiety symptoms they may experience as they worry over the possibility of failure. However, such children may also be more motivated to meet such standards. As a consequence, they may exert more effort, resulting in success. Success may in turn foster positive self-evaluation, thereby inhibiting depressive symptoms. Thus, because parents treat girls more than boys in a manner that may foster investment in standards, girls may be more prone than boys to anxiety, but not to depressive, symptoms prior to junior high school, when the criteria for success are generally not difficult to meet. During this time, even when failure is encountered, it may not foster depressive symptoms because it is not viewed as due to stable causes. Once children make the transition to junior high school and develop a conception of ability as a stable capacity, however, girls may become more prone to depressive symptoms than boys. Conclusions The approach we have described here integrates ideas from the prior gender stereotype and gender socialization models. It goes beyond these models in accounting for the complex findings on parents' differential treatment of girls and boys as well as those on sex differences in self-evaluation and consequent internalizing symptoms. There is clearly a long way to go toward fully validating the model, as we have only begun to test it empirically. However, the empirical evidence we have collected thus far is consistent with the model. It is our hope that this approach will lead to programmatic research on parental gender socialization and sex differences in self-evaluation. Moreover, in addition to elucidating the development of sex differences in self-evaluation, it is our hope that the model will elucidate the processes involved in self-evaluation in general. Despite the strengths of our model, a central weakness revolves around the consistent finding that the sex differences in self-evaluation depend in part on the extent to which gender stereotypes characterize the area in which the self-evaluation is made – females evaluate themselves more negatively than males in stereotypically male areas and more positively than males in stereotypically female areas. One explanation for females′ more negative self-evaluation in math that fits with our model may be found in Dweck′s (1986; Licht & Dweck, 1984) suggestion that the sex differences in math are due to the fact that the criteria for success are often ambiguous and difficult to meet. In addition, Ruble (Frey & Ruble, 1987; Ruble et al., 1993) and Crandall (1969) suggested that females′ tendency to evaluate themselves more positively than males in stereotypically female areas is merely a reflection of their greater skills in these areas. If analyses adjusted for actual skills, females would evaluate themselves more negatively than males. Another possibility is that the pattern of parental gender socialization observed by Pomerantz and Ruble (1998) only occurs in stereotypically male areas and that the reverse pattern occurs in stereotypically female areas. The empirical examination of these speculations is a fruitful direction for future research. GENERAL CONCLUSIONS We began this chapter by discussing the issue of what motivates self-evaluation. This issue has been at the heart of much research concerned with self-evaluation in general. Yet investigators concerned with sex differences in selfevaluation have generally not sought to explore this issue. As noted earlier, the sex differences in reactions to failure suggest that females may be motivated by self-assessment whereas males may be motivated by self-enhancement. Roberts (1991) is one of the few investigators who has focused on sex differences in motives. She suggested that females are more likely than males to approach achievement situations with self-assessment and self-improvement goals. In contrast, males are more likely than females to view such situations as competitive in nature and, consequently, to be guided by self-enhancement goals. The approach we have proposed suggests that females may be more motivated than males to maintain their self-worth by meeting standards. This may result in females possessing a stronger self-regulatory system than males (Higgins, 1991). As a consequence, females may be more likely than males to be guided by self-assessment and self-improvement goals in an attempt to meet their standards. However, when improved effort does not foster success, females may drop these goals and adopt self-enhancement goals. Hence, under such conditions, females may have stronger self-enhancement motives than males. In all, females′ motives may be more sensitive. Consistent with this view, Lenney (1977) suggested that females′ selfevaluation was more variable than that of males. Yet the fact that females evaluate themselves more negatively than males in stereotypically masculine areas and males evaluate themselves more negatively than females in stereotypically feminine areas points to several other motives that may also be important (see earlier discussion). The issue of what motivates self-evaluation for females and males needs to be pursued in future research. In this vein, it is important for future research to disen-

< previous page

page_72

next page >

< previous page

page_73

next page >

Page 73 tangle to what extent the sex differences are due to motivational factors and to what extent they are due to socialcognitive factors. Hence, sex differences in the wide range of social-cognitive processes implicated in self-evaluation need to be examined. In conclusion, we have argued that research needs to consider the role played by both gender stereotypes and gender socialization in the development of the sex differences in self-evaluation. An integration of these two perspectives in addition to more general perspectives taken in studying self-evaluation is essential to fully identify the circumstances under which the sex differences emerge. The approach we have proposed has the potential to account for the complexity in the sex differences in self-evaluation and internalizing symptoms. It is our hope that this approach, based on theory and research from a number of areas within psychology, will lead to programmatic research on the factors that lead to the sex differences in self-evaluation while also elucidating basic processes involved in self-evaluation in general. ACKNOWLEDGMENT Preparation of this chapter was facilitated by grant R29 MH57505-01A2 from NIMH and ORWH and grant 9809292 from NSF to the first author. We are grateful for the constructive feedback we received on the ideas presented in this chapter from the Social Cognition Group at the University of Illinois. We also appreciate Ellen R. Altermatt's helpful comments. ENDNOTES 1. Roberts' (1991) model of sex differences in self-evaluation is the only one that does not fit into either of these categories. Thus, given space limitations, we do not give much attention to this model in our review. However, we do highlight where it overlaps with the other models we discuss. 2. Conceptions of ability as a stable capacity that changes as children progress through elementary school appear to represent a different construct than the entity theories on which Dweck and Leggett (1988) have focused (Pomerantz & Ruble, 1997).

< previous page

page_73

next page >

< previous page

page_74

next page >

page_74

next page >

Page 74 This page is intentionally left blank.

< previous page

< previous page

page_75

next page >

Page 75 5 Evaluating the Self in the Context of Another: The Three-Selves Model of Social Comparison Assimilation and Contrast Hart Blanton State University of New York at Albany From the straightforward premise that evaluation of the self is relative, social comparison theory makes predictions on such diverse topics as self-regulation, subjective well-being, interpersonal attraction and rejection, group formation, group conflict, prejudice, and perceptions of injustice. Unfortunately, the history of this theory is anything but straightforward. To begin, social comparison is not one theory but two. One theory is about motivation and the factors influencing the type of social comparison information people seek from their social environment. The primary dependent variable in this research has been social comparison choice, To a lesser degree, research has also examined efforts to change the self or others so that desirable comparisons can be made. The second social comparison theory is about self-evaluation and the factors influencing the effects social comparisons have on judgments of the self. The primary dependent variables in this research have been mood and state changes in selfesteem. Of the two, the first theory has received the most empirical attention. A possible account for this is that motivation was Festinger's (1954a) interest. His views on motives were captured succinctly in the first two hypotheses of his original theory. In Hypothesis I, he argued, “There exists, in the human organism, a drive to evaluate his opinions and his abilities” (p. 117). This statement lead to Hypothesis II, “To the extent that objective, nonsocial means are not available, people evaluate their opinions and abilities by comparison respectively with the opinions and abilities of others” (p. 118). The research that followed these pronouncements has established that Festinger's perspective was lacking in two regards. First, it underestimated the variety of motives contributing to interest social comparison information. The motive to obtain accurate self-evaluation is one. Depending on the context, however, social comparison information may be sought to help in efforts toward self-enhancement, self-verification, and selfimprovement, or to help with any number of other personal or interpersonal motivations (see Major, 1994; Taylor & Lobel, 1989; Wood & Taylor, 1991). The second way in which Festinger's original perspective has proved lacking is that it underestimated the desire for social comparison information. People seek such information even when “objective, non-social means” are readily available. Thus, to the extent that one is motivated to self-evaluate (or to self-verify, or to improve, etc.), social comparison is often the first source of information sought (Wood, 1989). Whereas a great deal of research has advanced a theory of comparison motives, relatively little has advanced a theory of comparative evaluation. However, a host of seemingly contradictory findings suggest that more attention should be paid to this second theory (see Buunk & Ybema, 1997; Collins, 1996; Tesser, 1988). For instance, research has demonstrated that both upward comparison with superior others and downward comparison with inferior others can both raise and lower self-evaluation. As a result, it is often difficult determining what motive is driving an interest in social comparison information in any given situation. To illustrate, someone who seeks upward social comparison may be doing so because of a motive to self-enhance, if this is a circumstance that will raise selfevaluations. Alternatively, this person may be doing this because of a motive to improve, if this is a circumstance that will lower self-evaluation and increase commitment to a goal or if this is a circumstance that will raise selfevaluations and increase feelings of self-efficacy. Finally, this person may be doing this because of a motive to selfverify, if this is a circumstance that will validate an assumed similarity between the self and an exceptional other. Thus, to decode the motives driving different comparison choices across different contexts, it would be informative to clarify the effects that different comparison choices are likely to have on self-evaluation.

< previous page

page_75

next page >

< previous page

page_76

next page >

Page 76 A COMPARISON-BASED THEORY OF SELF-EVALUATION It is the goal of this chapter to introduce a framework that predicts the effects of social comparison on selfevaluation. This effort is indebted to a number of models identifying factors that moderate reactions to social comparison information (e.g., Brewer & Weber, 1994; Buunk & Ybema, 1997; Collins, 1996; Tesser, 1988). However, the current chapter uses an outside source as its guide. Rather than turning to research within the social comparison tradition, which has focused mostly on the motives for comparison, it turns to research within the social cognition tradition. Specifically, it looks at social-cognitive theory linking categorization and comparison processes to the evaluation of social objects. This research tradition is applied to the social comparison literature to predict how categorization of and comparison with another person influences evaluation of the self (cf. Mussweiler & Strack, 2000). It is beyond the scope of this manuscript to review and integrate the many disparate theories linking categorization and comparison processes to evaluation (e.g., Herr, Sherman, & Fazio, 1983; Martin, 1986; Parducci, 1965; Sherif & Hovland, 1961; Tajfel, 1959; Tajfel & Wilkes, 1963; Thibau & Kelly, 1959; Turner, 1987; see Stapel & Koomen, chap. 14, this volume). Instead, this chapter utilizes one theoretical perspective – the Schwarz and Bless (1992a) inclusionexclusion model. Although the Schwarz and Bless model draws on previous theories and, in turn has influenced a number of more recent frameworks, it seems to be the model best suited for organizing social comparison predictions related to self-evaluation. This is in large part because it is based on straightforward and established rules of categorization and comparison. Thus, application of the model does not require distinctions between different mechanisms that ultimately lead to the same evaluative outcome. This feature is desirable with social comparison, which is a complex social phenomenon potentially affected by a multitude of factors. Inclusion and Exclusion The inclusion-exclusion model assumes that any evaluation requires two mental representations: A representation of the target stimulus and a representation of a standard. The target stimulus is the object that is being evaluated, and the standard is a point of comparison allowing the perceiver to locate the standing of the target stimulus on the dimension for which it is being evaluated (Sherif & Hovland, 1961). The representation of the target and the representation of the standard are assumed to be constructed at the time of the evaluation on the basis of information that is both chronically and temporarily accessible (Kahneman & Miller, 1986). The latter information varies from situation to situation and forms the basis of context effects on social judgment. Consistent with prior models, the inclusion-exclusion model states that whether contextual information causes assimilation or contrast depends on how that information is categorized. If the information is included in the representation of the target, assimilation will result and evaluation of the target stimulus will be displaced toward the evaluation of the contextual information. If the information is included in the representation of the standard, contrast will result and the evaluation of the target stimulus will be displaced away from the evaluation of the contextual information. Finally, if the information is not included in either the representation of the target stimulus or the standard, neither contrast nor assimilation will result. As an example of inclusion-exclusion reasoning, a recent study by Stapel and Schwarz (1998a) assessed the effect that a context prime, General Colin Powell, had on evaluations of two separate target stimuli, the Republican Party and Republican Senator Bob Dole. Because Colin Powell is an example of a Republican, he should be included in the representation of the Republican Party, resulting in an assimilation effect in evaluations of the Republican Party. However, Colin Powell is not an example of Bob Dole because Bob Dole forms his own bounded category, of which Colin Powell is not a member. Thus, Colin Powell should be excluded from the representation of Bob Dole and no assimilation should result. However, the presence of Colin Powell could influence the evaluation of Bob Dole if Colin Powell is considered relevant as a standard against which to evaluate Bob Dole. Because most would evaluate Bob Dole in terms of his standing as a Republican presidential candidate, and because Colin Powell is a Republican who has also been considered as a potential candidate, Colin Powell would be a reasonable standard against which to evaluate Bob Dole. Thus, reflecting on Colin Powell should cause a contrast effect on evaluations of Bob Dole. These predictions were supported in the study by Stapel and Schwarz: Because Colin Powell was popular at the time of the study, reflecting on him prior to judgment raised evaluations of Republicans (assimilation) and lowered evaluations of Bob Dole (contrast; see also Moskowitz & Skurnik, 1999; Schwarz & Bless, 1992b; Stapel & Koomen, chap. 14, this volume). In summary, the same contextual variable can have either a contrast or assimilation effect on a target depending on how it is

< previous page

page_76

next page >

< previous page

page_77

next page >

Page 77 categorized. Inclusion in the target encourages assimilation, and inclusion in the standard encourages contrast. Ultimately, the size of assimilation and contrast effects depend on the extremity of the contextual information. For instance, if Colin Powell were rated no differently than the average Republican, including him in the representation of his party would have had no influence on evaluations. Similarly, if Colin Powell were not evaluated either more or less positively than Bob Dole, including him in the representation of the standard would have had no influence on evaluations (Moskowitz & Skurnik, 1999). Thus, whether a contextual variable will have measurable effects on evaluation depends both on the way in which that information is categorized and the extremity of that information in comparison to the other information already represented in the category.1 Terminology Before applying the inclusion-exclusion model to the domain of social comparison, it is necessary to clarify terms. In social comparison language, it is typically stated that a person engages in social comparison with one or more target persons. Here the term target is usually (although not always) meant to imply that the other person has been used as a temporary standard for self-evaluation. If one were to use the language of inclusion-exclusion, however, the target in this situation would be the self because the self is the social object being evaluated. The other person would then constitute the judgment context because the other person has introduced information that may alter the evaluation of the self. Because of these differences in terminology, a separate set of terms is used to avoid confusion. Throughout this chapter, social comparison refers to any situation in which a person's self-evaluation can be influenced by information made accessible by the real or imagined presence of another person. Because the focus of this chapter is on the effects of social comparison information on self-evaluation, social comparison is only considered as it relates to the comparison of abilities and valenced traits. The focus is specifically on the effects of upward social comparison with someone who is superior to the self versus downward social comparison with someone who is inferior to the self. In each case, it is assumed that the upward or downward social comparison information is extreme enough to influence self-evaluations. During social comparison, the individual who engages in social comparison is referred to as the perceiver and the other person who makes social comparison information accessible is referred to as a comparison other , or at times as an exemplar . An assimilation effect is said to occur when self-evaluation is displaced toward the evaluation of the comparison other (raising self-evaluation following upward social comparison and lowering self-evaluation following downward social comparison). A contrast effect is said to occur when self-evaluation is displaced away from the evaluation of the comparison other (lowering selfevaluation following upward social comparison and raising self-evaluation following downward social comparison). Based on the inclusion-exclusion model, it is assumed that self-evaluation requires two mental representations: The first is a representation of self, which is the social object being evaluated. The second representation is the standard , which is a point of comparison used to locate the self on the dimension for which it is being evaluated. In traditional social comparison theory, the standard is nothing more than a representation of the comparison other in terms of that person's standing on the dimension by which the self is being evaluated (e.g., a representation of the standard that is based on the comparison other's intelligence during evaluations of one's own intelligence; Morse & Gergen, 1970). In the current framework, the standard is any mental representation used to locate the self on the evaluative dimension. Social comparison occurs when a comparison other introduces evaluative information that is extreme enough to alter the judgment context. Thus, the comparison other can alter either the representation of the standard or the representation of the self. The latter effect would occur if the comparison other is used as a proxy for the self during self-evaluation (e.g., a representation of the self that is based on the comparison other's intelligence during evaluations of one's own intelligence; Lockwood & Kunda, 1997). This broader definition of social comparison , in which the comparison other can alter either the representation of the standard or the representation of the self, is necessary for predicting whether contrast or assimilation will result. INCLUSION-EXCLUSION AND SOCIAL COMPARISON How might the presence of exemplars influence self-evaluation? Three possibilities seem most likely. First, when accessible information about the exemplar is included in the representation of the standard, contrast results and selfevaluation is displaced away from evaluation of the exemplar. Second, if information about the exemplar is included in the representation of the self, assimilation results and self-evaluation is displaced toward evaluation of the exemplar. Finally, if information about the exemplar is included in some shared social

< previous page

page_77

next page >

< previous page

page_78

next page >

Page 78 category, of which both the perceiver and the exemplar are members, then assimilation occurs if self-evaluation is based on the evaluation of that category. To determine which of these three outcomes occurs, a mechanism is needed that can predict how comparison information is categorized. The current framework identifies two, both of which have received a great deal of research attention. First is the perception of similarity . Social comparison research has long considered the role that similarity plays in determining whether another person is considered a relevant source of social comparison information. For the most part, similarity has been found to translate into relevance. However, whether comparison relevance leads to inclusion in the self or inclusion in the standard depends on the second mechanism, which is the type of working self-concept that is accessible in memory. Consistent with prior theory, the self is conceived as a complex collection of interrelated self-conceptions that become differentially accessible depending on the current judgment context (e.g., Cantor & Kihlstrom, 1983; Gergen, 1967; Markus & Wurf, 1987; Mead, 1934; Rosenberg, 1979; Turner, 1985). When individuals form judgments about themselves, the type of self-concept that is active predicts how comparison information from similar others can be categorized in relation to the self. The current framework distinguishes three broad classes of working self-conceptions. These are: individual selves, possible selves, and collective selves. Because the effects of similarity are assumed to depend on the self that is active, predictions for self-evaluation are covered separately in the sections that follow. Individual Selves The individual self refers to perceivers' conceptions of their current and personal attributes. In many ways, it appears that the individual self-concept is the default value of the working self-concept (cf. Markus & Kitayamma, 1991). For instance, when asked to describe themselves, people spontaneously mention attributes that distinguish them from salient others (McGuire & McGuire, 1980; McGuire, McGuire, Child, & Fujioka, 1978; McGuire & PadawerSinger, 1976; see also Gilbert, Giesler, & Morris, 1995), and their self-evaluations are based more on evaluations of their distinct attributes than their evaluations of common attributes (Ditto & Griffin, 1993). Apparently for these reasons, people often base their behavioral choices on the perceived social consequences of distinguishing the self from others within the social environment, instead of basing them on the perceived consequences of becoming similar to others (Blanton, Stuart, & VandenEijnden, in press). Interestingly, research on social comparison has shared this default interest in the individual self. Of the studies that have examined comparison effects on self-evaluation, most have occurred in contexts that would increase the salience of individual attributes. As examples, participants have been given comparison information in the context of standardized testing (e.g., Major, Schiacchitano, & Crocker, 1993; Tesser, Millar, & Moore, 1988) or during competition with another (e.g., Morse & Gergen, 1970; Pleban & Tesser, 1981). Based on the principles of categorization, comparison information in these studies should not cause assimilation because comparison others are excluded from the representation of the self. However, comparison others can cause contrast effects when they are considered relevant to comparison by being included in the representation of the standard. For this to occur, it is typically necessary for the comparison other to be perceived as similar to the self. Similarity Perhaps no sentence has guided social comparison research more than ‘‘Given a range of possible persons for comparison, someone close to one's own ability or opinion will be chosen for comparison” (Festinger, 1954a, p. 121). This, Corollary III A of “A Theory of Social Comparison Processes,” was based on Festinger's (1954a) third hypothesis that the desire for social comparison decreases as a target becomes less similar. For Festinger (1954a, 1954b), a similar other was thought to provide a meaningful standard for self-evaluation, whereas a dissimilar other was not. Interestingly, Corollary III A seems to have inspired research as much because it generated confusion about what Festinger meant by a similar other as by any desire to test his proposition (Wheeler, 1991; Wood, 1989). At times, he seemed to be referring to similarity on the evaluative dimension , which is the performance, trait, or ability being assessed. At other times, he seemed to mean similarity on surrounding dimensions, which are the attributes, traits, or abilities not being evaluated. Most of the initial studies testing the similarity hypothesis measured interest in similarity on the evaluative dimension (Wheeler, 1966; Wheeler et al., 1969). These studies used the rankorder paradigm, in which participants were given performance feedback and then allowed to select, on the basis of rank, the test score(s) of others that they would like to view. Findings generally supported the similarity hypothesis: People preferred to compare their

< previous page

page_78

next page >

< previous page

page_79

next page >

Page 79 performances with others at a similar rank (e.g., Gruder, 1971; Gruder, Korth, Dichtel, & Glos, 1975; Wheeler, 1966; Wheeler, Koestner, & Driver, 1982; cf. Wheeler, 1991). However, subsequent research gave more attention to the features of a surrounding dimension that might compel a social comparison. The Related-Attributes Hypothesis . The first major revision of the similarity hypothesis to address surrounding dimensions was best articulated by Goethals and Darley (1977; see also Wheeler et al., 1969; Wilson, 1973). They argued that, by similarity, Festinger (1954a, 1954b) meant people who are similar on surrounding dimensions that are “related to and predictive of performance” (p. 265). This idea, termed the related-attributes hypothesis (Wheeler & Zuckerman, 1977), was based on attribution theory. People were thought to seek comparison with others who would be expected to perform at about the same level as themselves, based on background variables, to uncover diagnostic standards against which to assess internal attributes and abilities. For instance, a runner who compares her speed to other women of the same age and training history can get a better sense of her underlying running ability than if she were to compare herself to men of a different age and training history. Research has generally supported the claim that people seek comparisons with others who are similar on related attributes (e.g., Gastorf & Suls, 1978; Gastorf, Suls, & Lawhon, 1978; Sanders, Gastorf, & Mullen, 1979; Wheeler & Koestner, 1984; Wheeler, Koestner, & Driver, 1982). For instance, people typically prefer comparisons with others who are similar on related dimensions even when they already have information on the evaluative dimension (Wheeler & Koestner, 1984). However, related attributes are not the only criteria used to select comparison others, and people often seek comparison with others who are similar on the evaluative dimension as well (Wheeler, Koestner, & Driver, 1982). Identity Comparisons . Although the related-attributes hypothesis clarified the conditions under which someone should use another person as a standard of self-evaluation, it did not account fully for the preferences for surrounding similarity. In fact, much as early attribution theory proved effective in identifying the conditions under which people deviate from rationality (Jones, 1985), the related-attributes hypothesis proved useful in identifying the conditions under which comparisons deviate from diagnosticity. For instance, various studies found that participants preferred comparisons with others of the same gender even when they knew that gender was not a related attribute (e.g., Feldman & Ruble, 1981; Major & Forcey, 1985; C.T. Miller, 1984; Suls, Gaes, & Gastorf, 1979) and even when they had information that was more diagnostic of performance (Tesser & Campbell, 1980). Analogous preferences were shown for comparing with others similar in physical attractiveness (C.T. Miller, 1982), age, and race (Tesser, 1986). Various explanations have been offered for the robust preference for similarity (see Miller & Prentice, 1996; Wood, 1989), and it appears that identity concerns play a consequential role. C.T. Miller (1984) found that women who identified strongly with their gender made comparisons within gender even when it was unrelated to the evaluative dimension. However, women who did not identify with their gender compared within gender only when it was a related attribute. D.T. Miller, Turnbull, and McFarland (1988) linked such comparisons to reference group theory (Hyman, 1960; Merton & Kitt, 1950; Singer, 1966) and demonstrated that people often choose standards that help them determine how they stack up within personally relevant reference frames. In this regard, gender and ethnicity seem particularly meaningful, such that people often ignore out-group standards and base their self-evaluations on their standing within their in-group (Crocker & Major, 1989; Major, 1994). Effects of Similarity. In summary, similarity exerts a strong influence on perceiver's choice of comparisons. However, studies on comparison choice do not demonstrate that similar others have greater effects on self-evaluations than dissimilar others. They also do not demonstrate that contrast always results. The guiding assumption behind virtually all of the studies just cited is that it is more meaningful to assess one's own ability relative to a similar other than it is to assess one's abilities relative to a dissimilar other. It follows that contrast should be magnified when comparison others are similar. The most direct test of this proposition in terms of evaluative similarity was conducted by Sanders (1982). He exposed undergraduates to social comparison information regarding their performance on an analogy test and found that small discrepancies in performance between the self and comparison others produced the greatest changes in ability evaluations. With small discrepancies, a contrast effect emerged. As the comparison discrepancy widened, the effects of comparison information diminished to the point that extremely discrepant comparison information had no greater effect on self-evaluations than slightly discrepant comparison information. In theory, if this discrepancy between self and other continued to widen, comparison information would have ceased to influence self-evaluations (see Radloff, 1966; Wilson, 1973; Wood, Taylor and Lichtman (1985). Just such an effect was demonstrated in a

< previous page

page_79

next page >

< previous page

page_80

next page >

Page 80 recent study of pay satisfaction by Brandstätter (1999). Participants estimated the satisfaction of two individuals differing in degree of similarity in salary. At first, small decrements between the two lead to contrast estimates on pay satisfaction. As this gap widened, the contrast effect leveled off and, at the extremes, estimates began returning to baseline. The ironic consequence of this pattern of responses is that it suggests that people will at times prefer the company of others who are much better than them on an evaluative dimension, rather than choosing the company of others who are only slightly better than them (see Mettee & Smith, 1977). Research on surrounding similarity also indicates that increased similarity can lead to increased contrast (i.e., Brewer & Weber, 1994; Major, Schiacchitano, & Crocker, 1993; Nadler & Fisher, 1976). As an example, Major et al. (1993) found that the superior performance of a comparison other lowered self-evaluation when comparison others shared the same minimal group, but this had no effect on self-evaluation when the comparison was a member of the outgroup. Most of the remaining evidence linking surrounding similarity to contrast has been conducted by Tesser and colleagues. Tesser (1986, 1988) proposed what he termed a self-evaluation maintenance model of social comparison. This model states that social comparison should have contrast effects on self-evaluation when the perceiver is psychologically close to the target. Closeness in this formulation is linked to what Heider (1958) referred to as a unit relationship. This is a broad term that encompasses surrounding similarity in terms of both related attributes and a shared identity. Consistent with the model, when perceivers are outperformed on important domains, closeness has been linked to negative affect (Tesser, Millar, & Moore, 1988), task disengagement (Tesser & Campbell, 1980), and distancing from the comparison other (Pleban & Tesser, 1981). Although these studies did not directly measure effects on self-evaluation, findings can be explained by assuming that the participants′ responses reflected a desire to maintain positive evaluations of the self in contrast to close others. Interestingly, Tesser (1988) also proposed that closeness can lead to assimilation. People are thought to cease evaluating the self when the evaluative dimension is not important to self-esteem. Freed from caring about their relative standing, people do not seek out standards against which to self-evaluate. Instead, they reflect on their associations with close others, causing assimilation (Cialdini et al., 1976; Cialdini & Richardson, 1980). Because this part of his model is not concerned with evaluation of individual attributes, discussion on this is held until later. In summary, social comparison theory has traditionally focused on comparisons of individual attributes. Although much of the work in this area has refined the definition of similarity or focused on comparison choice, similarity does appear to increase contrast effects in contexts in which the self-concept is based on individual attributes. Possible Selves Although the individual self may be the spontaneous representation for the perceiver, alone it is not sufficient for interpreting the diversity of human experience. One reason that the working self-concept is thought to be contextually fluid is that flexibility is necessary for people to interpret the full range of their current, past, and future actions (e.g., Cantor, Markus, Niedenthal, & Nurius, 1986; Gergen, 1972; James, 1910; Little, 1983). In this regard, Markus and Nurius (1986) argued for the importance of possible selves. These are defined as self-representations “that have not been verified or confirmed by social experience” but that nonetheless exert “a significant influence on individual functioning” (p. 955). Markus and Nurius suggested two primary functions of possible selves. First, they act as incentives that guide future behavior; second, they act as interpretive structures that lend meaning to current experience. As incentives, possible selves can either be positive “hoped-for selves” that one tries to attain or they can be negative “feared selves’’ that one tries to avoid. For example, high school students should exert effort into their school work when they have hoped-for selves as college students and feared selves as highschool dropouts. As interpretive structures, possible selves guide reactions to current situations. For instance, a student with a hoped-for self as a college student would probably feel like a failure after getting a “C” in a class, whereas the student who fears failing out of school might feel like a conquering hero. Numerous studies support both the motivational and interpretive functions of possible selves, indicating that they influence achievement striving and subjective well-being above and beyond what would be predicted from current self-concepts (e.g., Ogilvie, 1987; Oyserman, Gant, & Ager, 1995; Oyserman & Markus, 1990; Oyserman & Saltz, 1993; Quinlan, Jaccard, & Blanton, 1999; Ruvolo & Markus, 1992; Ryff, 1991). For clarification, it should be noted that positive and negative possible selves are often viewed as similar to Higgins′ (1987) notions of ideal and ought self-guides. However, theory related to possible selves addresses how self-evaluation, affect, and motivation are influenced by representations of the self, whereas Higgins′ self-discrepancy theory addresses how these same outcomes are influenced by

< previous page

page_80

next page >

< previous page

page_81

next page >

Page 81 the type of standard against which the self is evaluated. The current model assumes that these two processes are inter-related, with the representation of the self influencing the representation of the standard and vice versa (Kahneman & Miller, 1986). For clarify of focus, however, the current model considers only on the effects that changing self-representations have on comparison and evaluation. Possibility Through Comparison As a working self-concept, the possible self differs from the individual self by extending beyond individual attributes to include attributes that others possess. In the present framework, comparison others can influence possible selves in one of two ways. First, they can make new possibilities available that would have been unknown to perceivers in the absence of social comparison. As an example, any successes of the American women's soccer team may cause young girls in the states to consider the possibility of becoming soccer players, when they would not have entertained this possibility before. Second, comparison others can make previously stored possibilities accessible in working memory in situations in which they would not otherwise have been considered. Thus, a usually confident soccer player who is experiencing a moment of self-doubt may take inspiration from a teammate who scores an unexpected come-from-behind goal. In each instance, the comparison other alters the temporary representation of the self, which can then influence self-evaluation. Although it may be intuitive that people assess what is possible for the self by considering the qualities of others, it is important to note that the alternative to this does not seem likely. Specifically, it does not seem that people will infer possible selves from considering attributes that others do not have. Thus, it is not predicted that a child who observes that most of her peers are different from her and that most of her peers are bad athletes will then conclude she has the potential of becoming a good athlete. This observation was made early in social comparison theory. Before the dominance of downward comparison theory (Wills, 1981), in which people were thought to maintain esteem by contrasting themselves with worse-off others, it was upward social comparison that was linked to positive self-evaluations. This prediction was based on the notion of the positive instance , which is mostly ignored in current research (Collins, 1996; Wheeler, 1991). The positive instance was the idea that it is easier to imagine the self possessing a desired trait by observing people who have that quality than it is by observing those who do not (see Berger, 1977; Gruder, 1977; Mettee & Smith, 1977; Thornton & Arrowood, 1966). A related and more general construct is the feature-positive effect , which is the tendency for both animals and humans to find it easier learning from an event occurrence than from an event nonoccurrence (Jenkins & Sainsbury, 1969, 1970). Fazio, Sherman, and Herr (1982) linked this to self-perception by demonstrating that people are more likely to draw inferences about themselves through their own actions than through their inaction (see also Allison & Messick, 1988; Cioffi, 1994). The parallel in social comparison is that people should be more likely to infer a possible self from observing people who demonstrate a possibility than from observing people who do not. In summary, when the working self-concept is represented in terms of its possibility, comparison others are predicted to influence self-evaluation through the exemplar's inclusion in the representation of the self rather than its exclusion. As a result, assimilation is generally predicted and contrast is not.2 Including Possibility The classic study demonstrating that comparison information influences self-evaluation was conducted by Morse and Gergen (1970). Participants were exposed to a job applicant who was competing against them for a desired job. The primary finding was a contrast effect: Subjects who viewed a competent applicant (“Mr. Clean”) reported more negative self-evaluations than those who observed a less competent applicant (“Mr. Dirty”). The reason for this effect is presumably that the evaluative context focused participants on their current attributes, which then seemed inadequate in the presence of a better job candidate. The current framework would predict an assimilation effect in Morse and Gergen (1970), had the evaluative context focused participants on the possible self. This focus would have increased participants' perceptions that they can interview at the level of the superior job applicant. Evidence in support of this prediction can be found in a set of studies by Lockwood and Kunda (1997). In Study 1, participants read a newspaper article about an exceptional and award-winning working professional. Much like the “Mr. Clean’’ in Morse and Gergen, this “Ms. Clean” represented an upward comparison other for the participants. However, the comparison other in this study was too different from the participants on both evaluative and surrounding dimensions for them to view her as a relevant standard against which to evaluate current abilities. This woman was no longer in school, presumably older, and receiving an award for extreme

< previous page

page_81

next page >

< previous page

page_82

next page >

Page 82 accomplishments in the “real world.” Thus, the similarity hypothesis would lead one to expect that this person would not be considered a relevant standard of comparison, and so she should not influence self-evaluations. Nevertheless, Lockwood and Kunda (1997) made it so that participants in some experimental conditions viewed this woman as a future possible self. This was done by manipulating the personal relevance of the exemplar's job title. Participants were preselected to be either education or accounting majors, and the exemplar was described either as a school teacher or an accountant. As predicted, this led to an assimilation effect in conditions in which the exemplar's career matched the participant's career aspirations. The school teacher raised self-evaluations of education majors and the accountant raised the self-evaluations of accounting majors. When the major/career profiles did not match, there was no effect of comparison information on self-evaluations. Similarity As with the individual self, an exemplar only has an effect on evaluation of the possible self if it is considered relevant to self-evaluation. As before, relevance is determined in large part by perceptions of similarity. Although similarity maintains its various meanings, what is most important with the possible self is that the perceiver can imagine becoming like the comparison other at some point in the future. This was demonstrated in Lockwood and Kunda (1997), in which assimilation depended on a shared career path. A variety of other features might contribute to this type of similarity. Recently, Buunk and Ybema (1997) argued that perceivers typically require some individuating information about a person to view their performance level as attainable. They noted that in the typical studies that demonstrate a desire for downward comparison, participants are asked to choose their comparisons based on a single test score or other outcome measure (Gibbons & Gerrard, 1991; Pyszczynski, Greenberg, & LaPrelle, 1985; Swallow & Kuiper, 1993; cf. Taylor & Lobel, 1989; Wheeler, 1966; Wheeler et al., 1969). Preference for downward comparison in these cases probably reflects a desire for self-enhancement. Without any basis for identifying with that person, this is achieved by contrasting the self with worse-off others. However, when people are presented with individuating information about the comparison other or when they have contact with that person, people more typically seek upward comparison (e.g., Molleman, Pruyn, & van Knippenberg, 1986; Rabbie, 1963; Ybema & Buunk, 1993). With more information to work with, it is easier for them to see their own future in the performance of another (see Klein & Kunda, 1992; Pyszczynski & Greenberg, 1987). Thus, the same selfenhancement motive as before can be served by looking for upward comparison with a better-off other, which raises self-evaluations through assimilation. Another important determinant of perceived similarity with possible selves is a perception of control over the attribute under evaluation (Major, Testa, & Bylsma, 1991). By definition, people must believe that they can achieve a target's success if they are to view that person as a possible self. Similarly, for a downward target to become a feared possible self, perceivers should doubt their own ability to keep from becoming like that person. Recently, Buunk and Ybema (1997) linked control perceptions to a variety of different features of the comparison context. For one, control depends on the comparison dimension (Buunk & Ybema, 1997). For instance, Buunk, Schaufeli, and Ybema (1994) found that people were more likely to seek upward comparisons with people at work who had more experience at their job than they were to compare upward with people who were more competent at their job. This preference presumably reflected the relative ease with which participants could imagine gaining experience over gaining competence (see also Buunk, 1995). In addition to the comparison dimension, control depends on time. People are more optimistic that they can achieve a positive target's success if they have the time to become like them (Aspinwall, 1997; Lockwood & Kunda, 1997). An example of this can be found in Lockwood and Kunda (1997). In their second study, they demonstrated that first-year graduate students showed assimilation after reading a vignette about an exceptional fourth-year student, whereas fourth-year students did not. With a couple of years ahead of them in which to achieve the successes of the comparison other, first-year students could see their own future superstar status in the successes of the comparison other. With most of their graduate training behind them, fourth-year students did not see this possibility. In summary, comparison others are predicted to cause assimilation effects when those others are considered possibilities for the self. This occurs when these others are perceived as similar in the sense that their attributes seem attainable and under personal control. Collective Selves Various theories of the self, most notably social identity theory (Tajfel, 1978; Tajfel & Turner, 1979) and its extension, self-categorization theory (Turner, Hogg,

< previous page

page_82

next page >

< previous page

page_83

next page >

Page 83 Oakes, Reicher, & Wetherell, 1987; Turner, Oakes, Haslam, & McGarty, 1994), contend that the working self-concept shifts in focus between representations of the self in terms of individual attributes and in terms of membership in social collectives or groups (see also Markus & Kitayama, 1991; Stryker & Statham, 1985). Although there are important distinctions between the different theories (see Deaux, 1996; Jost, chap. 6, this volume), each shares the assumption that people represent the self in terms of collective selves and that, when this occurs, self-evaluation is based on evaluations of one's group memberships (see Luhtanen & Crocker, 1992). Therefore, when the collective self is accessible, comparison others should influence self-evaluation by altering group evaluation. This then carries over to the evaluation of the self as a member of the group (see Stapel & Schwarz, 1998a). The current framework predicts two ways in which this occurs. First, if an exemplar is a member of the perceiver's in-group, that person can be included in the representation of the in-group. This categorization leads to assimilation effects on group evaluation, which can carry over as assimilation effects on self-evaluation. Second, if an exemplar is excluded from the representation of the in-group (possibly due to membership in an out-group), that person can be included in the standard against which the ingroup is evaluated. This categorization should result in contrast effects on group evaluation, which should then carry over as a contrast effect on self-evaluation. Although both possibilities seem likely, only the former is discussed. This is due in part to the paucity of research on the latter and in part to the belief that discussion of the first instance provides the necessary information for making predictions about the second. As with the individual self, whether the comparison other is considered relevant for comparison depends on perceptions of similarity. People who are perceived as similar have the potential to be perceived as fellow group members, which would make assimilation the predicted outcome. At first glance, this prediction appears to contradict earlier arguments that increased similarity leads to contrast. However, assimilation is predicted with increased similarity when it shifts attention to a shared group, thereby shifting the working self-concept to a representation of the self as a member of that group. Interestingly, similarity on both the evaluative dimension and surrounding dimensions can cause this shift in the representation of the self. Evaluative Similarity Recall that traditional social comparison theory assumes that similarity on the evaluative dimension increases contrast (e.g., Brandstätter, 1999; Sanders, 1982). To illustrate, one can imagine students in a math class evaluating their math ability in terms of their standing relative to others in their class. Due to dissimilarity on the evaluative dimension, however, they view the one exceptional student in class as a math whiz and thus not relevant for comparison (Wood et al., 1985).3 Thus, as already reviewed, comparison between two students in the class who are similar in ability should lead to contrast effects, but comparisons between an average student in the class and the lone math whiz should have no effect on ability evaluations for either student. Interestingly, most social judgment models do not predict this intuitive example because they do not predict increased contrast with increasing similarity, but instead predict increased assimilation (e.g., Eiser, 1990; Herr, 1986; Herr, Sherman, & Fazio, 1983; Manis, Nelson, & Shedler, 1988; Parducci & Wedell, 1990; Sherif & Hovland, 1961). For instance, Herr (1986) had participants evaluate another person who engaged in a behavior that was ambiguously hostile in the context of a moderately or extremely hostile exemplar. When participants were primed with a list of extremely hostile people (e.g., Adolf Hitler), a contrast effect occurred and the target person was judged as relatively less hostile. However, when participants were primed with a reading list of moderately hostile people (e.g., Joe Frazier), assimilation resulted and the target person was judged as relatively more aggressive. Although the various social judgment models have different bases for linking similarity to assimilation and not contrast, the primary reasoning is that people are less likely to differentiate two stimuli if they are similar on the evaluative dimension. Lack of differentiation can occur because discrimination becomes more difficult with increased similarity (e.g., Sherif & Hovland, 1961), because stimuli are seen as sharing the same conceptual category (e.g., Herr et al., 1983), or because one stimuli assists with the encoding of the second stimuli (e.g., Philippot et al., 1991; Stapel & Winkielman, 1998; Stapel & Schwarz, 1998b). The reason that social comparison theory seems to make the wrong prediction is that a failure to differentiate the self from the comparison other is not a possibility in contexts in which the self is represented in terms of its individual attributes. By definition, the individual self creates its own bounded category, of which the exemplar cannot be a member. Thus, similarity can only increase the comparison other′s relevance as a standard, leading to contrast (Stapel & Winkielman, 1998). However, the self is not always distinguished from others and so this should not be the

< previous page

page_83

next page >

< previous page

page_84

next page >

Page 84 universal prediction. In fact, similarity on the evaluative dimension can cause a shift in the working self-concept from the individual to the collective self, causing the perceiver to no longer differentiate the self from the comparison other (Turner, 1985). To illustrate how this happens, consider again the situation in which the comparison other earns the label math whiz. If both the comparison other and the perceiver are doing extremely well in math class – so well that they beat all of their other classmates – this label would be appropriate for describing not just one of them but the pair. As the only two members of the math-whiz group, each should come to think of the other in terms of this shared membership, causing them to see one another as psychologically interchangeable when they are evaluating that category. Turner (1985) captured this feature of categorization with a comparative rule that he termed the meta-contrast principle . This states that, in any given context, two individuals will cease to differentiate one another on an evaluative dimension to the extent that they are less different from one another on that dimension than they are, on average, from others in their frame of reference. For instance, two students should develop a shared math-whiz identity when they both are so much better at math than the rest of the class that their abilities distinguish them from the rest of the class more than from one another (Haslam & Turner, 1992; Oakes & Turner, 1990; Oakes, Turner, & Haslam, 1991; see also Mullen, 1991). Because shared categorization, as defined by the meta-contrast principle, depends on the average similarity within the reference frame, the likelihood of assimilation should vary accordingly. Consider, for instance, what would happen if the two math whizzes earned membership into a school for the gifted. If this new school brought them in contact with many others of similar ability, their previous identity would no longer create a meaningful categorization. Moreover, if the two students now found themselves sharing a similar rank, but only now buried within the middle of the distribution instead of at the extremes, they may begin to see one another as relevant standards against which to evaluate their own personal abilities in the way predicted by the similarity hypothesis. In this context, contrast should result (Oakes & Turner, 1990; cf. Gibbons, Benbow, & Gerrard, 1994). Surrounding Similarity Much like similarity on the evaluative dimension, similarity on surrounding dimensions can shift representation to the collective self. Gender, race, ethnicity, national origin, and even college major can be the basis for a meaningful group identity. Comparison others who share these attributes can cause assimilation effects by altering the perceived status of these shared groups. Of course, the difficulty with predicting assimilation due to similarity on surrounding dimensions is much the same as with predicting it for similarity on the evaluative dimension: A shared group membership can be a basis for considering a person a relevant standard of comparison for the individual self, and so it can increase contrast as well. Thus, predicting assimilation requires some clarification of the conditions under which shared group membership draws attention to the collective self instead of the individual self. Group Threat. One important factor moving evaluation to the group level is a perceived group threat, or a challenge to group status. When the qualities of an important group are called into question, people often become preoccupied with how they will be perceived as members of the group (e.g., Steele & Aronson, 1995). As a result, they should become focused on how they will be perceived through their associations with other in-group members. They should avoid associations with individuals who diminish the status of the group and seek associations with members who enhance it (e.g., Cooper & Jones, 1969; Mackie, 1984; Marques, Robalo, & Rocho, 1992; Marques, Yzerbyt, & Leyens, 1988; Novak & Lerner, 1969). As an example, Mackie (1984) found that ethnic minority school children were more likely to make upward in-group comparisons if they were making comparisons on an evaluative dimension in which negative group stereotypes were made salient. This was interpreted as a largely defensive reaction on the part of participants, who presumably sought evidence of in-group excellence as a way of dispelling negative stereotypes. Although suggestive, Mackie′s (1984) study did not establish that in-group comparison would yield an assimilation effect. Direct evidence can be found in a recent set of studies by Blanton, Crocker, and Miller (2000). They gave African-American participants a bogus test of mathematical ability under conditions that were designed to increase the salience of negative intellectual stereotypes about African-American students. Before taking the test, a White experimenter informed participants that he was in the process of standardizing a measure of math ability for the purpose of determining how a representative sample would perform. This statement was included to create a concern among participants about their group′s performance. After taking the test, participants were told that they had done about average. They then overheard feedback being given to

< previous page

page_84

next page >

< previous page

page_85

next page >

Page 85 either a White or a Black4 confederate, who ostensibly performed extremely well (i.e., 99th percentile) or extremely poorly (i.e., 20th percentile). As predicted, when the confederate was White, a contrast effect occurred: Participants reported more positive self-evaluations following social comparison with a White confederate who had done poorly on the test than they reported after social comparison with a White confederate who had done well (e.g., Morse & Gergen, 1970). However, when the confederate was Black, an assimilation effect occurred: Participants reported more positive self-evaluations following social comparison with a Black confederate who had done well on the test than they reported after social comparison with a Black confederate who had done poorly. In summary, group threats can cause individuals to evaluate the self in terms of the group status, leading to an assimilation effect following in-group comparison. However, group threat is not sufficient to cause this focus. For this to happen, the individual must identify with the group and view the lowered status as illegitimate. When this does not occur, individuals cannot take pride in the superior performance of an in-group member (Crocker & Blanton, 1999). For instance, Blanton, Christie, and Dye (2000) found that the presence of negative gender stereotypes, unlike negative ethnic stereotypes, increased the tendency for women to contrast their performance with members of the in-group (i.e., other women). One explanation for this effect is that women seemed to endorse the negative gender stereotype when it was invoked. In this light, it would make sense that increasing the salience of the stereotype would increase contrast because the presence of real gender differences in ability would make gender a relevant attribute for including a comparison other in the standard (Goethals & Darley, 1977). This finding bears resemblance to Biernat and Manis' (1994; also Biernat, Manis, & Nelson, 1991) finding that perceivers evaluate people using different standards based on their stereotypical expectations for them. Thus, when judging the competencies of women on stereotyped dimensions (Biernat & Manis, 1994, Study 1), both male and female perceivers spontaneously used standards that assessed how well the women had done as women. It seems reasonable then that women in Blanton, et al. (2000) would evaluate their own abilities by determining their standing relative another woman to a greater degree when (believed) gender stereotypes were activated (see also Jost & Banaji, 1994; Major, 1994). A second reason that threat to the group may fail to cause assimilation with gender is that gender, unlike ethnicity, is not psychologically distinct. Recall that people typically think of themselves spontaneously in terms of particular attributes when these attributes are distinct (e.g., McGuire & Padawer-Singer, 1978). In this vein, they also think of themselves spontaneously in terms of shared in-groups when these groups are distinct as opposed to common (Brewer, 1991; Levine & Moreland, 1987; Mullen, 1991; Simon, 1992). Because gender is not distinct in most social contexts, contact with someone of the same gender should not lead to identification with that person as the member of a shared group (cf. Taylor & Fiske, 1978). This principle was illustrated in a set of studies by Brewer and Weber (1994). They told participants that a group they shared with a comparison other was either relatively common (80% prevalence) or relatively distinct (20% prevalence). Then, much like in the “Mr. Clean/Mr. Dirty” study (Morse & Gergen, 1970), they exposed participants to either upward or downward comparison information regarding performance on a job interview. As predicted, social comparison resulted in assimilation effects when comparison others shared a distinct in-group, but it resulted in contrast when the comparison other shared a common in-group.5 In summary, group threat may focus attention on group evaluation, leading to in-group assimilation effects. However, this is not predicted when the group threat is perceived as legitimate or when the group is common. Evaluation Indifference. The study by Brewer and Weber (1994) suggested that group threat does not lead to ingroup assimilation effects if the group is not distinct. This study makes another important point as well: Group distinctiveness can be sufficient for creating in-group assimilation effects even in the absence of a group threat . The fact that participants identified with another person simply on the basis of a minimal group assignment suggests that distinctiveness alone is enough to create a curiosity about group status. A more dramatic demonstration of this principle can be found in a set of studies by Brown and colleagues (Brown, Novick, Lord, & Richards, 1992). They found assimilation effects as a result of a social comparison with someone who simply shared the same birthday. This shared category is distinct but all the more meaningless. What is surprising about studies by Brewer and Weber (1994) and Brown et al. (1992) is that it is hard to imagine why participants would form a collective identity on the basis of meaningless groups. However, results are consistent with other studies suggesting that assimilation occurs when people are not motivated to self-evaluate. Recall the findings by Tesser (1988), which indicated that assimilation occurs when people make comparisons

< previous page

page_85

next page >

< previous page

page_86

next page >

Page 86 with close others, provided the evaluative dimension is not self-relevant. When people do not care about their standing on an evaluative dimension, they presumably do not care how they stack up relative to others with whom they share a close bond. As a result, they cease evaluating their abilities and shift to thinking of themselves in terms of their associations with close others. Integrating Tesser's (1988) findings with those from studies using meaningless groups (Brewer & Weber, 1994; Brown et al., 1992), assimilation seems to occur when people do not care about self-evaluation – either because the group or the evaluative dimension is not self-relevant. Recently, Pelham and Wachsmuth (1995) proposed a model that accounts for this emerging pattern of results. Based on arguments by Martin, Seta, and Crelia (1990) they suggest that assimilation is the default processing strategy whenever people are not motivated to engage in explicit and deliberate self-evaluations, whereas contrast is expected when people are so motivated. This proposal is consistent with Schwarz and Bless (1992a), who argued that the default response to contextual information is to include it in the representation of the target stimulus (see also Higgins, 1996; cf. Gilbert et al., 1995; Stapel & Koomen, chap. 14, this volume). Thus, when the comparison other is categorized as similar (“close’’), assimilation should occur in contexts that do not elicit interest in self-evaluation, but contrast effects should occur in situations that do. In support of their model, Pelham and Wachsmuth found assimilation effects when perceivers were certain of their abilities (and thus not motivated to evaluate them), provided the comparison other was close. However, low certainty and high closeness lead to contrast presumably on account of the desire to self-evaluate. This framework would account for the findings in Brewer and Weber (1994) and Brown (Brown et al., 1995) if one assumes that meaningless groups do not evoke self-evaluative concerns, but do evoke feelings of similarity. In summary, similarity on surrounding dimensions is predicted to cause assimilation effects under conditions that focus attention on the collective self. Although many factors may be relevant, group threat and group distinctiveness are two important determinants. Group threat can increase the concerns for the status of the group when it is perceived as illegitimate, thus causing assimilation effects in response to comparison with in-group members. Distinctiveness may be necessary for these effects to occur, but it can also cause assimilation in the absence of threats to status. Assimilation seems to occur in the absence of threat because distinctiveness can create a basis for shared categorization, but it typically does not evoke interest in deliberate self-evaluation. Thus, assimilation is predicted in two very different circumstances: when the group is meaningful and its status is under question, and when the group is incidental and its status is meaningless. Contrast is predicted with common but meaningful groups under conditions in which interest in self-evaluation is high. FUTURE DIRECTIONS The current framework applies basic principles of categorization and comparison, as outlined by Schwarz and Bless (1992a), to predict the effects of comparison on self-evaluation. Although many factors that can influence comparison processes were discussed, many more could not be addressed. One serious limitation to the current framework is that it focused exclusively on evaluations of the self in terms of valenced traits and abilities. It is almost certain that effects would be different if the same principles of categorization were applied to attitude evaluation. The most serious challenge to the current framework, however, is the limited consideration of contextual factors that alter the type of self-concept that is active in working memory during comparison. Although a few mechanisms were discussed (e.g., the meta-contrast principle, group threat), most of the discussion treated the type of selfrepresentation activated as something determined prior to self-evaluation. Predictions had an if-then logic to them: “If the self is represented in terms of individual attributes, then relevant comparison will be included in the standard, etc.” This is admittedly simplistic. A perceiver does not have to focus attention first on the individual self to feel diminished by a superior job candidate. Similarly, one does not have to go looking for an inspiring possible-self to find it in a role model. Based on norm theory (Kahneman & Miller, 1986), it seems more likely that comparison others tell the perceiver simultaneously what self and what standard are appropriate for self-evaluation. In the words of Kahneman and Miller, the appropriate representations are determined “on the fly,” based on both the features of the comparison other and the nature of the judgment context. Explaining how this occurs is one of the more interesting challenges to face social comparison theory in the future. ACKNOWLEDGMENT I would like to thank Bram Buunk, Tonya Dodge, Frederick X. Gibbons, Jim Jaccard, Nawaf Madi, Julie Osland, Anne Stuart, Ladd Wheeler, and Jan Fekke Ybema for comments and assistance with earlier drafts. Thanks especially to Norbert Schwarz for providing an

< previous page

page_86

next page >

< previous page

page_87

next page >

Page 87 expert opinion on Schwarz and Bless (1992a). ENDNOTES 1. It should also be mentioned that the impact of a new piece of information is affected by the amount of information already in the category. Inclusion of one new piece of information will have less influence on evaluations the more information one has represented in the category. 2. There are situations in which representation of the self in terms of its possibility can lead to contrast. This will occur when the comparison others are used as standards against which to evaluate possible selves. This was demonstrated by Lockwood and Kunda (1999). They exposed participants to upward role models after first having them reflect on their future ideal selves. Contrast effects were observed presumably because the upward role models were superior to participants' ideal selves. In short, the upward comparison others showed participants that their greatest hopes for the self fell short of what others had already achieved. 3. Thanks to Dale T. Miller for this example. 4. In this chapter, the term African-American is used to refer to the perceiver because stereotype threat is conceived as the result of acculturation to American stereotypes about Blacks. However, the term Black is used when referring to the comparison other because the perceiver only knew comparison racial appearance. 5. Some subjects were exposed to comparison others who were members of the out-group. It was predicted that out-group standards would not be relevant for comparison when the in-group was common but would be relevant when the in-group was distinct. Results were consistent with these predictions (cf. Major et al., 1991).

< previous page

page_87

next page >

< previous page

page_88

next page >

page_88

next page >

Page 88 This page intentionally left blank.

< previous page

< previous page

page_89

next page >

Page 89 6 Outgroup Favoritism and the Theory of System Justification: A Paradigm for Investigating the Effects of Socioeconomic Success on Stereotype Content John T. Jost Stanford University And in the moment you are born since you don't know any better, every stick and stone and every face is white, and since you have not yet seen a mirror, you suppose that you are, too. It comes as a great shock around the age of 5 or 6 or 7 to discover that the flag to which you have pledged allegiance, along with everybody else, has not pledged allegiance to you. It comes as a great surprise that Gary Cooper killing off the Indians, when you were rooting for Gary Cooper … that the Indians were you.(Interview with James Baldwin, The Price of the Ticket ) It is not only the discipline of psychology that has undergone a cognitive revolution in the latter half of the 20th century. Social science in general has shifted toward an increased reliance on mental states and processes in explaining the behaviors of individuals and groups. Sociologists and political scientists, for example, have focused more and more on expectations, attitudes, beliefs, perceptions, decisions, and judgments in explaining such diverse phenomena as political socialization, group dynamics, voting behavior, and other responses to structures of status, power, and prestige (e.g., Berger & Zelditch, 1998; Howard, 1994; Iyengar & McGuire, 1993; Ridgeway, 2001). Anthropologists, too, have moved increasingly toward beliefs, construals, interpretations, and other intentional states in their descriptions of cultural systems and practices (e.g., Geertz, 1983; Shweder & LeVine, 1984; Sperber, 1990). Even philosophers have adopted the language of cognitive science to the point where traditional metaphysical and epistemological approaches have almost disappeared (e.g., Goldman, 1988; Kornblith, 1994; Solomon, 1992). In the field of organizational behavior, social-cognitive constructs such as attributions, accounts, scripts, and justifications have been used to shed light on such applied topics as job satisfaction, division of labor, employee relations, task design, and corporate strategy (e.g., Baron & Pfeifer, 1994; Martin, 1982; Salancik & Pfeifer, 1978; Weick, 1993). Neo-institutionalist theories of organizations have further analyzed the ways in which ideas and symbols are used to structure and legitimate business cultures and spread influence (e.g., Powell & DiMaggio, 1990). Somewhat improbably given the subject matter, cognitive theories of social movements and revolutions have become paradigmatic (e.g., Eyerman & Jamison, 1991; Moore, 1978; Snow & Oliver, 1995), even among Marxist scholars, who are traditionally among the least individualistic of social theorists (e.g., Elster, 1985). These intellectual developments, spread as they are across a variety of disciplines, mean that constructs such as attitudes, thoughts, and beliefs have proved useful indeed for explaining behavior that clearly falls outside of the original domain of cognitive psychology – in this case, behavior that is collective, coordinated, and downright political. More precisely, one might say that in the post-cognitive revolution world, sociologists, psychologists, anthropologists, political scientists, and organizational theorists all accept the fundamental assumption that social systems are maintained at least in part through attitudes and beliefs that support them. In the language of social cognition, researchers would say that conscious and unconscious thought processes play a pivotal role in the acceptance or rejection of particular social and political forms (Jost, 1995). One variable in particular, the appraisal of legitimacy , has emerged as an important social-psychological predictor of responses to inequality (e.g., Major, 1994; J. Martin, 1993; Tajfel & Turner, 1986; Tyler, 1997). Historical and ethnographic studies make abundantly clear that, in the absence of militant or totalitarian rule, authorities, procedures, and social arrangements are stable and enduring to the extent that they are perceived as having legitimacy (e.g., Gurr, 1970; Moore, 1978). Inequality among groups and individuals is accepted and perpetuated, even by those who stand to lose the most from it, so long as it is perceived as fair and legitimate. This is one of the starting points of the theory

< previous page

page_89

next page >

< previous page

page_90

next page >

Page 90 of system justification (e.g., Jost & Banaji, 1994), which is intended as a social-cognitive theory of intergroup relations and political behavior, and it is also one of the central contentions of this chapter. THE JUSTIFICATION PRINCIPLE IN SOCIAL PSYCHOLOGY In their ground-breaking work on The Social Construction of Reality , Berger and Luckmann (1967) observed that “the institutional world requires legitimation, that is, ways by which it can be ’explained’ and ’justified’” (p. 61). In other words, we do not support social structures unless they satisfy our cognitive needs for validity and rationality. Sociologists and organizational theorists have described the excuses, accounts, and explanations we use to smooth over social interaction (e.g., Scott & Lyman, 1968), spread organizational influence (e.g., Powell & DiMaggio, 1990), and preserve public reputations (e.g., Elsbach & Sutton, 1992). However, it is the discipline of psychology that reminds us of the fact that people must generate reasons and justifications privately, for themselves, as well as for others (e.g., Aronson, 1992; Festinger, 1957; Tetlock & Manstead, 1985; Weick, 1993). Psychologists know better than anyone else that we are cognitive creatures; we need reasons and arguments to justify both action and inaction. If we behave in an inconsistent or counterattitudinal way, we must come up with rationalizations for the departure. If we do something that causes social disapproval, we must defend ourselves with ideas. And sometimes, we just want to know that there are reasons and explanations for the actions of others and the way things are. These principles are assumed by such diverse theories and research programs in social psychology as balance theory, cognitive dissonance theory, equity theory, attribution theory, and social cognition. In other words, one of the central contentions of social psychology from the 1950s to the present has been that people justify themselves, their associates, and the world around them in the sense that they use ideas to provide validation and legitimacy for all of these things (Jost & Banaji, 1994). Justifications differ from explanations, in that the former (but not necessarily the latter) render social events right and appropriate. It is no accident that justice and justification derive from the same Latin root. The existence of class systems may be justified – that is, made legitimate – by postulating wide individual differences in effort or motivation; patriarchal systems may be justified by asserting insurmountable gender differences in achievement or ability; and systems of ethnic or racial segregation may be justified by claiming incommensurable group differences in intelligence or morality. In this way, people use ideas about groups and individuals to reinforce existing social systems and preserve the sense that those systems are fair, legitimate, and justifiable (Jost & Banaji, 1994). This need not be a conscious process, as Bem and Bem (1970) pointed out in their analysis of gender socialization and the subtle, non-conscious ways in which girls are taught to “know their place” and to participate in a sexist world. Yet why would people perceive the world around them to be justified and legitimate when so many features of their environment seem unfair and illegitimate? One answer that social psychologists have given is that there is a general motivation to “believe in a just world” (e.g., Lerner, 1980; Olson & Hafer, 2001). The guiding thesis is that living in circumstances that are unpredictable, uncontrollable, and capriciously unjust would be too psychologically threatening, and so we cling to the illusion that people get what they deserve and deserve what they get. The theory of system justification builds on this essential insight (Jost & Banaji, 1994), deemphasizing somewhat the universal, psychodynamic aspects of the process and stressing instead the impact of social learning, institutionalized norms, and the power of ideology (cf. Bem & Bem, 1970; Berger & Luckmann, 1967; Major, 1994; Tyler & McGraw, 1986). Stereotypes as Ideological Justifications One of the most common ways in which people use ideas and beliefs to justify the social world around them is by stereotyping members of disadvantaged groups in ways that rationalize the inequality. The stereotype that African Americans (or Hispanic Americans or blue-collar workers) are not as intelligent or hard working as other, more successful groups, according to this perspective, serves as an ideological justification for the substantial socioeconomic differences between these groups and others. This justification or rationalization function was recognized by Katz and Braly (1933) and Allport (1954), but it was not studied directly until relatively recently, when cognitive and ideological analyses were again combined. In sociology, Jackman and Senter (1983) provided survey data leading to the conclusion that ideological values and stereotypes were relatively consensual and favoring of dominant groups′ interests across racial and gender contexts. C. Hoffman and Hurst (1990) conducted experimental research in which people formed stereotypes of fictional groups as a way of rationalizing unequal divisions of labor and social roles; the authors explicitly called into question the kernel of truth view of

< previous page

page_90

next page >

< previous page

page_91

next page >

Page 91 stereotypes, arguing that stereotype contents reflect rationalization rather than perceptual processes. Ridgeway (2001), too, has summarized field and experimental studies, leading to the conclusion that people form status beliefs in the course of face to face interaction in such a way that structural inequalities tend to be legitimized and perpetuated. In all of these research programs, we see that stereotypes operate as ideological devices to justify or rationalize inequality between groups. The most surprising and powerful cases of system-justifying stereotypes arise when members of low-status groups internalize unfavorable stereotypes of themselves and favorable stereotypes of others as a way to justify the existing hierarchy (Jost & Banaji, 1994). This process may give rise to the attitudes and beliefs that are out-group favoring. Such ideas and justifications are taught to us as children until they begin to operate non-consciously (Bem & Bem, 1970). Ultimately, system-justifying ideologies and stereotypes become imperceptible – like water to the fish. Outgroup-Favoring Stereotypes Under Extreme and Ordinary Circumstances From a psychological standpoint, it is especially striking that disadvantaged people would justify the very social system that places them at a disadvantage. Nevertheless, it seems that people sometimes do favor the preservation of the social order even over their own personal and collective interests, as the opening quote from James Baldwin attests. Familiar examples in the domain of public opinion include women′s widespread failure to support the Equal Rights′ Amendment (Gurin, 1985; Mansbridge, 1986) and the significant lack of support among the American working class for policies of economic redistribution (Kluegel & Smith, 1986; Piven & Cloward, 1977). Even more dramatic examples are found in historical and journalistic accounts of war camps and slavery. Psychoanalytically inspired work by Anna Freud, Bruno Bettelheim, and others on the phenomenon of ‘‘identification with the aggressor” suggested that, among victims of extreme injustice and deprivation, there was an implicit rejection of the ingroup and a preference for the outgroup. A historical novel set during the Holocaust provides a stirring literary example in which a young Gypsy boy encounters his first Nazi officer. In The Painted Bird, Jerzy Kosinski (1966) writes: His entire person seemed to have something utterly superhuman about it … he seemed an example of neat perfection that could not be sullied: the smooth, polished skin of his face, the bright golden hair showing under his peaked cap, his pure metal eyes … I thought how good it would be to have such a gleaming and hairless skull instead of my Gypsy face which was so feared and disliked by decent people (p. 100). Similarly disturbing cases of system justification exist in contemporary accounts of the ongoing slave trade in Africa and Latin America. A 1992 article published in Newsweek magazine (Masland, Nordland, Liu, & Contreras, 1992), for example, quoted a 25-year-old Mauritanian slave as follows: I am a slave, my whole family are slaves … Sometimes they treat us well, sometimes they treat us badly, but only the children get beaten … A master is a master and a slave is a slave. Masters are white, slaves are black … Naturally, we blacks should be the slaves of the whites (p. 32). These anecdotal examples are of rare and extraordinary circumstances. They are relayed to suggest that psychological investment in the status quo may occur even under the most horrific systems of inequality and exploitation. If some degree of system justification arises under such dramatic conditions, system-justifying impulses should be even more likely to arise in ordinary life. It was a widespread assumption, in fact, of early researchers of intergroup relations that members of disadvantaged groups such as Jews and Blacks could not help but internalize society′s biases against them and exhibit a kind of inferiority complex at the group level (e.g., Allport, 1954; Bettelheim, 1960; Lewin, 1941). This was also the conclusion reached by Clark and Clark (1947) in their famous studies of African-American children′s preferences for White dolls. A series of laboratory studies conducted by Sachdev and Bourhis (1985, 1987, 1991) demonstrated convincingly that assignment to high-status or powerful groups leads people to display ingroup favoritism, whereas assignment to low-status or powerless groups produces outgroup favoritism. Because evidence of outgroup favoritism is often deemphasized in the literature on intergroup relations, the data from tables reported in an influential metaanalytic study conducted by Mullen, Brown, and Smith (1992) are rearranged and presented in Table 6.1 according to the percentage of experimental groups showing ingroup favoritism, outgroup favoritism, and exact equality – broken down according to relative status of the ingroup. What this reveals is that members of high-status and equal-status groups are indeed going overwhelmingly with ingroup favoritism, but 85% of the low-status groups are displaying outgroup favoritism. That is, they

< previous page

page_91

next page >

< previous page

page_92

next page >

Page 92 TABLE 6.1: Number and percentage of experimental groups showing ingroup favoritism, exact equality, and outgroup favoritism by relative status of the ingroup. StatusHighEqualLowIngroup favoritismNumber20273Percentage100%73 %15 %Exact equalityNumber020Percentage0%5%0%Outgroup favoritismNumber0817Percentage0%22%85%Note. Table adapted from a meta-analysis reported by Mullen, Brown, and Smith (1992). Ingroup favoritism indicates a positive Z-score reported by Mullen et al., exact equality indicates a Z-score of zero, outgroup favoritism indicates a negative Z-score. are saying that members of the other group are more intelligent, more industrious, and so on, than are members of their own group. This type of evidence led Jost and Banaji (1994) to argue that many low-status groups accept as legitimate their alleged inferiority. Rather than attempting to raise self-esteem or enhance ingroup solidarity, they use their evaluations and judgments to reinforce and justify the existing system of inequality. In some ways, this interpretation implies a rediscovery of the Marxian concept of false consciousness (Jost, 1995; Jost & Banaji, 1994). Contrary to some claims (e.g., Mullen et al., 1992), systematic evidence of outgroup favoritism does not seem to be restricted to laboratory groups. Field studies conducted by, among others, R. Brown (1978), Hewstone and Ward (1985), Jost, Burgess, and Mosso (2001), Mlicki and Ellemers (1996), Skevington (1981), and Spears and Manstead (1989) have turned up strong evidence of outgroup favoritism among members of various low-status groups. In fact, evidence coming from many real-world groups supports Roger Brown's (1986) observation that: Subordinate groups like black Americans, South African Bantus, the Mayans of Guatemala, and the lower castes of India either do, or until recently did, derogate or look down on the in-group and show positive attitudes toward the depriving out-group (p. 558). It has often been suggested that the prevailing theory of intergroup relations, social identity theory, is not wellequipped to handle the phenomenon of outgroup favoritism (e.g., Hewstone & Ward, 1985; Hinkle & Brown, 1990; Jost & Banaji, 1994; Sidanius, 1993), although it is true that the subject has been addressed in some detail by social identity theorists such as Turner and Brown (1978), Tajfel and Turner (1986), Spears and Manstead (1989), and others. In many ways, system justification theory seeks to build on the foundation laid by social identity theory, much as social identity theorists sought to build on theories of social comparison and realistic conflict (Tajfel & Turner, 1986). Outgroup Favoritism and Social Identity Theory Social identity theory was developed to account for the initially unexpected finding that minimal laboratory groups with no history of interaction displayed ingroup favoritism with regard to social stereotyping, performance evaluation, and resource allocation (Tajfel & Turner, 1986). Drawing extensively on Festinger's social comparison theory, it was argued that because people need to evaluate themselves favorably and because group membership is an important constituent of the self-concept, people tend to evaluate their ingroups more favorably than they evaluate other groups (e.g., Tajfel & Turner, 1986; Turner, 1975). Thus, according to social identity theory, there is a general drive to enhance individual and collective self-esteem by making favorable comparisons between the ingroup and relevant outgroups (e.g., Hogg & Abrams, 1988; Tajfel & Turner, 1986). Although it is true that social identity theory emphasizes the generalizability of ingroup favoritism among members of many different types of social groups (Hogg & Abrams, 1988; Mullen et al., 1992; Tajfel & Turner, 1986), it has also done much to frame the social-psychological understanding of how and when groups that are low in social standing will accept their alleged inferiority and when they will attempt to challenge it (e.g., Ellemers, Wilke, & van Knippenberg, 1993; Mummendey & Schreiber, 1984; Spears & Manstead, 1989; Turner & Brown, 1978; van Knippenberg, 1978; Wright, Taylor, & Moghaddam, 1990). From this body of literature, it is possible to discern three distinct explanations for the phenomenon of outgroup favoritism among low-status groups. One account has to do with self-categorization processes of identification and dis-identification (Tajfel & Turner, 1986), another account distinguishes between comparative dimensions that are relevant and irrelevant to the status differences (van Knippenberg, 1978), and a third account is related to perceptions of the legitimacy

< previous page

page_92

next page >

< previous page

page_93

next page >

Page 93 and stability of status differences (Turner & Brown, 1978). According to the first type of explanation, members of low-status groups exhibit outgroup favoritism to the extent that they shun identification with their own negatively valued group and identify instead with members of a positively valued outgroup (Tajfel & Turner, 1986). It has been argued that disidentification is the preferred choice among low-status group members in general (e.g., Ellemers, van Knippenberg, de Vries, & Wilke, 1988; Lewin, 1941; Tajfel, 1978), and research suggests that when the option of individual mobility or exit is available, low-status group members tend to take it (Lalonde & Silverman, 1994; Wright et al., 1990). Research has also demonstrated that people who are made to identify only weakly with a low-status ingroup are less committed to the group and more likely to express a desire for individual mobility to another group than are people who are made to identify strongly with the ingroup (Ellemers, Spears, & Doosje, 1997). Although it seems plausible that levels of ingroup identification would predict the direction and magnitude of ingroup versus outgroup favoritism among members of low-status groups, there are reasons to think that ideological factors such as the perception of system legitimacy play a more determining role. For instance, group consciousness-raising among women and minority groups requires not merely an identification with one's group, but a perception that the group's low status is illegitimate and unfair (e.g., Gurin, 1985; Kluegel & Smith, 1986). Thus, from our perspective, group identification is probably a necessary but not sufficient condition for the rejection of outgroup favoritism. According to the second type of explanation, members of low-status groups may accept their inferiority and engage in outgroup favoritism on dimensions that are highly relevant to the status differences, but they may exhibit ingroup favoritism on irrelevant dimensions as a way to compensate for an otherwise negative social identity (e.g., van Knippenberg, 1978). In fact, there is a wealth of evidence to support the notion that members of low-status groups accept their inferiority and exhibit outgroup favoritism on dimensions that are highly relevant to the status differences, but they exhibit ingroup favoritism on irrelevant dimensions as a way to compensate for an otherwise negative social identity (e.g., Mullen et al., 1992; Mummendey & Schreiber, 1984; van Knippenberg, 1978). For instance, Skevington (1981) examined intergroup relations among professional nursing groups that differed in status and found that low-status group members judged the other group to be more intelligent, ambitious, responsible, organized, and confident than their own group, but they saw themselves as more cheerful, thoughtful, happy, and practical than the outgroup. The strategy of compensating for the effects of a low-status position by displaying strong ingroup favoritism on dimensions that are unrelated to the status difference has been referred to as social creativity (Tajfel & Turner, 1986). A third account – one that is most closely related to the concerns of system justification theory – has to do with perceptions of the legitimacy and stability of the social system (e.g., Tajfel & Turner, 1986). Specifically, it has been found that members of a low-status group accept their inferiority on dimensions related to the status differences and display outgroup favoritism unless they perceive the status differences to be both illegitimate and likely to change (Turner & Brown, 1978). Thus, whether low-status group members accept or reject their alleged inferiority is hypothesized to depend on whether they perceive “cognitive alternatives” to the social system, which are said to be brought on by appraisals of illegitimacy and instability. Although there is no published research to date linking perceptions of legitimacy and stability to counterfactual thinking with regard to social systems, research in social identity theory has highlighted these variables as important predictors of group identification and intergroup behavior (e.g., Caddick, 1982; Ellemers et al., 1993; Turner & Brown, 1978). The Influence of Social Identity Theory on the System Justification Perspective Social identity theory is an important precursor to the theory of system justification in at least three ways. First, it brings a social-psychological perspective to bear on intergroup relations. Differences in status or success between groups are predicted to affect group members′ perceptions of their own group and other groups, and these perceptions are theorized to affect the future course of relations between the groups as well as the viability of the social system (e.g., Tajfel & Turner, 1986; Wright et al., 1990). Second, the theory introduces ideological factors such as perceptions of the legitimacy and stability of the status system as relevant to ingroup and outgroup favoritism (e.g., Ellemers et al., 1993; Tajfel & Turner, 1986; Turner & Brown, 1978). Third, the notion that social groups invent ideologies that justify their competition against other groups, when combined with perspectives emphasizing the persuasive power of dominant groups′ ideologies, helps explain why stereotypes and other ideas justifying social and material inequalities eventually come to be endorsed even by members of subordinate groups (e.g., Hogg & Abrams, 1988; Jost &

< previous page

page_93

next page >

< previous page

page_94

next page >

Page 94 Banaji, 1994). Nevertheless, there are some limitations of the social identity perspective that make it more of a jumping off point than a terminus for the theory of system justification (cf. Jost & Banaji, 1994). Most important, theorizing about when low-status groups accept the status quo and when they reject it has tended to be relatively undeveloped in the social identity tradition. Turner and Brown (1978) argued that “subordinate groups will seek positive distinctiveness from dominant groups to the degree that their inferiority is not perceived as inherent, immutable or fully legitimate” (p. 207), but the theory fails to specify when the system is perceived as “inherent, immutable, or fully legitimate” and when it is not. Perceptions of legitimacy and stability have been addressed by social identity researchers, but they have entered into the theory as independent variables (e.g., Caddick, 1982; Ellemers et al., 1993; Turner & Brown, 1978). As a result, not much is known from social identity theory about the causes of perceived legitimacy or about why members of low status groups would ever find the system to be legitimate, when such a perception clearly conflicts with group-serving motivations. The guiding assumption of social identity theory is that people are motivated to favor their own group over other groups, but this motivation seems conspicuously lacking in any display of outgroup favoritism, even outgroup favoritism under conditions of legitimacy and stability (Hinkle & Brown, 1990). System justification theory, by contrast, draws on the vast literature on the tolerance of injustice (see Jost, 1995) and posits a motive to invest in and rationalize the status quo, and this motive is thought to be present even among members of disadvantaged groups, although typically to a lesser degree. Furthermore, research on system justification theory has begun to treat legitimacy and stability as moderating and dependent variables, demonstrating, for instance, that the act of stereotyping increases perceptions of the magnitude, legitimacy, and stability of status differences (see Jost, Burgess, & Mosso, in press). Thus, system justification theory seeks to elaborate further some of the socio-structural variables identified by social identity theorists. In fact, Tajfel (1984) seemed to realize some of the limitations of social identity theory when he wrote that: This disymmetry between the ‘superior’ and ‘inferior’ groups has been recognized to some extent in the social identity approach to intergroup relations which has specified the different strategies for achieving distinctiveness that can be adopted by members of groups which differ in status … But this is not enough (p. 700). Elsewhere, he noted the importance of justice perceptions in particular for an understanding of when group members will accept and when they will reject the social system: [an] important requirement of research on social justice would consist of establishing in detail the links between social myths and the general acceptance of injustice, and research which would attempt to specify the sociopsychological conditions which could be expected to contribute to the dissolution of these patterns of acceptance (1982, p. 164). Here, Tajfel seemed to be appealing to justice researchers to determine when people will engage in system justification and when they will not. He alluded, it seems, to the need for a theory of false consciousness (Jost, 1995). Obviously, these are top priorities of the system justification approach, the implications of which are still being developed and tested in emerging research paradigms. After a brief overview of the theory, we describe an experimental paradigm that has been used to shed further light on the dynamics of ingroup and outgroup favoritism and the role of several variables identified by social identity theory, including group identification, attribute relevance, and perceived legitimacy. THE THEORY OF SYSTEM JUSTIFICATION Although phenomena such as outgroup favoritism and internalization of inferiority may be puzzling to social, political, and psychological theorists, who assume that attitudes and behaviors are driven largely by self-interest, groupinterest, or needs for personal or collective self-esteem, there is a rich tradition of Marxist and feminist scholarship on the problem of false consciousness (see Jost, 1995, for a review). This work especially emphasizes the cognitive dimensions of oppression and system preservation, building on Marx and Engels' (1846/1970) observation that, “The class which has the means of material production at its disposal, has control at the same time over the means of mental production ” (p. 64; emphasis added). This theme is carried on in 20th century Marxism, most especially in György Lukács′ historical analysis of class consciousness and Antonio Gramsci′s cultural theory of hegemony and consent. Contemporary sociologists working under the banner of dominant ideology theory have continued to explore the extent to which subordinate groups are persuaded to hold beliefs that are at odds with their objective social interests (Abercrombie, Hill, & Turner, 1990; Kluegel & Smith, 1986).

< previous page

page_94

next page >

< previous page

page_95

next page >

Page 95 System justification theory, then, is a theory of social cognition that takes its impetus from the Marxian ideological tradition, with its focus on the justifcation of inequality and exploitation (Jost & Banaji, 1994). One of the main theoretical assumptions of this perspective is that, all other things being equal, people tend to use ideas about groups and individuals to justify the way things are, so that existing social arrangements are perceived as fair and legitimate, perhaps even natural and inevitable. There are at least seven established social-psychological phenomena that we draw on and take to support the general system justification perspective. They may be summarized as follows: 1. Members of groups low in social standing exhibit ‘‘outgroup favoritism” by internalizing unfavorable stereotypes of their own group and subscribing to favorable stereotypes of successful outgroups (Hinkle & Brown, 1990; Jost & Banaji, 1994; Sidanius, 1993). 2. People form stereotypes as a way of “rationalizing” unequal divisions of roles, especially in terms of essentialistic biological categories (Eagly & Steffen, 1984; Hoffman & Hurst, 1990; Jackman & Senter, 1983). 3. Members of disadvantaged groups tend to draw intrapersonal and intragroup social comparisons rather than intergroup comparisons, as when women judge the legitimacy of their own income against standards of the income of other women and of their own income in the past (Jost, 1997; Major, 1994). 4. People perceive existing institutions, procedures, and outcomes as fair and legitimate, even when there are reasons to suspect that they are not (Lerner, 1980; Martin, 1986; Tyler & McGraw, 1986). 5. People exhibit decision-making biases in favor of whatever option is perceived as the “status quo” and avoid choices that are perceived to entail change (Samuelson & Zeckhauser, 1988; Silver & Mitchell, 1990). 6. People stick disproportionately with past behavioral practices simply because they are familiar or habitual and fail to consider innovative alternatives (Hackman & Oldham, 1980; Silver & Mitchell, 1990). 7. People display “outcome biases” in their evaluations of groups and individuals, so that people described as “winners” are selectively perceived as possessing enduring attributes that are consistent with their success and people described as “losers” are seen as always having possessed attributes that are consistent with their failure (Allison, Mackie, & Messick, 1996). What these distinct bodies of evidence have in common is the notion that what is tends to be experienced as what ought to be; Although some of these phenomena, most especially status quo and outcome biases, are usually explained in purely cognitive terms by social psychologists, there is an ideological tenor to them that adds a layer of political significance and motivation to the basic information processing functions. According to the present view, cognition is deployed in the service of the social system. An Experimental Paradigm On occasion, the phenomenon of outgroup favoritism has been dismissed as something of an experimental artifact that does not occur in real-world groups. For instance, Mullen et al. (1992) wrote that “a concentration on transitory, task-specific conceptualizations of status would lead to the misguided conclusion that ingroup bias occurs predominantly in higher status groups” (p. 118). One of the goals of the research paradigm summarized here is to examine ingroup and outgroup favoritism by using an experimental manipulation of status that is neither transitory nor task-specific. Instead, we sought to devise an experimental paradigm in which ingroup and outgroup favoritism could be investigated in the context of real-world group memberships, whereas relative social status could be manipulated experimentally so that differences due to social status of the ingroup could be attributed solely to variations in status and not to other factors associated with particular real-world groups. Additionally, the manipulation of status was defined in terms of socioeconomic success. This was done to improve on the ecological validity of previous experimental manipulations of status in terms of performance feedback on tests of creativity (Sachdev & Bourhis, 1987, 1991), reasoning skills (Turner & Brown, 1978), or other “transitory, taskspecific conceptualizations” (Mullen et al., 1992, p. 118). Because system justification theory is especially relevant for understanding cognitive responses to wealth and poverty (e.g., Jost, 1995; Lane, 1962; Major, 1994), status of the ingroup was operationalized as relative socioeconomic success. The basic procedure for varying perceived socioeconomic success is as follows. Shortly after arriving for an experiment billed as “The Inter-Collegiate Study of Abstract Thought,” university students are told that they are about to participate in a research project aimed at understanding “why differences in social and economic success exist between graduates of different colleges and universities.” Half of the participants are presented with statistics indicating that alumni members of their own university group are significantly less successful in terms

< previous page

page_95

next page >

< previous page

page_96

next page >

Page 96 of socioeconomic achievement than are members of a rival outgroup, and the other half are led to believe that the ingroup is more successful than the outgroup. These statistics include information concerning average financial income, career advancement and promotions, status of professions entered, rates of admission to graduate and professional schools, and years of postgraduate education completed. The materials used for a study involving University of Maryland students (low-success condition) are presented in Table 6.2 (see also Jost & Burgess, 2000, for more information). This procedure has been used successfully in a series of experiments involving students at Yale University, the University of Maryland, and the University of California at Santa Barbara (U.C.S.B.), with comparison outgroups of Stanford University, the University of Virginia, and the University of California at Los Angeles (U.C.L.A.), respectively. These three experiments are described in abbreviated form in this chapter; two of them are presented more systematically elsewhere (see Jost, 1996; Jost & Burgess, 2000). Manipulation checks indicate that, in all studies conducted thus far, the statistics concerning socioeconomic success differences (ascribed to sources such as U.S. News and World Report and the Chronicle of Higher Education ) were accepted as credible and convincing. These studies also demonstrate that relative socioeconomic success has a major impact on the stereotypes and evaluations that people have about ingroup and outgroup members. As a general rule, random assignment to the high-success condition leads people to display ingroup favoritism – that is, to express beliefs that their own group is superior on a number of stereotypical characteristics, whereas assignment to the low-success condition leads people to display outgroup favoritism – that is, to express beliefs that the more successful outgroup is superior. This pattern has been observed on qualitative, open-ended measures of attribution in which respondents are not constrained by the expectancies or questions of the researchers. In the Yale study, following the experimental induction of relative success, participants were given the following instructions: “Think about the differences in social and economic success between Yale and Stanford alumni/ae. Can you think of any explanations or justifications for why Yale and Stanford graduates would have different rates of socioeconomic success?” Participants were then asked to spend 2 to 3 minutes listing up to five responses in clearly numbered spaces. Two independent judges coded the open-ended responses as focusing either on the ingroup (Yale) or the outgroup (Stanford) and as expressing something favorable, unfavorable, or neutral about that ingroup or outgroup. TABLE 6.2: Sample materials for manipulating perceived socioeconomic success. Virginia AlumniMaryland AlumniMean Financial Incomeafter 5 years$38,500$24,700after 10 years$53,200$39,500after 20 years$69,700$54,100at retirement$78,300$62,500Career AdvancementMean number of promotions after 5 years2.41.3Mean number of promotions after 10 years5.33.0Mean number of promotions after 20 years9.56.2Number of CEOs of major corporations4117postgraduate EducationMean years of postgraduate education1.30.4% of applicants admitted to medical school44%21%% of applicants admitted to law school43%19%% of applicants admitted to business school57%30%% of applicants admitted to graduate school60%41%% receiving post-baccalaureate degrees49%23%Sources: U.S. News & World Report, 1993; The Chronicle of Higher Education, 1994 When Yale students were assigned to the high-success condition, explanations making reference to characteristics of the ingroup tended to be very favorable (e.g., “Maybe Yalies are smarter”; ‘‘Yale admits students with better records who are innately more driven”). According to the independent judges, 81.5% of ingroup-related statements were favorable in content, and only 2.4% of these were unfavorable. When members of high-success groups generated explanations pertaining to the outgroup, 42.1% of these were judged to be unfavorable (e.g., “Because Stanford is a sport scholarship-granting school, they are going to get athletes that are not as intelligent as the students who get in regularly”; “Stanford students are somehow superficial. They simply glide

< previous page

page_96

next page >

< previous page

page_97

next page >

Page 97 on the surface without seeking the deep reasons”). Only 15.8% of the explanations having to do with features of the outgroup were favorable. When Yale students were assigned to a position of low socioeconomic success, the results were very different. Under these conditions, only 12.3% of the explanations involving the ingroup were judged to be favorable in nature, whereas 42.5% were unfavorable (e.g., “Yale is full of your bookworms and your dorks”; “Yale students are too idealistic, and usually have impractical or false imaginations about real world life”). Of explanations involving the outgroup, 62.2% were judged as favorable (e.g., “Stanford offers a better education than Yale”; ‘‘Stanford is a more selective school, so it has smarter people”), and only 4.2% were unfavorable. Thus, members of low-success groups display outgroup favoritism in making open-ended attributions for the socioeconomic success differences. As with stereotypes, evaluations, resource allocations, and other types of social judgments, the research literature on intergroup relations has stressed the ethnocentric nature of attributions about ingroup and outgroup members (Cooper & Fazio, 1986; Hewstone, 1990; Pettigrew, 1979). The notion that people generate groupserving attributions for outcomes is also highly consistent with social identity theory. Our research suggests, however, that low-status group members do not attribute their inferior position-to situational factors or extenuating circumstances, but rather seem to internalize the inequality in the form of internal attributions about the unfavorable characteristics of the ingroup and the favorable characteristics of the outgroup. According to a system justification analysis, members of groups that are low in social or material standing should exhibit ingroup derogation and outgroup favoritism to the extent that they perceive the overarching social system to be fair, legitimate, and justifiable. Thus, it is hypothesized that perceived legitimacy is negatively related to ingroup favoritism among members of low-status groups, insofar as people who accept ideological justifications for a status quo that places them at a disadvantage should be more likely to consent to their own inferiority. However, perceived legitimacy is hypothesized to relate positively to ingroup favoritism among high-status groups, insofar as they gain confidence and esteem from the sense that their advantage is legitimized; their sense of superiority is increased by the perception that the system is fair, legitimate, and justified. This interaction hypothesis is slightly different from that which has been predicted by social identity theorists. Turner and Brown (1978) hypothesized that “[g]roups with illegitimate status relations would display more ingroup bias than those with legitimate status relations” (p. 210), regardless of the status of the ingroup. Their reasoning was that perceived illegitimacy should render the system of status differences unstable and insecure, leading both groups to vie for a position of superiority (see also Caddick, 1982). This implies a main effect hypothesis such that ingroup favoritism should be greater among people who perceive the status differences to be low in legitimacy than among people who perceive them to be high in legitimacy. Thus, although social identity and system justification theories make the same prediction with regard to the behavior of low-status group members, the two perspectives differ when it comes to predictions about high-status group members. In an experiment involving students at the University of Maryland, the procedure described earlier was used to manipulate perceived socioeconomic success. Following this induction, participants were asked how fair or unfair, how justifiable or unjustifiable, and how legitimate or illegitimate the socioeconomic success differences were between the ingroup (University of Maryland) and the outgroup (University of Virginia), and their responses to these three items were averaged to create a general index of perceived legitimacy. Ingroup and outgroup ratings on status relevant (intelligent, hard working, and skilled at verbal reasoning) and status irrelevant (friendly, honest, and interesting) attributes were solicited. As illustrated in Fig. 6.1, perceived legitimacy increased ingroup favoritism among members of high-success groups, but it decreased ingroup favoritism (and increased outgroup favoritism) among members of low-success groups. This interaction pattern was also observed in the Yale study (Jost, 1996), suggesting that the focus on legitimation in intergroup relations is well placed. Perceived legitimacy seems to have opposite effects on high-status and low-status group members, as system justification theory predicts, and not as Turner and Brown (1978) suggested. In addition, Jost and Burgess (2000) found that Maryland students assigned to the position of low socioeconomic success showed significantly greater attitudinal ambivalence directed at their own group than did students assigned to the high-success condition. This was explained in terms of a psychological conflict between opposing tendencies toward group justification and system justification – a conflict that faces members of low-status but not high-status groups (see also Jost et al., 2001). We also reasoned that members of a psychologically meaningful group such as this (for whom at least moderate levels of group justification motives would be present), ambivalence toward the ingroup

< previous page

page_97

next page >

< previous page

page_98

next page >

Page 98 Fig. 6.1. Means on ingroup favoritism (stereotyping) by ingroup success and perceived legitimacy

would be increased as levels of system justification were increased for members of the low-status group, but ambivalence would be decreased as levels of system justification were increased for people assigned to the highstatus condition. And this is what we found. Perceptions of the legitimacy of the status differences were associated with increased ambivalence among low-status group members and decreased ambivalence among high-status group members (Jost & Burgess, 2000). One potential limitation of the Maryland study is that perceived legitimacy was measured rather than manipulated, and so there was no random assignment to conditions of legitimacy or illegitimacy. A follow-up study involving students at U.C.S.B. did employ an experimental manipulation of perceived legitimacy. After learning that their own group was less socioeconomically successful than the comparison outgroup of U.C.L.A. students, participants were exposed to a pair of persuasive essays that were allegedly written by members of the ingroup (as part of a cover story concerning abstract verbal reasoning); these essays were in actuality designed to alter perceptions of the legitimacy of the socioeconomic differences. In the high-legitimacy condition, for example, one of the essays read: There are two good reasons why UCSB students are less economically successful than UCLA students: (1) UCLA admits students with more varied and more experienced backgrounds, and these people have a better sense of what they want to do later in life; and (2) there is a perception out there (and it's probably right!) that UCSB students are partyers who do not take academics seriously enough. Both of these reasons would easily explain the disparities. With regard to the first, everyone knows that people who are ambitious and knowledgeable are in a better position to succeed economically, and they deserve that success. With regard to the second reason, potential employers are probably sensitive to legitimate differences in the qualifications of students at the two schools. The corresponding essay for the low-legitimacy condition was as follows: There are two main reasons why UCSB students are less economically successful than UCLA students: (1) UCLA admits more students with privileged backgrounds, and these people have more advantages to begin with and more connections later in life; and (2) there is a misperception that UCSB students are just partyers who do not take academics seriously. Neither one of these reasons is fair. With regard to the first, everyone knows that ’wealth begets wealth’ and that it is far easier for people of higher social classes to succeed economically, whether they deserve that success or not. With regard to the second reason, potential employers are probably relying on false perceptions, without paying enough attention to the merits of qualified individuals at UCSB.

< previous page

page_98

next page >

< previous page

page_99

next page >

Page 99 Fig. 6.2. Effects of legitimacy manipulation on ingroup and outgroup favoritism on the part of low status group members

Although a subset of participants (12%) were not persuaded by the essays, those who were persuaded that the system was either legitimate or illegitimate showed changes in their stereotypes (see Fig. 6.2). Compared with students who were assigned to the low- legitimacy condition, those who were assigned to the high-legitimacy condition exhibited stronger outgroup favoritism on status-relevant attributes (intelligent, hard working, skilled at verbal reasoning) and lesser ingroup favoritism on status-irrelevant attributes (honest, friendly, interesting). In addition to clarifying the important role of perceived legitimacy in determining the degree of ingroup or outgroup favoritism exhibited by low-status group members, the experimental paradigm we developed also provides us with some insight concerning the issues of disidentification and attribute relevance. Does Disidentification Account for Outgroup Favoritism in Low-status Groups? According to one prominent account derived from social identity theory, members of low-status groups exhibit outgroup favoritism to the extent that they disidentify with their own group. However, it may not always be feasible for people to avoid perceiving themselves in terms of ascribed group memberships and to persist in thinking of themselves as part of a group to which they do not belong. Research indicates, in fact, that ingroup identification tends to be stronger among members of some low-status groups (e.g., African Americans, Hispanic Americans) than among members of high-status groups (e.g., European Americans), insofar as the former group memberships are more numerically distinctive (rarer) than the latter (McGuire & McGuire, 1988). An even bigger challenge to the disidentification thesis is the fact that correlations between ingroup identification and ingroup favoritism are weak and inconsistent (Hinkle & Brown, 1990). A study by Mlicki and Ellemers (1996), for example, finds that Polish citizens identify especially strongly with their own national image, but the image that they hold is a predominantly negative one, presumably because of recent failures of their economic system. Therefore, it is doubtful that decreased ingroup identification among members of low-status groups, even if it can be demonstrated, would be sufficient to produce the extent of outgroup favoritism that has been observed. Although ingroup identification has not been a focal point of the research program reported on here, it was included as a measured variable in the Yale study and as a manipulated variable in the U.C.S.B. study. In the Yale study, no evidence was obtained for the disidentification

< previous page

page_99

next page >

< previous page

page_100

next page >

Page 100 Fig. 6.3. Stereotypic evaluations of high status and low status target groups by an observer group

hypothesis that members of low-success groups would identify less with the ingroup than would members of highsuccess groups. Ingroup identification was in fact nonsignificantly higher among members of low-success groups than among members of high-success groups, possibly because cognitive dissonance is aroused by belonging to a group that is low in social standing, and this dissonance may be reduced by redoubling one's investment in the group (cf. Turner, Hogg, Oakes, & Smith, 1984). In the U.C.S.B. study, a bogus pipeline procedure was used to convince members of low-status groups they were either especially high or low in ingroup identification (cf. Doosje, Ellemers, & Spears, 1995). No significant main effect of ingroup identification on ingroup favoritism was obtained, although this should have been expected on the basis of social identity theory (Tajfel & Turner, 1986). Taken as a whole, these findings deepen existing worries that social identity theory insufficiently explains status differences on ingroup favoritism (Hinkle & Brown, 1990; Jost & Banaji, 1994; Sidanius, 1993). None of this is to say that disidentification processes play no role in the phenomenon of outgroup favoritism or that issues of group justification are unrelated to intergroup evaluations. What our results suggest, however, is that disidentification with the ingroup does not seem to be a necessary prerequisite for the sort of system justifying outgroup favoritism observed among members of low-status groups . Does the Moderating Role of Attribute Relevance Support Social Identity Theory? Studies employing our experimental paradigm for manipulating perceived socioeconomic success replicate the finding that attribute relevance moderates the display of ingroup and outgroup favoritism on the part of low-status group members (e.g., Mullen et al., 1992; Mummendey & Schreiber, 1984; Skevington, 1981; Spears & Manstead, 1989; van Knippenberg, 1978). In the Maryland study (Jost & Burgess, 2000), for example, members of low-status groups exhibited strong outgroup favoritism on status-relevant attributes of intelligence, industriousness, and verbal reasoning ability, but they exhibited strong ingroup favoritism on status-irrelevant attributes of honesty, friendliness, and interestingness, a finding that was replicated in the U.C.S.B. study. Members of high-status groups, by contrast, exhibited ingroup favoritism on relevant and irrelevant attributes. Although we have replicated the pattern of results obtained by social identity theorists for relevant versus

< previous page

page_100

next page >

< previous page

page_101

next page >

Page 101 irrelevant attributes, another one of our studies casts doubt on the theoretical interpretation that has been offered repeatedly for this pattern. U.C.S.B. students were asked to rate two outgroups – Stanford and U.C.-Santa Cruz – on exactly the same stereotyping measures used in all of the prior studies. Although U.C.S.B. students did not belong to either of the groups being rated, the stereotype pattern they showed was almost identical to other studies in which the ingroup was implicated in the comparison: The higher status group (Stanford) was rated as more intelligent, more hard working, and more skilled at verbal reasoning, whereas the lower status group (U.C.-Santa Cruz) was rated as more honest, friendly, and interesting (see Fig. 6.3). In fact, the mean differences corresponded almost exactly to mean levels of ingroup and outgroup favoritism in our other studies. This suggests that the tendency among members of low-status groups to favor the ingroup on status-irrelevant dimensions of comparison may not actually be driven by group-justifying needs to compensate for threatened social identification, as has always been assumed. The possibility that people subscribe generally to lay or folk theories in which a negative correlation exists between social or economic status and favorable socioemotional characteristics (such as honesty and friendliness) is being addressed in ongoing research. In fact, it may be that such beliefs serve system-justifying ends and that highand low-status group members might all feel that the system is more legitimate if they can sustain stereotypes of the “poor but honest” or “poor but happy” variety. FUTURE DIRECTIONS IN THE THEORY OF SYSTEM JUSTIFICATION One question that has not been addressed sufficiently in the research reported herein is “why do people engage in system justification?” Our response is that it is probably overdetermined. There are many forces, internal and external to the individual, that pull for system justification sorts of attitudes and behaviors, and these should be investigated in future research. There are cognitive factors, such as genuine attempts to explain and understand the world, as well as tendencies to preserve existing attitudes and beliefs and to achieve certainty and closure. Thus, we have argued that political conservatism and other system-justifying attitudes serve to reduce uncertainty and satisfy the “need for cognitive closure’’ (e.g., Jost, Kruglanski, & Simon, 1999). In addition, there are motivations to stave off existential terror by preserving the sense that the world is a fair and manageable place in which people “get what they deserve and deserve what they get” (e.g., Lerner, 1980). Situational factors also determine the strength of system-justifying responses. In recent research with Yephat Kivetz, we have found that the presence of an ideological threat directed against the national system increases stereotypic differentiation between high-status Ahskenazi Jews and low-status Sepharadic Jews in Israel. In general, too, people face structural constraints, such as rewards for obeying authority and tolerating unequal outcomes, and punishments for challenging the legitimacy of the system. Therefore, we argue that one major (and overdetermined) function of attitudes, stereotypes, and social judgments is to justify existing social arrangements. System justification refers to a set of social, cognitive, and motivational tendencies to preserve the status quo through ideological means, especially when it involves rationalizing inequality among social groups. Our research (in collaboration with Grazia Guermandi, Monica Rubini, and Cristina Mosso) has begun to explore the ways in which people use stereotypes to justify socioeconomic differences between northerners and southerners in the United States, England, and Italy. There are other issues, too, that future research would do well to address. For one thing, it would be important to know the extent to which system justification and outgroup favoritism are truly internalized on the part of low-status group members, as opposed to being strategic, self-presentations displays that do not reflect privately held beliefs and attitudes. Jost and Banaji (1994) argued that outgroup favoritism might be even stronger at an unconscious or unexamined level of awareness insofar as such impulses would be less subject to controlled processing and to conscious activation of ego and group justification motives. Our research has begun to explore these issues, using reaction time paradigms and other unobtrusive methodologies to estimate the extent of outgroup favoritism on nonconscious cognitive, affective, and behavioral measures (Jost, Pelham, & Carvallo, 2000). Finally, the future of system-justification theory will have to accommodate exceptions to the rule by explaining when and why people fail to provide ideological support for the existing social system. In other words, the theory should be useful also for identifying opponent processes that govern group consciousness-raising and inspire social and organizational change. Some of these are likely to be associated with processes of ego justification and group justification, which are hypothesized to stand in an inverse relation to system justification processes for members of low-status groups (Jost et al., 2001). Individual difference variables might also identify

< previous page

page_101

next page >

< previous page

page_102

next page >

Page 102 people who are especially likely or unlikely to engage in system-justifying outgroup favoritism. Candidates include the “belief in a just world” scale and the “social dominance orientation” scale (see Jost & Burgess, 2000; Jost & Thompson, 2000), as well as political orientation (Jost, Burgess, & Mosso, in press). In collaboration with Grazia Guermandi and Erik Thompson, we have also been developing a scale of “economic system justification’’ (see Jost & Thompson, 2000). Evidence presented here – that perceptions of illegitimacy are associated with the rejection of outgroup favoritism – is also a step in the right direction, but more research is needed to determine when and why people will shed the layers of false consciousness and begin to substantially challenge the existence of social inequality in all of its various guises. ACKNOWLEDGMENT Portions of this chapter are based on writings submitted in partial fulfillment for the Ph.D. degree granted by the Department of Psychology at Yale University under the supervision of William J. McGuire. Funding for the research described herein was provided by a Robert M. Leylan Fellowship in Social Science from Yale University, a grant from the Faculty Research Assistant Program at U.C.-Santa Barbara (both awarded to the author), NIMH Grant #5R01MH32588 awarded to William J. McGuire, and NIMH Grant #R01-MH52578 awarded to Arie W. Kruglanski. I am grateful to Joyce Liu, Cristina Mosso, and Oliver Sheldon for assistance with manuscript preparation and to Sonya Grier and Michael Morris for helpful suggestions concerning revisions of this chapter.

< previous page

page_102

next page >

< previous page

page_103

next page >

Page 103 Part II.Building Blocks of Social Cognition: Representation and Structure

< previous page

page_103

next page >

< previous page

page_104

next page >

page_104

next page >

Page 104 This page intentionally left blank.

< previous page

< previous page

page_105

next page >

Page 105 7 A Case for the Nonconscious Self-Concept John J. Hetts Brett W. Pelham Ohio State University, Columbus State University of New York at Buffalo In fact, I cannot totally grasp all that I am. Thus, the mind is not large enough to contain itself: but where can that part of it be which it does not contain? – St. Augustine Virtually all contemporary research on the self-concept has examined people's self-reported, consciously accessible self-conceptions (for reviews, see Banaji & Prentice, 1994; Baumeister, 1998; Brown, 1998; Kihlstrom & Klein, 1994; Wylie, 1974, 1979). However, emerging evidence suggests there are many aspects of experience that are outside our conscious awareness and control, including aspects of memory, learning, attitudes, stereotypes, and, notably, self-esteem (e.g., Bargh, 1997; Bargh & Chartrand, 1999; Berry & Dienes, 1993; Greenwald & Banaji, 1995; Schacter, 1992). Evidence has thus begun to suggest that St. Augustine was correct. Apparently a large part of who we are remains inaccessible to us. That is, there appear to be important aspects of the self-concept that reside outside of conscious awareness. The idea that there are domains of the mind and the self to which people have little access and over which people have little control is not new. In fact, this idea can be traced back to ancient Greek and Roman philosophers. For example, the idea we are often unable to exert our will over our behavior was expressed both by Ovid, “I see and approve better things but follow worse,” and Euripedes, “We know the good, we apprehend it clearly, but we can not bring it to achievement.” Throughout history, a variety of philosophers, writers, poets, play-wrights, and physicians (inter alia St. Thomas Aquinas, Cervantes, Galen, Montaigne, and Shakespeare) have explored mental processes over which people have little awareness or control (for reviews, see Margetts, 1953; Whyte, 1978). In recent years, researchers have developed techniques for examining unconscious beliefs, and we believe such techniques contribute to the development of a complete theory of the mind and the self-concept. Before discussing contemporary research on nonconscious aspects of the self-concept, however, we briefly review the history of how people have understood nonconscious thought. HISTORICAL FOUNDATIONS OF THE MODERN CONCEPT OF THE UNCONSCIOUS Many revered thinkers have tackled the concepts of the unconscious and the self. Perhaps more than any other thinker, René Descartes (1641/1993) laid the foundation for a modern understanding of nonconscious aspects of the self. However, he did so not by developing the concept of the unconscious, but by developing a dualistic model of human experience. Two of Descartes' premises proved critical. The first premise was that human beings represent a union between a self-conscious, thinking entity and an unthinking, reflexive mechanical body. The second premise was that the self resides in the thinking entity, even in thought itself (thus the famous phrase “Cogito, ergo sum”). Descartes assumed that self-understanding (in addition to understanding the world more generally) could begin only with the awareness of one's existence. Additional principles of self-understanding could subsequently be deduced on a logical basis. This rational model of selfhood had a profound influence on the understanding of nonconscious elements of the mind. Descartes' dualistic position on mind and body dictated that anything outside of awareness must reside in the body rather than in the mind. Thus, the body became the source of drives, needs, and instincts. Reactions against Descartes' emphasis on consciousness facilitated the development of the idea that there are aspects of the mind and self of which we are not aware. These ideas laid the foundation for our contemporary understanding of the mind and, as we hope to demonstrate here, the self. The Cognitive Unconscious One of the first thinkers to organize arguments against

< previous page

page_105

next page >

< previous page

page_106

next page >

Page 106 Descartes' position on mind and self was Blaise Pascal. Pascal (1670/1966) held that aspects of both self and behavior lay outside of our awareness. As he put it, “The heart has its reasons, which reason knows nothing of.’’ Pascal further held that the conscious, self-aware thinking entity (Descartes' self) sometimes interfered with the achievement of happiness (cf. Wilson & Schooler, 1991). Both Spinoza and Leibniz emphasized similar perspectives. For example, Leibniz (1704/1996) foreshadowed Freud's conception of the preconscious, Weber's (1834) conception of absolute threshold, and many contemporary analyses of consciousness (e.g., Dennett, 1991) by arguing that the perceptions of which we are aware are the accumulation of smaller perceptions that are of insufficient size to yield awareness. Like many modern students of the cognitive unconscious (e.g., Kihlstrom, 1990; Schacter, 1992) Leibniz also postulated that people's experiences could influence later thoughts and behaviors in the absence of any conscious memory for the experiences. The Affective and Motivational Unconscious In the wake of Descartes and his contemporaries, other philosophers laid the groundwork for a second dimension of modern understandings of the unconscious: the idea of nonconscious emotions, instincts, and drives. Rousseau (1781-1788/1995) suggested that many uncontrollable behaviors, both remarkable and mundane, are primarily grounded in emotions or feelings. Hume (1777/1976) went further, suggesting that nearly all behaviors, including those we perceive to be consciously controlled, are grounded in instinct or physiology. In a similar way, Goethe (1957) located the imagination completely in the unconscious and posited that the mind was the unified interplay of conscious and unconscious aspects of existence. He thus proposed that no conscious thought existed without the operation of unconscious elements and thus anticipated contemporary views of intuition and insight (e.g., Dorfman, Shames, & Kihlstrom, 1996). The Pathological Unconscious As thinking about the unconscious continued to evolve, philosophers began to speculate about harmful aspects of the unconscious. For example, Schopenhauer (1819/1995) suggested that the primary purpose of consciousness was to rationalize unconscious desires and instincts. Similarly, Nietzche (1954/1977) held that humans are not rational but rather rationalizing, self-deceptive beings strongly influenced by unconscious drives. Nietzche anticipated a great deal of Freud's description of the unconscious. For example, Nietzche foreshadowed Freud's argument that difficult or disturbing memories are often forced out of consciousness. Nietzche also suggested that consciousness is a late and incomplete evolutionary development (cf. James, 1890/1950). Thus, to Nietzche, consciousness was a narrow, superficial epiphenomenon whose impact on behavior paled in comparison with unconscious influences. The growing emphasis on pathological aspects of the unconscious during the late 18th and early 19th centuries was partly fueled by scientific attempts to understand hysteria (Kelly, 1991). Franz Mesmer was at the forefront of such efforts. Mesmer discovered that he could either elicit or relieve various psychosomatic symptoms through “animal magnetism,” now known as hypnotism. Expanding on these mesmerizing ideas, Charcot began the differentiation between neurological and unconscious causes of mental illness, paving the way for the subsequent psychological study of the unconscious by Janet, Freud, and Jung. Janet (1907) was one of the earliest thinkers to employ careful observation in the study of unconscious aspects of psychopathology. Moreover, he argued forcefully that the understanding of the unconscious mind was crucial to the treatment of certain mental disorders. His work, largely overshadowed by Freud's, helped pioneer the use of projective tasks such as hypnosis, automatic writing, and dream interpretation, techniques that were adopted in earnest by Freud and Jung. Although earlier thinkers anticipated much of his work, Freud did more than anyone else in history to increase people's interest in the unconscious. He did so by postulating an intelligent, motivated, dynamic unconscious. In his early work, Freud (1900/1965) theorized that the human mind could be divided into conscious and unconscious parts. He further theorized that the unconscious mind could be divided into a portion that could be brought into consciousness (the preconscious) and a fully unconscious portion that contained material too difficult for the conscious mind to accept. Freud (e.g., 1927/1950, 1933/1965) later revised this division, postulating three distinct components of the mind: the now familiar id, ego, and superego. From this perspective, only the ego was primarily, although not entirely, conscious. Freud felt that the source of many mental illnesses was the ego's unsuccessful negotiation of the conflict between the constraints of reality and the motives of the two primarily unconscious components (id and superego). Freud took a number of routes to bringing people's unconscious thoughts into consciousness, including dream interpretation and free association. It is noteworthy that many of

< previous page

page_106

next page >

< previous page

page_107

next page >

Page 107 these projective tasks bear more than a superficial resemblance to modern tasks used to assess implicit memory (e.g., Toth & Reingold, 1996). Because of Freud's emphasis on controversial ideas such infantile sexuality, catharsis, and repression, many researchers interested in the mind and the self failed to appreciate the importance of Freud's insights about the unconscious (see Epstein, 1994; Westen, 1998a, 1998b, for related discussions). Nonetheless, Freud and his successors generated a great deal of interest in the unconscious, and many early ideas about the unconscious have been validated by contemporary research (see Newman, Duff, & Baumeister, 1997; Westen, 1998a, 1998b). Much of this work could also be considered an extension of Jung's less exotic view that the unconscious performs most of the same functions that more typically take place in the conscious mind. Jung (1961) also anticipated contemporary work on the unconscious by emphasizing that we often experience the products rather than the processes of unconscious mental events. These products then form the contents of consciousness (cf. Nisbett & Wilson, 1977). THE PSYCHOLOGICAL UNCONSCIOUS Psychoanalytic theorists were not the only ones who paved the way for the study of the unconscious. Early students of perception, learning, and memory also did so by anticipating phenomena such as implicit memory and implicit learning. Ebbinghaus (1885/1964), for example, observed that even when people could not consciously recall previously studied stimuli, they could learn these stimuli more quickly than could people who had never before seen the stimuli. Ebbinghaus specifically noted that these savings represented a measure of memory independent of people's conscious recollections. Helmholtz (1866/1925) reintroduced the idea of unconscious inferences, reasoning that some forms of learning (e.g., depth perception) are unconscious. William James and Functionalism One of the foremost scholars of the self, William James also proved to be highly influential in developing our modern understanding of the nonconscious mind. One of the major contributions of James' perspective on human behavior was to shift study away from the contents toward the functions of consciousness. This change greatly expanded existing theories of mental processes and laid the foundation for later interest in the idea that many everyday psychological processes occur outside of conscious awareness. James (1890/1950) took issue with the idea that consciousness is epiphenomenal and has no causal influence on behavior. Questioning why consciousness should closely parallel something that it has nothing to do with, he suggested that consciousness has discrete functions and observable effects. James concluded that the primary purpose of consciousness is selective attention. It directs attention when there is competition from multiple impinging sensations (e.g., it allows one to keep writing despite the impinging sensations of hunger). What function can this serve? James postulated that consciousness and selective attention allow people to behave in highly efficient ways in a variety of circumstances. That is, consciousness functions in the short run in much the same way that evolution functions in the long run: It quickly produces adaptive behaviors that could take eons to evolve in the absence of consciousness. According to James, consciousness utilizes the flexibility of higher brain functions while allowing people to take advantage of well-learned habits. James took it as evidence for his position that consciousness (i.e., self-awareness) is highest at moments of greatest indecision (i.e., in the face of danger, uncertainty, or competing cues for behavior; see Baumeister, 1998; Carr, 1925; Wegner & Bargh, 1998). At the same time, James also disputed contemporary descriptions of the unconscious as a font of desires, wills, repressed pains, and instinct, dismissing such theorizing as untestable (a claim, we suggest, that proved to be incorrect). However, despite his emphasis on the prepotency of consciousness, James made substantial contributions to the position that unconscious processes are important to daily experience. He did so primarily in his treatments of habit and emotion. Habit . James thought of habits as consistent modes of conduct that become easier to activate with repeated use. Just as a path through a thicket becomes easier to traverse with each additional trip, habits become more inviting with each additional expression. Thus, according to James, one of the important functions of habit is to diminish the conscious attention necessary to perform a task. Thus, habits become self-sustaining: One aspect leads to the next without the necessity of conscious oversight (cf. Bargh & Chartrand, 1999; Posner & Snyder, 1975). Emotion. According to James, emotional experience derives partly from bodily movements and actions: “We feel sorry because we cry, angry because we strike, afraid because we tremble” (James, 1890/1950, II, pp 449-450). This suggests that the conscious experience of emotion has important determinants outside of our awareness (namely, behavioral and physiological responses). This

< previous page

page_107

next page >

< previous page

page_108

next page >

Page 108 perspective anticipated the contemporary view that emotions are often a joint product of arousal and labeling (cf. Schacter & Singer, 1962). Unfortunately, as the ideas of James and Freud were beginning to take root in psychology, a paradigm shift toward behaviorism lead to a deemphasis of consciousness, the self, and mental processes. Behaviorism's emphasis on objective verification and the intractability of mental processes relegated the self – including the unconscious self – to the philosophers armchair. However, behaviorists and neobehaviorists eventually provided researchers with the paradigms, tools, and hypothetical constructs (from shuttle boxes and latent learning, to Skinner boxes and cognitive maps) that eventually made it possible to study unconscious mental processes. Moreover, they began the accumulation of evidence of automatic behavior (by showing that stimuli often lead automatically to responses). Paradoxically, then, behaviorists and learning theorists invented and refined many of the tools that are now used to open the black boxes they typically considered impenetrable. MODERN EVIDENCE FOR NON-CONSCIOUS PROCESSES AND THEIR IMPLICATIONS FOR THE SELF The cognitive revolution in psychology revived psychologists interest in mental processes, eventually including unconscious mental processes. This development, however, owes more to attempts to understand the cognitive and affective components of the unconscious than to the study of Freudian dynamics (Kihlstrom, 1984, 1987, 1990). Although Freud deserves the credit for popularizing the importance of the unconscious, modern treatments of the unconscious are rooted less in Freud's analysis of the dysfunctional and maladaptive and more in James analysis of the functional and adaptive (cf., however, Westen, 1998a, 1998b). For example, reviewing a great deal of past and recent research, Bargh and Chartrand (1999; Bargh, 1997) argued that the vast majority of our thoughts, emotions, and behavior are determined by preconscious automatic processes (processes that depend solely on the presence of a triggering stimulus or condition). They further argued that this routine automatization of virtually every aspect of our mental lives liberates our conscious mind from monitoring and directing the multitudes of our mental activities. Drawing on inferences based on the study of skill acquisition and knowledge compilation (as reviewed in Bargh, 1996), dating from James chapter on habit, Bargh and Chartrand argued that, over time, our mental activities are increasing delegated to the efficient and reliable preconscious automatic processes. This allows conscious resources to be allocated to tasks that require attention and cognitive resources.1 To develop our case for the existence of nonconscious elements of the self-concept, we first examine current conceptualizations of the nonconscious mind as it pertains to memory, social cognition, and affect. We then suggest how the findings in those areas may be applied to our understanding of the self. However, before specifying the likely architecture of the nonconscious self, we begin by outlining the general terrain of the nonconscious. What are the hallmarks of automatic or nonconscious cognition? Bargh (1989, 1994) identified four important attributes of nonconscious or automatic processing (see also Shiffrin & Schneider, 1977). They include (a) a lack of awareness of a stimulus or of the stimulus effects on subsequent experience, (b) a lack of intention to make a judgment or act in a certain way, (c) high levels of efficiency (automatic processes require little processing capacity), and (d) low levels of controllability. Few processes meet all four of these criteria. Rather, most processes involve a mixture of nonconscious and conscious processes (Bargh, 1994; Jacoby, 1991). We thus focus on processes that reveal at least some of the qualities of nonconscious processing. Nonconscious Cognition: Implicit Memory and Implicit Learning Research on implicit learning and memory holds substantial promise for informing nonconscious aspects of the self. Researchers who have tackled implicit learning and memory have learned to identify introspectively inaccessible effects of previous experience on the facilitation or disruption of current task performance (for reviews, see Jacoby & Kelley, 1987; Richardson-Klavehn & Bjork, 1988; Roediger, 1990; Schacter, 1987, 1989, 1992). For example, researchers have assessed implicit memory for previously viewed stimuli using performance on perceptual identification tasks such as the recognition of degraded words (e.g., Jacoby & Dallas, 1981): Other researchers have assessed implicit memory using word-stem or word-fragment completions. Even when participants cannot consciously recall words to which they were recently exposed, this previous exposure increases the likelihood that participants will generate such words in the word-completion task. For example, participants recently exposed to the word caper may not be able to recall the word, but they will become especially likely to generate this word in response to fragments such as “____APER” or “CAP____” (Tulving, Schacter, & Stark, 1982). These implicit effects of prior exposure occur

< previous page

page_108

next page >

< previous page

page_109

next page >

Page 109 even when exposure to the critical stimuli is incidental (Eich, 1984) or extremely brief (e.g., Kunst-Wilson & Zajonc, 1980; Murphy & Zajonc, 1993), and they occur even for patients with profound forms of amnesia (Schacter, 1996). Not only have unconscious, introspectively inaccessible effects of previous experiences been systematically identified, but they have also been carefully distinguished from the conscious remembering that people more commonly think of as memory. In an influential review, Schacter (1987) summarized the observed distinctions between implicit and explicit memory. Four of his points are highly relevant to our position here. First, Schacter noted, manipulations that influence explicit memory (e.g., depth of processing) typically have little or no effects on implicit memory (e.g., Graf & Mandler, 1984; Jacoby & Dallas, 1981; Winnick & Daniel, 1970). Second, seemingly superficial changes in the stimuli between exposure and testing (e.g., moving from auditory to visual presentation) have little effect on explicit memory but can substantially disrupt implicit memory (e.g., Jacoby & Dallas, 1981; Roediger & Blaxton, 1987). Third, the interval from exposure to test often has substantially different effects on implicit and explicit memory. Interestingly, implicit memory sometimes decays very quickly relative to explicit memory (e.g., Graf & Mandler, 1984) and sometimes persists much longer (Jacoby & Dallas, 1981; Tulving et al., 1982). Fourth, measures of implicit and explicit memory are typically uncorrelated (e.g., Eich, 1984; Graf & Schacter, 1985; Jacoby & Witherspoon, 1982; Tulving et al., 1982). Schacter (1987, 1989, 1996; see also Moscovitch, Goshen-Gottstein, & Vriezen, 1994; Shimamura, 1986; Weiskrantz, 1989) also noted that, unlike explicit memory, implicit memory is highly robust in the face of physiological insult, including amnesia. Similarly, whereas explicit memory declines with age, implicit memory does not (Light, 1996). Conversely, in infants and children, explicit memory increases with age, whereas implicit memory remains relatively constant (Rovee-Collier, 1997). Although other evidence has begun to suggest that the dissociations noted earlier are not as extreme as Schacter originally suggested (e.g., Graf & Schacter, 1985; Schacter & Graf, 1986; Sloman, Hayman, Ohta, Law, & Tulving, 1988), his basic conclusions seem to be well supported (Kelley & Lindsay, 1996; Schacter, 1989). To account for the observed dissociations, some have proposed different underlying memory systems (e.g., Tulving, 1985; Tulving & Schacter, 1990). However, others have provided alternate interpretations based on similarities and differences between encoding and retrieval processes (i.e., transfer appropriate processing; Roediger, 1990; Roediger & Blaxton, 1987). From our point of view, however, whether the dissociations represent different types of encoding and retrieval processes or distinct memory systems is not critically important. Regardless of the precise source of these dissociations, they suggest that our understanding of memory and, by extension, the self can be greatly expanded by attending to these dissociations and what they suggest about nonconscious processes. One important controversy that has proved useful to understanding conscious and nonconscious aspects of memory is the controversy over “process-purity’’ (Jacoby, 1991). Frequently, implicit and explicit memory tasks are assumed to represent implicit and explicit memory processes. However, many memory tasks may reflect a combination of implicit and explicit memory processes (e.g., Richardson-Klavehn & Bjork, 1988). Consequently, some researchers have attempted to develop techniques that distinguish between conscious and nonconscious aspects of memory. Schacter, Bowers, and Booker (1989), for example, suggested that implicit and explicit measures of memory should be made as similar as possible except for their instructions. For example, some participants might be instructed to complete a word fragment (e.g., _APER) with whatever word first comes to mind, whereas others might be instructed to use the word fragment as a recall cue (Richardson-Klavehn, Gardiner, & Java, 1996). Alternately, other researchers have attempted to place implicit and explicit processes in direct opposition to one another. To this end, Jacoby and colleagues (Jacoby, 1991; Jacoby, Toth, & Yonelinas, 1993; Jacoby, Toth, Lindsay, & Debner, 1992; Toth, Reingold, & Jacoby, 1994) have developed the “process dissociation” technique.3 Jacoby and colleagues compare participants responses on an implicit memory task under two conditions. In one condition, participants are instructed to base their responses on previously encountered stimuli. Both conscious and nonconscious processes ostensibly influence these responses. In the other condition, participants are instructed to base their responses on anything but previously encountered stimuli. Thus, task performance should only reflect nonconscious processes in the absence of any conscious contamination. Thus, the difference between these two conditions reflects the contribution of conscious processes to performance on the memory task. The contribution of the nonconscious processes to task performance, however, depends on the underlying assumption regarding the degree of independence between conscious and unconscious processes,

< previous page

page_109

next page >

< previous page

page_110

next page >

Page 110 which is a matter of some controversy. To date, however, such techniques have produced dissociations between nonconscious and conscious aspects of memory consistent with previously observed patterns, although the size of the nonconscious effects now appears to be smaller than previously thought (Kelley & Lindsay, 1996; RichardsonKlavehn, Gardiner, & Java, 1996). Whether such techniques could be adapted to refine the measurement of people's implicit beliefs about themselves remains an open and important question. In addition to research on implicit memory, research on implicit learning (for reviews, see Berry, 1996; Berry & Broadbent, 1995; Berry & Dienes, 1993; Seger, 1994) also suggests that people possess and acquire knowledge of which they are unaware. Specifically, research on implicit learning indicates that people can learn to respond appropriately to complex situations in the absence of the ability to explain how they are managing to do so. One of the most mundane and important examples of this is the ability of young children to produce grammatically correct language without being able to explain the rules of the acquired grammar. Implicit learning has also been demonstrated in the learning of artificial grammar. For example, Reber (1967) demonstrated that when participants are asked to memorize letter strings that have an underlying grammar,they can readily discriminate subsequent letter strings as grammatical or ungrammatical. This performance occurs in the absence of awareness that one is studying letter strings with an underlying grammar. In fact, instructions to attend to an underlying grammar during learning appear to interfere with learning (Reber, 1967)4 Nonconscious Affect: The Primacy and Automaticity of Affect Several areas of research suggest that attitudes toward social objects are activated relatively automatically and in the absence of conscious processing, even in the absence of explicit attitudes for the object (for a further review, see Roese, chap. 9, this volume). First, research has revealed that mere exposure to a stimulus leads to positive evaluations of that stimulus, with liking increasing as a function of number of exposures (Mandler, Nakamura, & Van Zandt, 1987; Seamon, Brody, & Kauff, 1983; Zajonc, 1980). These effects occur in the absence of immediate identification or later recognition of the stimulus (Zajonc, 1980), or even when the stimuli are presented subliminally (Kunst-Wilson & Zajonc, 1980). Second, recent evidence suggests that attitudes are activated automatically by the mere presence of an attitude object (Bargh, Chaiken, Govender, & Pratto, 1992; Fazio, Sanbonmatsu, Powell, & Kardes, 1986). For instance, for many graduate students, exposure to the stimulus qualifying exam might automatically activate a wide variety of negative associations, making it easier for them to respond to subsequent negative targets (e.g., nonsignificant) and more difficult to respond to positive targets (Ph.D.). This effect becomes more robust when the judgments about the object are more likely to preclude conscious evaluation (Bargh, Chaiken, Raymond, & Hymes, 1996). Finally, neurological studies suggest that affective or evaluative processing is governed by relatively primitive brain centers such as the amygdala and the hippocampus (for reviews, see LeDoux, 1992, 1995; MacLean, 1993; Zajonc, 1998). For example, according to Damasio's (1994) somatic marker hypothesis, perceptual information is relayed directly to these centers, and the affective responses that occur in these emotion centers then shape identification and higher order processing of the information. Damasio's position dovetails nicely with the evidence for the automatic attitude activation effect. Nonconscious Social Cognition Implicit Inference. Both the extensive research on implicit learning and memory and the cognitive revolution in social psychology have spurred a new approach to understanding nonconscious processes in social inference. An early reflection of this influence appeared in research on social judgment and trait inference. One area of this research focused on construct accessibility – the process through which recently activated concepts influence the categorization or evaluation of subsequent (usually ambiguous) stimuli. This research documented that recent exposure to a construct (e.g., priming) increases the likelihood that the construct will be employed in subsequent social judgments (e.g., Bargh, Bond, Lombardi, & Tota, 1986; Higgins, Bargh, & Lombardi, 1985; Higgins, Rholes, & Jones, 1977), especially under conditions of repeated exposure (e.g., Srull & Wyer, 1979, 1980) and even when people are not aware of the exposure (e.g., Bargh & Pietromonaco, 1982). Repeated activation of a construct over long periods also increases the likelihood of the constructs application (e.g., Bargh & Pratto, 1986; Bargh & Thein, 1985; Higgins, King, & Mavin, 1982). Moreover, the application of chronically accessible constructs appears to be relatively efficient, occurring even under cognitive load (Bargh & Thein, 1985; Bargh & Tota; 1988)5 A second area of research on nonconscious aspects of social inference is research on spontaneous trait

< previous page

page_110

next page >

< previous page

page_111

next page >

Page 111 inferences (for reviews, see Newman & Uleman, 1989; Uleman, Newman, & Moskowitz, 1996). Participants unconsciously and automatically categorize certain social behaviors as reflecting underlying traits, and they do so in the absence of the intention to do so (e.g., Winter & Uleman, 1984; Winter, Uleman, & Cunniff, 1985). Moreover, these spontaneous categorizations can influence subsequent judgments (Moskowitz & Roman, 1992), as in research on construct accessibility. For example, participants who are instructed to unscramble sentences containing hostile content later perceive ambiguous targets as more hostile than usual (Srull & Wyer, 1979)6 The work of Gilbert and colleagues (e.g., Gilbert, 1989; Gilbert, Pelham, & Krull, 1988; Gilbert, Krull, & Pelham, 1988) on the correspondence bias (i.e., the tendency to attribute behavior to the actors disposition rather than to aspects of the situation) represents a third example of nonconscious aspects of social inference. Gilbert and colleagues argued that an action or behavior is automatically categorized and then attributed to the disposition of the actor (what they referred to as characterization). They further argued that correcting such characterizations (e.g., to make a situational attribution) takes mental effort and cognitive resources. Consistent with this position, participants deprived of cognitive resources exhibit enhanced tendencies to attribute behavior to actors dispositions. Implicit Stereotyping and Automatic Prejudice. In much the same way as research on construct accessibility, spontaneous trait inferences, and the automaticity of attitudes, contemporary research on stereotyping and prejudice has also applied cognitive models of the unconscious to the understanding of human social behavior. More specifically, there has been a rapid proliferation of research on the nonconscious aspects of stereotypes and prejudice. First, building largely on research in implicit memory, research on implicit stereotyping has examined the implicit influence of gender or race-related information on subsequent judgment (e.g., Banaji & Greenwald, 1995; Banaji & Hardin, 1996; Banaji, Hardin, & Rothman, 1993; Blair & Banaji, 1996; Gilbert & Hixon, 1991) and behavior (e.g., Bargh, Chen, & Burrows, 1996; Chen & Bargh, 1997). For example, Banaji and Hardin (1996) found that participants primed with words associated with gender (e.g., nurse) responded more quickly than usual to genderconsistent target pronouns (e.g., she ) and less quickly than usual to gender-inconsistent target pronouns (e.g., he). This was true whether participants were required to judge whether the pronouns were male or female or whether they simply reported whether target words were pronouns). Second, researchers have also attempted to measure people's automatic evaluative associations to members of stigmatized groups, usually Blacks, building on research on construct accessibility and automatic attitude activation. That is, researchers have developed automatic measures of prejudice (Devine, 1989; Dovidio, Evans, & Tyler, 1986; Dovidio, Kawakami, Johnson, Johnson, & Howard, 1997; Fazio & Dunton, 1997; Fazio, Jackson, Dunton, & Williams, 1995; Gaertner & McLaughlin, 1983; Greenwald, McGhee, & Schwartz, 1998; Lepore & Brown, 1997; Moskowitz, Wasel, Gollwitzer, & Schaal, 1999; von Hippel, Sekaquaptewa, & Vargas, 1997; Wittenbrink, Judd, & Park, 1995). Automatic prejudice is usually operationalized in terms of the relative accessibility of positive and negative thoughts after exposure to race-related stimuli. For example, Fazio and colleagues (Fazio et al., 1995; Fazio & Dunton, 1997) primed participants with either a Black or a White face and then asked participants to quickly categorize subsequent positive and negative adjectives as positive or negative. They argued that if a people's automatic associations to a group are positive, people primed to think about the group should be able to subsequently identify positive adjectives more quickly than negative adjectives (Bargh et al., 1992; Fazio et al., 1986). Conversely, if people have negative associations about a group, these processing differences should be reversed. People's response to target words under different priming conditions can be used to generate implicit stereotyping scores (in this case, for implicit stereotypes about Blacks). Most research on automatic prejudice relies on some variation of this technique. Regardless of the assessment technique used, research on implicit stereotyping and automatic prejudice has yielded a number of consistent findings. First, like work on implicit learning and memory, research on implicit stereotyping has typically shown that both implicit stereotyping and automatic prejudice are unrelated to their explicit, consciously controlled analogs (cf., however, Dovidio et al., 1997, Study 2; Lepore & Brown, 1997; Wittenbrink et al., 1995). Furthermore, both those high and low in explicit stereotyping typically show evidence of implicit stereotyping, suggesting that implicit stereotyping occurs even among those who wish to avoid stereotyping others. Finally, preliminary evidence suggests that measures of implicit stereotyping predict a number of variables better than explicit measures of stereotyping. These include nonverbal and spontaneous behaviors, the use of ethnic or gender categories in person perception, and other implicit measures of racial attitudes. In contrast, explicit measures of racial attitudes tend to predict more deliberative

< previous page

page_111

next page >

< previous page

page_112

next page >

Page 112 judgments such as the decisions of mock jurors or explicit interpersonal evaluations. Although this area clearly represents a promising new avenue of exploration for understanding stereotyping and prejudice, a number of questions remain. For instance, the meaning and significance of implicit measures of stereotyping are under contention. Some researchers characterize measures of automatic prejudice as a “bona fide pipeline” to participants true attitudes (Fazio et al., 1995; see also Greenwald et al., 1998). Others suggest that implicit measures simply represent a different aspect of stereotyping and prejudice, a set of overlearned associations that subsequently influence behavioral responses to category members (e.g., Moskowitz, Salomon, & Taylor, 2000). Perhaps the most consequential debate revolves around the inevitability of the influence of implicit stereotypes on judgment and behavior. If implicit stereotypes and automatic prejudice are typically unconscious, difficult to disrupt, and widely held, then their impact on judgment and behavior may be much more pervasive than has often been suggested. However, emerging research suggests that implicit stereotypes and automatic prejudice are not always activated in the presence of a stereotyped target, they may be activated but not applied in judgment, and their activation or application may be influenced by current processing goals (see Blair, chap. 22, this volume; but cf. Bargh, 1999). BEYOND SELF-REPORT: PEERING DEEPER INTO THE MINDS I By far the most important object of social cognition, however, is the person doing the cognizing. (Kihlstrom & Klein, 1994, p. 154) Given that the self is one of the most widely studied domains in social psychology, we have been surprised by the lack of attention given to nonconscious elements of the self. Indeed, in many reviews of nonconscious social cognition, the applications of these findings to the self often receive little consideration (cf. Greenwald & Banaji, 1995). Consequently, we briefly summarize some reasons to suspect that many of the unconscious processes reviewed thus far are likely to play a role in the self-concept. Nonconscious Elements of the Self-Concept First, memory and the self appear to be closely intertwined. Numerous thinkers have suggested that a persons stable sense of self arises from autobiographical memory (Greenwald, 1980; Greenwald & Banaji, 1989; Greenwald & Pratkanis, 1984; James, 1890/1950; Kihlstrom et al., 1988). Recent research in social cognition has strongly implicated memory, learning, and stereotypes as important aspects of self-construction (see Kihlstrom & Klein, 1994; Linville & Carlston, 1994). For example, research on the self-reference effect has demonstrated that information related to the self is typically remembered better than other information (e.g., Rogers, Kuiper, & Kirker, 1977). Moreover, information relevant to the self is more easily and more quickly activated than other types of information. For example, in her research on self-schemata (i.e., cognitive representations of important self-relevant characteristics), Markus (1977) found that people can decide more quickly whether trait terms describe them when the traits under consideration are part of people's self-schemata. Similarly, Bargh (1982) found that participants performance in a dichotic listening task was significantly impaired when self-relevant stimuli were presented in the non-attended ear, suggesting that selfrelevant stimuli are highly accessible and readily (perhaps automatically) capture people's attention. These and other egocentric biases in encoding, processing, and retrieval led some to suggest that the self was a unique cognitive structure. Although the exact interpretation of these and similar findings has been the subject of some debate (e.g., Greenwald & Banaji, 1989; Higgins & Bargh, 1987; Klein & Kihlstrom, 1986), most now agree that these effects are not unique to the self. Rather, all kinds of well-elaborated, highly accessible cognitive structures share similar qualities, generating the observed advantages in attention, encoding, and retrieval. However, even critics of the “self is special” hypothesis agree that the self-concept is one of the most elaborated, evaluatively laden, and accessible of all cognitive structures.7 Clearly the self shares features with other well-elaborated cognitive structures. Thus, it seems reasonable to expect that the processes involved in implicit memory and implicit learning generally should apply to the self as well. That is, a persons experiences should exert nonconscious influences on the persons self-conceptions and self-evaluations. At the same time, the self should exert nonconscious influences on social perception and judgment. Implicit Influences on the Self-Concept. One major area of research that can be understood in terms of the implicit influences on self-relevant beliefs is research on self-presentation. It is well established that our attempts to present ourselves in a particular fashion to others can lead to changes in how we perceive ourselves (for a review, see Tice, 1994). That is, self-presentation often

< previous page

page_112

next page >

< previous page

page_113

next page >

Page 113 leads to self-concept change (e.g., Fazio, Effrein, & Falender, 1981; Jones, Rhodewalt, Berglas, & Skelton, 1981; Tice, 1992), especially when the self-presentation occurs publicly (Tice, 1992, 1994) and when the initial Selfconceptions are not held with great certainty (Rhodewalt & Agustsdottir, 1986; Schlenker & Trudeau, 1990). The desire to feel good about oneself can also have an implicit influence on ones self-views (Dunning, Perie, & Story, 1991; Gump & Kulik, 1995; Kunda, 1990; Kunda & Sanitioso, 1989; Sanitioso, Kunda, & Fong, 1990). For example, Kunda and Sanitioso (1989) found that people altered their self-ratings of introversion and extroversion after reading a story about a student who either did very well or very poorly in school and who was either very introverted or very extroverted. When extroversion was linked with success in the story, participants described themselves as more extroverted. When the reverse was true, participants described themselves as more introverted. Many other temporary situational influences may influence our self-conceptions in ways of which we are unaware. For example, moods appear to exert a largely unconscious influence on people's self-views. For example, induced negative mood increases the accessibility of negative self-relevant thoughts (e.g., Brown & Taylor, 1986; Teasdale & Fogarty, 1979; Teasdale, Taylor, & Fogarty, 1980; see also Sedikides, 1995). Moreover, some kinds of chronic negative moods (e.g., depressive symptoms) may lead to the development of automatic negative associations to the self (Bargh & Tota, 1988). As another example, manipulations of the distinctiveness of self-attributes (e.g., being the only psychologist in a room full of politicians) can cause those attributes to take on a prominent role in self-descriptions and subsequent behavior (Markus & Kunda, 1986; McGuire & McGuire, 1981, 1988). If manipulations of distinctiveness are relatively enduring (such as the gender composition of a persons family), the distinctive attribute can become a stable part of the persons self-description (e.g., McGuire, McGuire, Child, & Fujioka, 1978; McGuire, McGuire, & Winton, 1979). Similarly, chronic social circumstances can also implicitly shape self-conceptions. For example, students from elite schools perceive themselves to be of lower academic ability than do students of equivalent ability levels from less elite schools (e.g., Bachman & O'Malley, 1986; Davis, 1966; Lieberman, 1956; Marsh & Parker, 1984). Implicit Influences of the Self-Concept. One might also expect that the self exerts influences on perceptions and actions that operate outside of awareness. For example, research has shown that existing categories or prototypes are imposed on newly encountered stimuli in a relatively nonconscious way (e.g., Higgins, Rholes, & Jones, 1977; Hill, Lewicki, Czyzewska, & Boss, 1989; Lewicki, 1986; Lewicki, Hill, & Czyzewska, 1992; Lewicki, Hill, & Sasaki, 1989). Presumably, the self includes an important set of highly accessible categories and prototypes and thus might influence the perception and encoding of ambiguous self-relevant stimuli (cf. Greenwald, 1980; Swann, 1987, 1990). Recent research by Story (1998) confirms this idea (see also Tafarodi, 1998). When participants were given ambiguous feedback, participants with high self-esteem recalled the feedback as more positive than did participants with low self-esteem. Moreover, this effect was not mediated by participants conscious perceptions of feedback credibility. Research on self-schemata has shown similar effects of people's strongly held self-views. Self-relevant information often captures people's attention (e.g., Cherry, 1953), particularly on dimensions about which people have a great deal of well-articulated self-knowledge (Bargh, 1982; Kihlstrom & Cantor, 1984; Markus, 1977; see also Kuiper & Derry, 1982). Presumably, self-schemata constitute stable, chronically accessible dimensions of the self that influence both self- and social perception (Carpenter, 1988; Dunning & Hayes, 1996; Fong & Markus, 1982; Lewicki, 1983; Markus & Smith, 1981; Markus, Smith, & Moreland, 1985; Markus & Wurf, 1987). People are more likely to recall, accept the validity of, and be confident in their judgments about self-relevant information that is consistent as opposed to inconsistent with their self-schemata. Moreover, people who are schematic on a particular dimension (e.g., height) are quick to categorize and judge others based on this dimension. Similarly, people with high selfesteem tend to exhibit more positive evaluations of others than do people with low self-esteem (e.g., Brown, 1986; Crocker, Thompson, McGraw, & Ingerman, 1987; Epstein & Feist, 1988). Even people's undesired self-views (i.e., things they do not want to believe about themselves) may lead to perceptions of precisely these traits in others (e.g., Newman et al., 1997). People's self-views even appear to influence their memory about what they were like in the past (Conway & Ross, 1984; Ross, 1989; Ross & Conway, 1986; see also Klein & Kunda, 1993; Pratkanis, Eskenazi, & Greenwald, 1994). That is, the content and structure of a persons current self appears to make self-consistent (as opposed to selfinconsistent) experiences and beliefs easier than usual to retrieve. Presumably, this implicit influence of the self on memory enhances the perception of personal consistency and continuity.

< previous page

page_113

next page >

< previous page

page_114

next page >

Page 114 NONCONSCIOUS ELEMENTS OF SELF-EVALUATION: IMPLICIT SELF-REGARD In their influential review, Greenwald and Banaji (1995) summarized research suggesting that self-esteem has a wide variety of implicit influences on a persons perception and behavior, including instructor evaluations, ingroup bias in the minimal intergroup paradigm, denial of disadvantage, and self-serving biases. We would argue that some of these effects reflect implicit (i.e., introspectively unidentified) influences of consciously considered or explicit selfevaluations. Indeed, the relationship between self-esteem and self-serving biases is well established (Blaine & Crocker, 1993; Taylor & Brown, 1988). Although controversial, there is evidence that people high in explicit selfesteem are more likely to exhibit ingroup bias than are those low in self-esteem (Crocker et al., 1987). To the extent that people are unaware of the role that their self-esteem plays in these effects, we agree that effects such as these may well be implicit (i.e., introspectively unidentifiable) effects of explicit self-esteem. We believe that another implicit phenomenon discussed by Greenwald and Banaji may be an even more important implicit aspect of the self. Based on the primacy and automaticity of affect and automatic attitude activation, coupled with the high levels of accessibility of self-relevant information, people will likely develop overlearned evaluative associations (i.e., automatic attitudes) about themselves. These automatic self-relevant associations are likely to reside outside of participants awareness. That is, it is possible to distinguish three different types of effects of self-evaluations on decisions and judgments. The most straightforward is the traditional effect of explicit self-esteem on consciously considered judgments and behaviors. For example, most people seem to realize that when they are feeling bad about themselves, they judge others more negatively than usual. The next two effects appear to be intermixed in Greenwald and Banaji's analysis. The first is the introspectively unidentified (i.e., implicit) effects of explicit self-esteem on judgments or behavior. For example, people probably are not aware of the degree to which their feelings about themselves influence their feelings about the groups to which they belong. The last is the introspectively unidentified effects of people's automatic evaluative associations to the self (i.e., the implicit effects of implicit self-esteem).8 To distinguish the latter effect from the implicit effects of explicit self-esteem (which Greenwald and Banaji also refer to as implicit selfesteem), we refer to these automatic, chronically accessible self-evaluations as implicit self-regard (Hetts, Sakuma, & Pelham, 1999; Pelham & Hetts, 1999b, 1999c). The Functions of Implicit Self-Regard What hypothetical functions might implicit self-regard serve? One way to answer this question might be to ask what functions explicit self-esteem (or the need for positive esteem) is thought to serve, and to ask whether implicit selfesteem might serve similar functions. One perspective – self-affirmation theory – advanced by Steele (1988; see also Steele, Spencer, & Lynch, 1993) argues that positive self-regard helps inoculate people against the stresses, strains, and defeats of daily life (see also Taylor & Brown, 1988). Positive implicit self-regard might thus serve as a kind of shield against such threats to the self. Given the evidence that chronically accessible self-schemata help people reject self-inconsistent information and make sense of ambiguous information, implicit self-regard might serve as a kind of automatically deployed, strategic self-threat defense initiative. That is, positive implicit self-regard might serve to fend off threats to the self before they become the fixation of the conscious self and thus undermine selfevaluations and self-integrity. Another proposed function of positive self-esteem is to protect people against more existential threats. Terrormanagement theory (Pyszczynski, Greenberg, & Solomon, 1997; Solomon, Greenberg, & Pyszczynski, 1991) posits that anxiety surrounding our awareness of our own mortality intensifies our support of culturally shared belief systems. In this formulation, feelings of self-esteem reflect the degree to which we believe we are meeting standards represented by our cultural world-views. Thus, self-esteem buffers us against the fear of death by connecting the self to meanings, values, and institutions that will persevere beyond our individual lifetimes. Interestingly, recent research suggests that terror-management processes operate most powerfully at nonconscious levels (e.g., Arndt, Greenberg, Pyszczynski, & Solomon, 1997; Simon et al., 1997). This suggests that implicit self-regard rather than explicit self-esteem might be a better yardstick by which to measure people's freedom from existential angst. A third perspective holds that the need for positive regard is merely an expression of the more fundamental desire to gain the esteem and approval of others, especially close others and ingroup members (Baumeister & Leary, 1995; Rogers, 1951; Tesser, in press). In this view, positive regard for the individual or social self is a reflection of others evaluations of us (Cooley, 1902/1983). That is, our regard for ourselves serves as a

< previous page

page_114

next page >

< previous page

page_115

next page >

Page 115 marker of the degree to which we are succeeding at gaining social approval (Leary & Downs, 1995; Leary, Tambor, Terdal, & Downs, 1995). We suspect that to the extent this hypothesis is true, implicit self-regard may represent a more direct sociometer, than explicit self-regard because it is probably less prone to distortion or rationalization. Additionally, making implicit self-regard the display for the sociometer suggests a way in which positive self-regard may play a larger role than as a simple reflection of acceptance. Rather, the development of positive implicit selfregard (i.e., chronically accessible positive evaluations of the self) might lead one to interpret others behavior, especially when ambiguous, more positively. Consequently, one might treat those others more positively, thus generating positive behavior in others (through self-fulfilling prophecy effects). This, in turn, should lead to more positive and stable relationships, which should cycle back into the long-term maintenance of positive implicit selfregard (Murray & Holmes, 1993; Murray, Holmes, & Griffin, 1996). A final perspective on the origins of the need for positive regard is provided by attachment theory. Attachment theory (Bowlby, 1969, 1973, 1980) posits that attachment behaviors develop to facilitate caregiving between parents and infants. Attachment researchers theorize that infants develop “internal working models” that reflect both infants self-characterizations and their characterizations of their primary caregivers (see Bartholomew & Horowitz, 1991; Griffin & Bartholomew, 1994). Despite their presumed development during infancy and childhood, attachment representations exhibit substantial stability over the life span (Baldwin & Fehr, 1995; Elicker, Englund, & Sroufe, 1992; Kirkpatrick & Hazan, 1994). Moreover, because implicit belief systems appear to develop earlier than explicit belief systems (e.g., Seger, 1994), a great deal of what is learned about self and others during childhood is likely to be learned implicitly (see also Main, 1991; Reis & Knee, 1996). Indeed, the internal working models of attachment theory, if translated into the realm of nonconscious processes, would substantially resemble implicit self-regard and implicit caregiver regard (for lack of a better term). In fact, the study of attachment appears to be headed in just that direction (e.g., Cline, 1999). However, a second approach to understanding the functions of implicit self-regard might be to speculate what its functions might be from the broad perspective of nonconscious social cognition. That is, a chronically accessible, automatic evaluation of the self might have functions independent from those of explicit self-esteem. For instance, Bargh and Chartrand (1999; Bargh, 1997) might argue that one function of implicit self-regard is to free the conscious mind from continually monitoring the environment for self-relevant stimuli. Rather than having to constantly reevaluate ourselves as additional self-relevant information arrives, people who engage in nonconscious self-regulation might rarely have to devote the entirety of their conscious attention to dealing with threats. Thus, the integrity and positivity of the self could be maintained in ways that leave explicit self-evaluations unaffected. Development and Properties of Implicit Self-Regard If implicit self-regard represents the automatization of our self-evaluations, how might it develop? What are its likely properties? Given the early development of implicit cognition relative to explicit cognition, one might assume that implicit self-regard would begin to develop very early in life, possibly as early as infants are able to distinguish self from nonself. For infants even more than for adults, one might also expect implicit self-regard to operate in a primitive, associationistic fashion. Whereas counterarguing and reinterpretation in the face of negative or disconfirming feedback may influence conscious self-evaluations, we suspect that these strategies are less likely to play a role in participants implicit self-evaluations. For example, when Diane's significant other has a bad day and takes it out on her, we suspect that Jacks explicit self-evaluations will not usually change because he will be able to discount his partners behavior. However, we also suspect that the hurt caused by a critical partner may have longer lasting effects on Jacks implicit self-regard. Indeed, we assume that people automatically process the valence and frequency of many kinds of evaluative feedback (Hasher & Zacks, 1984; Lewicki, 1985). Similarly, we assume that people initially, and more or less automatically, accept incoming information as veridical, especially at the nonconscious level (cf. Gilbert, 1991; Gilbert, Tafarodi, & Malone, 1993). Presumably, these initial judgments are modified only when people are aware that the basis of the judgment is erroneous and possess the cognitive resources required to correct the judgment. However, this conscious correction may leave the nonconscious effects unaltered. Thus, implicit self-regard is likely to reflect the accumulated effects of social feedback, with only minimal benefit of reinterpretation or correction. Second, by extension, we also expect implicit self-regard to be more or less independent of explicit self-evaluations. That is, if explicit self-esteem can benefit

< previous page

page_115

next page >

< previous page

page_116

next page >

Page 116 from conscious repair and reinterpretation, and implicit self-regard must instead suffer the slings and arrows of life, it stands to reason that there will be many instances when the two are no longer in accord with one another. The dissociations among implicit and explicit learning, memory, social inference, and stereotypes reviewed earlier further suggest a similar dissociation between implicit and explicit self-evaluations. Third, implicit self-regard should have independent effects on social judgment and behavior. As people's chronically accessible evaluative associations to the self, implicit self-regard should frequently determine gut-level responses to self-relevant stimuli – responses that may precede their conscious responses to social stimuli. Although people may sometimes choose to modify these initial judgments, social perceivers who lack the motivation or cognitive resources to do so should be particularly likely to fall back on the habitual, automatic responses that are influenced by their implicit self-regard (James, 1890/1950; Posner & Snyder, 1975; see also Bargh, 1999; Bodenhausen, 1990; Bargh & Chartrand, 1999; Devine, 1989; Gilbert, 1989; Petty & Cacioppo, 1986a, 1986b). Thus, implicit self-regard should frequently translate into self-relevant judgments and behavior. This fact may help account for some inconsistencies in the relation between explicit self-esteem and individual and social outcomes (Baumeister, 1998; Wylie, 1979). Finally, we expect that implicit self-regard should be both durable and malleable. That is, we posit that implicit selfregard is formed by associations to the self and built up from numerous instances of self-relevant experience or feedback. Consequently, one would expect that it would take a great deal of contradictory feedback, across a variety of situations, to change a persons chronic level of implicit self-regard. However, this does not mean that implicit selfregard will remain highly stable from moment to moment. The particular set of self-relevant associations that are most accessible from moment to moment is likely to vary from situation to situation, and thus implicit self-regard should vary notably across social situations. However, this change should be firmly anchored around a core set of well-established self-relevant associations. To use an analogy, implicit self-regard has characteristics that resemble both climate and weather. The climate of a region typically reveals a predictable and stable pattern of temperature and precipitation. However, the weather in a particular locale can vary dramatically. Similarly, implicit self-regard should change only gradually but should also show substantial variability from day to day and situation to situation. Notice that this metaphor suggests that it may be difficult to get a clear picture of a persons chronic level of implicit self-regard without measuring the persons beliefs on more than one occasion. Existing Evidence for Implicit Self-Regard One simple source of evidence for automatic self-evaluation comes from the phenomenological experience of the self. As James (1890/1950) noted, “there is a certain average tone of self-feeling which each one of us carries about with him, and which is independent of the objective reasons we may have for satisfaction and discontent” (p.306). James also noted that a positive feeling and emotional attachment accompanies everything associated with the self – from relatives to material possessions. James argued that this attachment arose, at least in part, from its association to the self. Not surprisingly, recent research (see Greenwald & Banaji, 1995) provides a second source of evidence supporting the presence of automatic evaluation of self-related objects. Research on the name-letter, mere ownership, and endowment effects indicates that people evaluate stimuli or objects associated with the self more positively than comparable objects not associated with the self. This effect has been demonstrated for letters in people's names (Hoorens & Nuttin, 1993; Nuttin, 1985, 1987), numbers in their birthdays (Kitayama & Karasawa, 1997), computer icons associated with the self (Feys, 1991), and a variety of physical objects (e.g., pens, mugs) that belong to the self (Beggan, 1992; Belk, 1988; Kahneman, Knetsch, & Thaler, 1990). Second, Paulhus and colleagues demonstrated that manipulations such as arousal and cognitive load increase the favorability of people's self-evaluations (e.g., Paulhus, 1990; Paulhus, Graf, & Van Selst, 1989; Paulhus & Levitt, 1987; Paulhus & Lim, 1994). This suggests that, for most people, evaluating oneself positively is a dominant, automatic tendency. In contrast to Paulhus view, Swann and colleagues (Swann, Griffin, Predmore, & Gaines, 1987; Swann, Hixon, Stein-Seroussi, & Gilbert, 1990; see also Shrauger, 1975) suggested that the automatic, overlearned desire for positive feedback will sometimes be overridden by a desire for self-verification. Nonetheless, Swann and Paulhus agreed that, for most people, the desire to self-enhance or feel good about oneself is highly automatic. A third major theoretical perspective that suggests the potential existence of implicit self-regard, as discussed previously, is attachment theory. According to attachment theory, attachment behaviors develop to facilitate parental caregiving. However, caregivers vary notably in how responsive they are to their infants needs, and Bowlby theorized that infants develop ‘‘internal

< previous page

page_116

next page >

< previous page

page_117

next page >

Page 117 working models” that reflect their caregivers responsiveness. These models reflect both the infants selfcharacterizations and their characterizations of their principle caregivers (see Bartholomew & Horowitz, 1991; Griffin & Bartholomew, 1994), Because implicit belief systems appear to develop earlier than explicit belief systems (e.g., Seger, 1994), a great deal of what is learned about self and others during childhood is likely to be learned implicitly (see also Main, 1991; Reis & Knee, 1996). Epstein's (1994, 1998) research on cognitive-experiential self-theory (CEST) probably provides some of the clearest support for the existence of an implicit self-system. In his integrative model of the self-concept, Epstein posited that people's self-evaluations arise from two distinct conceptual systems. The first, the rational system, is consciously controlled, deliberative, and highly effortful. In contrast, the experiential system is simpler, more intuitive, and operates automatically. Thus, Epstein's conception of the experiential self-system is quite similar to our conception of implicit self-regard. Moreover, his careful organization of diverse research domains suggests the existence and importance of implicit self-regard in understanding the self, emotions, cognitions, and behavior. However, a full exposition of cognitive-experiential self-theory has been held up by the same stumbling block as Greenwald and Banaji's analysis of implicit social cognition – the lack of an effective procedure for assessing implicit self-regard or (in Epstein's case), experiential beliefs about the self.9 Current Research on Implicit Self-Regard As suggested by the research reviewed thus far, a measure of implicit self-regard is clearly needed (Greenwald & Banaji, 1995). To that end, we have developed two primary methods to assess implicit self-regard (Hetts et al., 1999; Pelham & Hetts, 1999c). The first method is an adaptation of the method used to measure the automatic activation of attitudes. Research on the automatic attitude activation effect has shown that the presentation of a prime (e.g., chocolate ) facilitates responses to subsequent targets of similar valence (e.g., wonderful ) and disrupts responses to subsequent targets of dissimilar valence (e.g., disgusting). Based on this research, we reasoned that we could use people's patterns of responses to valenced targets to determine their evaluation of primes that preceded them (for similar analyses at the individual and the group level, see Bylsma et al., 1991; Perdue, Gurtman, Dovidio, & Tyler, 1990). Thus, to measure participants implicit self-regard, we assessed their latencies for identifying clearly valenced targets (e.g., good and bad) after being primed with a self-relevant pronoun (e.g., me). When participants are repeatedly exposed to self-relevant primes, it is possible to compute a simple measure of implicit self-regard by subtracting participants response latencies for positive target words from their response latencies for negative target words. Of course, it is also possible to assess other beliefs in this way. For example, to assess implicit group-regard, one needs merely to replace self-relevant primes with group-relevant primes (e.g., we ). Similarly, as a simple control procedure, one might use neutral, self-irrelevant primes (e.g., it, that ). Other researchers have successfully adapted this procedure (e.g., see Spalding & Hardin, in press), whereas still others have developed alternate approaches that build on the same underlying logic (e.g., Farnham & Greenwald, 1998; Farnham, Greenwald, & Banaji, 1999; Greenwald et al., 1998). To construct measures of implicit self-regard and group-regard that do not require collecting response latency data, we adapted word fragment-completion tasks from implicit memory research (e.g., Tulving et al., 1982). In the case of these measures, we activate personal or social identities by asking people to respond to nonevaluative self- or group-relevant statements (e.g., “I am very sensitive to my inner thoughts and feelings”). Immediately after exposure to an identity-relevant prime, participants generate three separate completions (in the order they come to mind) for a specific word fragment. For example, for the fragments (1. _ICE 2._ICE 3. _ICE), we code for the serial position of the positive word NICE. For the fragments (1._OOR 2._OOR 3. _OOR), we code for the serial position of the negative word POOR . By subtracting the average serial position of the positive word completions from the average serial position of the negative word completions, we generate an implicit self-evaluation score that is logically analogous and theoretically comparable to the response latency measure discussed previously. We have found that both of these measures of implicit self-regard have many of the theoretically predictable properties we have identified in this review. We provide a couple of illustrative examples next. Implicit Self-Regard and Culture . If implicit self-regard represents associations to the self that accumulate over ones lifetime, then one of its primary determinants should be the socialization practices of the culture in which a person is reared. Research on the relation between culture and self-concept suggests that the members of Eastern and Western cultures understand and evaluate the self quite differently (Fiske, Kitayama, Markus, & Nisbett, 1998; Markus & Kitayama, 1991; Triandis, 1989, 1995). Specifically, this research suggests that Western cultures encourage an individualistic

< previous page

page_117

next page >

< previous page

page_118

next page >

Page 118 approach to the self. Along these lines, the typical members of most Western cultures are encouraged to create and affirm positive individual identities. In contrast, the typical members of most Eastern cultures are encouraged to create and affirm positive collective identities. That is, Eastern cultures predispose people to focus and obtain validation from their social rather than their individual identities. These differences in the understanding and construction of the self have important implications for implicit selfregard. For inhabitants of independent cultures, nonconscious associations to the individual self (i.e., implicit selfregard) should be predominantly positive. However, for inhabitants of interdependent cultures, nonconscious associations to the individual self may be much less favorable. Given our hypotheses about the stability of implicit self-regard, we might expect that when people immigrate from an interdependent to an independent culture their implicit associations about their personal and social identities might continue to bear the influence of their collectivistic upbringing. In contrast, such people should be able to quickly correct or adapt their explicit self-evaluations to the normative demands of the new culture. This is precisely what we found in three studies of participants who differed in their degree of exposure to Eastern and Western cultures (Hetts et al., 1999). In Studies 1 and 2, recent Asian immigrants to the United States exhibited significantly lower implicit self-regard than both European and Asian Americans, despite being similar to these groups when it came to their explicit self-evaluations. Moreover, the longer the recent immigrants had lived in the United States, the more positive their implicit self-regard became ( r =.46). In contrast, participants explicit self-evaluations were equally favorable regardless of how long they had been living in the United States. In these same two studies, we also found that, when it came to measures of implicit group-regard, the recent Eastern immigrants to the United States had somewhat more favorable implicit self-conceptions than their more fully Westernized counterparts. Presumably, the focus on group harmony and group identity that is common in many Eastern cultures leads the members of such cultures to develop favorable implicit associations about their social identities. Consistent with our assumption that implicit beliefs change slowly over time, we found that the longer immigrants had lived in the United States, the lower their implicit group-regard became ( r = -;.65). For these participants, learning to toot their own horns went hand in hand with feeling worse about the other members of their band. Hetts et al. (1999) also examined the implicit and explicit self-evaluations of a group of Japanese university students. Some of these students had substantial exposure to an independent culture (they had lived for 5 or more years in the United States or Canada). Others had never lived outside of Japan. Again, differences in participants cultural experiences influenced their implicit self- and group-regard but not their explicit self-evaluations. Specifically, participants who had never lived in a Western culture possessed lower implicit self-regard and higher implicit groupregard than did participants who had spent substantial time overseas. Again, we observed little differences in participants explicit self-evaluations. Taken together, these studies have a number of important implications for self-evaluation. First, they demonstrate the ability of implicit self-evaluation measures to identify differences in people's experiences that may not always be revealed by explicit measures. That is, a complete reliance on explicit measures of self-evaluation would have suggested that the effects of culture on self-evaluation are short-lived and can be eliminated very quickly when people immigrate to a new culture. These studies also demonstrate the relative stability of implicit self-evaluation (and, by extension, the long-lasting influences of culture on the self). Third, these studies reveal an important sense in which the typical members of Eastern cultures possess highly favorable associations about the self. Specifically, the higher levels of implicit group-regard observed among participants with greater degrees of exposure to Eastern cultures suggests that the need for positive regard (Pelham & Hetts, 1999b) may be common to both Eastern and Western cultures. However, this need may be expressed in very different ways in different cultures – through the individual self in independent cultures and through the social self in interdependent cultures.10 Implicit Self-Regard and Ethnicity. Another variable that should have an important impact on people's implicit selfevaluations is ethnicity. Numerous studies have shown that, relative to Euro-Americans, people of color face higher levels of prejudice and discrimination, are subject to disparate treatment by the criminal justice system, endure systematic economic disadvantages, and enjoy less social and professional prestige (e.g., Braddock & McPartland, 1987; Crosby, Bromley, & Saxe, 1980; Sidanius & Pratto, 1999). However, a great deal of research has indicated that people of color are able to marshal a number of strategies in defense of favorable self-conceptions. In fact, many studies have revealed that the self-esteem of people of color is at least as high as, if not higher than, that of Euro-Americans (Crocker & Major, 1989; Cross, 1985; Porter & Washington, 1979,

< previous page

page_118

next page >

< previous page

page_119

next page >

Page 119 1989, 1993; Rosenberg, 1979). At the same time, however, research has found that members of stigmatized or low-status groups sometimes display subtle effects of their poor treatment by others. For example, relative to high-status group members, lowstatus group members often engage in outgroup favoritism (e.g., Allport, 1954/1979; Hinkle & Brown, 1990; Jost, 1995; Jost & Banaji, 1994; Sidanius, 1993; Sidanius & Pratto, 1993). More specifically, a variety of studies have replicated Clark and Clark's (1947) observations of outgroup preference and negative self-stereotyping in African-American children ( inter alia Fine & Bowers, 1984; Gopaul-McNicol, 1995; Landreth & Johnson, 1953; Powell-Hopson & Hopson; 1988; Sagar & Schofield, 1980; Spencer, 1984; Williams & Morland, 1979; for similar results for Hispanic children, see Bernat & Balch, 1979, 1981; Teplin, 1976). Interestingly, the studies that have demonstrated outgroup preference or negative self-stereotyping have usually used indirect measures of self-evaluation (Simmons, 1978). Further, they appear to be largely focused on evaluations of the ethnic group rather than the individual self. In contrast, studies documenting that stigmatized group members possess positive self-conceptions have typically made use of direct or objective measures, most commonly survey measures of self- and group evaluation. A few studies have shown that African-American children possess positive explicitly reported self-concepts but demonstrate outgroup preferences toward Euro-Americans when it comes to projective tasks (e.g., Clark, 1982; Spencer, 1984) These findings are consistent with our conception of implicit self-regard as a relatively associationistic belief system. Thus, although people of color can successfully deploy a number of strategies to protect their explicit selfevaluations (Crocker & Major, 1989), their implicit self-evaluations, especially their implicit group-regard (the target of the stereotypes), may still reflect the adverse effects of socially shared stereotypes. We (Hetts & Pelham, 1999) have conducted preliminary investigations to examine this possibility. In keeping with previous research, we found that the explicit self-evaluations of people of color did not differ noticeably from those of Euro-Americans. In contrast, we found that people of color exhibit lower levels of implicit group-regard but higher levels of implicit self-regard than Euro-Americans (see also Hunter & Mackie, 1996; Steele, 1997). This is consistent both with our model of implicit self-evaluation and some otherwise puzzling findings in the literature on stereotyping and stigma. For example, it is consistent with the finding that the members of many stigmatized groups (e.g., women, people of color) often report that, whereas their fellow ingroup members are frequently the targets of discrimination, they themselves rarely feel the brunt of discrimination (e.g., Crosby, 1984; Ruggiero & Taylor, 1995; Taylor, Wright, Moghaddam, & Lalonde, 1990; Taylor, Wright, & Porter, 1994). Although generally supporting the finding that stigmatized group members often develop high self-esteem, these findings also suggest that defending ones social identity against stigmata and stereotypes may not be as straightforward as previously thought. Implicit Self-Regard and the Ghosts of Christmas Past . If implicit self-regard does in fact represent the accumulation of positive and negative self-relevant experiences over ones lifetime, then one might expect people who have received less special attention over their lifetimes to have lower implicit self-regard. For the most part, people in the United States have 1 day a year – their birthday – in which they are praised and celebrated for merely existing. In contrast to most other people, however, the unfortunate person who is born in close proximity to Christmas day may fail to receive the attention that normally accompanies ones birthday. That is, being born in close proximity to a holiday that so fully captivates the attention of most Americans virtually guarantees that people born at this time of year will be the victim of oversights and omissions – on a day in which they might otherwise expect to be praised and adored. For this reason, one might expect people born at this time of year to have lower implicit self-esteem than people born at other times (see Cooley, 1902/1983; Mead, 1934). However, because most people are able to understand the circumstance leading others to neglect them on their birthdays, one might expect that the effects of being born near Christmas day would not be as extreme for participants explicit self-evaluations. This is precisely what we found (Pelham & Hetts, 1999a). Participants born within a week of Christmas day had significantly lower implicit self-regard than people in general and than two control groups from early December and early January (to explore the possibility of time of year effects). In contrast, there were no differences between the groups in implicit group-regard or explicit self-esteem. In a second study, we showed these implicit self-esteem effects were not simply mood effects in disguise. In a third study (using an indirect, behavioral rather than a strictly implicit measure of self-esteem), we showed that these effects occurred for Christians but not for non-Christians (Jews, Muslims, or atheists). Construct Validity of Implicit Self-Regard. In our initial enthusiasm over our findings, we may have failed to pay sufficient attention to the less exciting task of gathering psychometric data on the basic properties of

< previous page

page_119

next page >

< previous page

page_120

next page >

Page 120 our implicit measures. However, evidence for the validity of the implicit self-regard construct is beginning to accumulate. For example, we have found implicit self-regard to be generally distinct from explicit self-esteem, collective self-esteem, explicit dimensions of specific self-evaluations, mood, and attributions (Hetts, 1999; Hetts et al., 1999; Hetts, Shimojo, & Hardin, 1999; Pelham & Hetts, 1999a, 1999c). At the same time, to the degree we have observed significant relations between implicit and explicit measures of self-evaluation, they have been highly predictable. For example, implicit self-regard is more highly correlated with explicit measures of individual selfesteem than with explicit measures of collective self-esteem. Similarly, implicit group-regard is more highly correlated with explicit measures of collective as opposed to individual self-esteem. Farnham and colleagues (Farnham & Greenwald, 1998; Farnham et al., 1999) found a similar pattern of relationships using an alternative measure of implicit self-regard based on the Implicit Association Test (IAT; Greenwald et al., 1998). Furthermore, in two initial examinations of the testretest reliability over a long delay (5 weeks), the testretest correlations for implicit self-regard were r (164) = .48, p

< previous page

page_121

next page >

Page 121 self-regard circumvent certain self-presentational concerns, we think it would be unwise to grant privileged status to measures of implicit self-regard as indicators of people's true beliefs about themselves. Although subject to certain biases, people's explicit self-views often serve as powerful predictors of people's judgment and behavior. They should not be cast aside when measures of implicit self-regard have yet to be linked conclusively to a wide range of judgments or behavior. We believe that reaching a full understanding of the self will likely require a careful examination of both the unique and joint contributions of implicit and explicit self-evaluations to social judgment and behavior. Along these lines, it seems likely that explicit self-evaluations are closely associated with certain kinds of carefully considered judgments (e.g., people's explicit evaluations of their close relationship partners; e.g., Murray & Holmes, 1997; Murray, Holmes, & Griffin, 1996). In contrast, implicit self-evaluations may be more closely associated with many nonverbal behaviors. Because many nonverbal behaviors are extremely difficult to regulate through conscious efforts (e.g., DePaulo, Lanier, & Davis, 1983; DePaulo, Lassiter, & Stone, 1982; Ekman & Friesen, 1969; Ekman, Friesen, & O'Sullivan, 1988; Zuckerman, Spiegel, DePaulo, & Rosenthal, 1982), these behaviors may be more intimately linked to implicit than to explicit self-evaluation. Spalding and Hardin (in press) recently observed support for exactly this position. They found that the degree to which people behaved nervously during a self-threatening interview was strongly related to people's level of implicit self-regard, but not to their level of explicit self-esteem (see Fazio et al., 1995, for similar findings with respect to automatic prejudice). In contrast, participants explicit selfevaluations predicted their explicit judgments of the quality of the interview. The Relation Between Implicit and Explicit Self-Evaluation Although researchers are beginning to document potential correlates and consequences of implicit self-regard (e.g., Bosson & Swann, 1999; Farnham & Greenwald, 1998; Farnham et al., 1999; Hetts et al., 1999; Pelham & Hetts, 1999a, 1999b, 1999c; Spalding & Hardin, in press), the lessons of implicit memory and learning research suggest that further efforts to differentiate implicit and explicit self-evaluation are still warranted. For example, it is possible that participants explicit self-evaluations sometimes have long-term effects on their level of implicit self-regard (e.g., when a person makes a carefully considered decision to spend more time with an emotionally supportive friend). It seems equally possible that implicit self-regard may sometimes influence participants explicit self-evaluations. Despite the unique connections between implicit and explicit self-evaluation and social experiences documented to date, further attempts to separate these two kinds of self-evaluation (e.g., by adapting Jacobys “process dissociation” technique) may help us better understand the origins, functions, and consequences of these alternate forms of self-evaluation. Second, despite our expectation of a dissociation between implicit and explicit self-evaluations, we also recognize that at least some experiences may have the same effects on both implicit and explicit self-evaluation. Thus, another important task in the development of the construct of implicit self-regard is to determine when it is and is not related to explicit self-evaluations. For example, a number of potential moderators of the relation appear worth pursuing. If one assumes that explicit measures of self-evaluation represent a mixture of implicit and explicit processes, than participants explicit self-evaluations should reflect different mixtures of implicit and explicit self-evaluation depending on whether they have the motivation or resources to think carefully about the self. Thus, manipulating variables such as performance, self-awareness, arousal, or cognitive load prior to assessment of self-evaluations may provide an opportunity to more carefully specify the relation between implicit and explicit self-evaluation. Indeed, the works of Swann et al. (1990) and Paulhus et al. (1989) provide initial evidence that variables such as cognitive load or arousal may in fact moderate the relation between implicit and explicit self-evaluation. Likewise, if social desirability or self-presentation interferes with the measurement of true attitudes about the self at the explicit level, then one might expect to observe stronger than average relations between implicit and explicit self-evaluations under conditions of anonymous rather than public judgment. Along somewhat different lines, the precise origins of people's beliefs may influence the degree of overlap between implicit and explicit self-evaluations. More specifically, when explicit attitudes are based on direct experience rather than speculation, second-hand opinion, or reconstruction, we might expect to see a good deal of overlap between implicit and explicit attitudes (self-relevant or otherwise). That is, explicit attitudes forged through direct experience are especially stable, especially resistant to change, and especially likely to predict behavior (e.g., Fazio & Zanna, 1981; Fazio, Zanna, & Cooper, 1978; Regan & Fazio, 1977; Zanna, Olson, & Fazio, 1981). Consequently, one might expect that

< previous page

page_121

next page >

< previous page

page_122

next page >

Page 122 explicit attitudes might have these properties precisely because they are in synch with implicit attitudes that have grown out of the very same experiences. Thus, in the case of self-evaluations that have grown out of a great deal of consistent and consensual validated social experience (e.g., Cal Ripken, Jr.'s self-evaluations on the dimension of durability), implicit and explicit self-evaluations might be closely related. CONCLUDING REMARKS In this chapter, we have reviewed a wide variety of evidence attesting to the potential importance of nonconscious elements of self-concept and self-evaluation. Moreover, we have provided preliminary evidence that implicit selfevaluations can be measured, that they exhibit surprising stability, and that they are systematically related to aspects of human behavior that are not predicted by explicit self-evaluations. If we assume that people do possess both implicit and explicit self-evaluations, this might raise the question of why these two separate systems of selfevaluation exist in the first place. Our suspicion is that implicit self-regard exists for the same reason that many other implicit belief systems exist. Like learning to comprehend and produce language, learning to engage in the myriad forms of self-regulation that are necessary for survival in a social world requires an enormous investment of time and energy. If these complex skills had to be acquired solely based on carefully considered reasoning, or if they never became automatized, many complex aspects of human social experience would be severely compromised. The more we can self-regulate without devoting cognitive resources to the effort, the more capacity we free up to deal with other pressing problems of survival. From this perspective, implicit self-regard may represent an important and highly adaptive piece of the puzzle of human adaptation to a complex social world. ENDNOTES 1. One omission in this analysis is a careful delineation of what precisely the resources of the conscious mind are being saved for. Bargh and Chartrand hint that one answer might be decision making or self-regulation, as suggested by the recent work of Baumeister, Bratslavsky, Muraven, and Tice (1998; Muraven, Tice, & Baumeister, 1998). Three other possibilities seem plausible. Conscious resources may be conserved for instances in which automatized thoughts and behaviors will lead to mistakes – for example – in novel situations or conflicts between nonconscious processes (e.g., Cart, 1925). Another possibility is that conscious resources may be conserved for the strategic direction of nonconscious processes. That is, the conscious mind may be a commander in chief: It attempts to direct the entire force of the organism through the manipulation of processes that reflect a mixture of conscious and nonconscious processes (high-ranking officers), which subsequently guide processes reflecting even less conscious processing, until one reaches the nearly automatic processes (sergeants) and the most highly automatized processes (privates). Finally, if Bargh (1997) is correct that conscious processes are necessary for the development of preconscious processes, it may be that conscious resources may be conserved not for themselves, but to hasten the development of additional nonconscious processes to create an increasingly efficient mind. 2. Implicit memory is contrasted with explicit memory, which involves the conscious recollection or recognition of previously encountered stimuli. Other researchers have employed different terms to describe a very similar distinction, including (a) direct and indirect memory tasks (Richardson-Klavehn & Bjork, 1988), (b) memory with and without awareness (Jacoby & Witherspoon, 1982; cf. Roediger, 1990), and (c) memory as a tool and memory as an object (Jacoby & Kelley, 1987). 3. Richardson-Klavehn and colleagues (Richardson-Klavehn & Gardiner, 1995; Richardson-Klavehn, Gardiner, & Java, 1996; Richardson-Klavehn, Lee, Joubran, & Bjork, 1994) took a complementary but distinct approach. They divided the implicit-explicit distinction into two dimensions – the intentionality of memory (whether the participant is trying to remember items) and the awareness of recall. They did so to highlight the possibility of conscious involuntary memory (when a stimuli springs to mind in full awareness without effort) and voluntary unconscious memory (when trying to remember previously studied stimuli influence responding despite the phenomenal experience of inability to retrieve the stimulus). 4. Implicit learning has been similarly demonstrated in the control of complex systems (e.g., Berry & Broadbent, 1984; Stanley, Mathews, Buss, & Kotler-Cope, 1989), such as the regulation of production in a fictional sugar factory. Such studies have shown that: (a) practice increases performance but not the ability to explain ones performance, (b) the ability to correctly manipulate a complex system comes before explicit knowledge of that system, and (c) performance and explicit knowledge tend to be unrelated (see also Bright & Burton, 1994; Mc-

< previous page

page_122

next page >

< previous page

page_123

next page >

Page 123 George & Burton, 1990; Willingham, Nissen, & Bullemer, 1989). 5. A great deal of additional work has been done examining different sources of temporary and chronic accessibility and the interplay between the two types of accessibility that necessarily lie outside the scope of this manuscript (see Higgins, 1996; Higgins & Bargh, 1987, for reviews). 6. As with construct accessibility, a great deal of additional work has been done examining the parameters of spontaneous trait inference and employing a variety of indirect (i.e., nonconscious) dependent variables that cannot be covered herein (for reviews, see Newman & Uleman, 1989; Uleman, 1999; Uleman, Newman, & Moskowitz, 1996). 7. For evidence that the structure may not exist, or that it may exist solely for negative self-attributes, see Higgins, van Hook, and Dorfman (1988). 8. With some imagination, one can imagine the possibility of a fourth effect: the effects of our automatic evaluations of our selves on our judgments and behaviors that we are aware of (i.e., the explicit effects of implicit self-regard). Although we suspect that the awareness of the influence of our automatic evaluations of ourselves is rare, we nonetheless believe that it does occur. For example, the point of some psychotherapy techniques is precisely to achieve such a state (i.e., to make people aware of the chronic, negative evaluations of themselves that are exhibited in some form of nonoptimal behavior). 9. In our opinion, the Basic Beliefs Inventory (Catlin & Epstein, 1992) and the Faith in Intuition scale (Epstein, Pacini, Denes-Raj, & Heier, 1996) are likely too explicit to satisfactorily measure nonconscious aspects of self-evaluation. 10. One additional implication is that explicit self-evaluations may be heavily influenced by the normative demands of the situation or culture. That is, they may not always reflect participants self-concepts, but rather what people believe their self-concepts should be. Thus, cross-cultural comparisons relying exclusively on explicit self-evaluations may tend to exaggerate existing differences in the self-concept while concealing underlying variation within cultures.

< previous page

page_123

next page >

< previous page

page_124

next page >

page_124

next page >

Page 124 This page is intentionally left blank.

< previous page

< previous page

page_125

next page >

Page 125 8 The Role of Theories in Mental Representations and Their Use in Social Perception: A Theory-Based Approach to Significant-Other Representations and Transference Serena Chen University of Michigan, Ann Arbor Social cognition researchers have made substantial progress in elucidating the nature of mental representations and their role in social perception. A diverse array of perspectives on representational content and structure exists (for reviews, see Carlston & E.R. Smith, 1996; E.R. Smith, 1998; Wyer & Carlston, 1994), along with a rich understanding of how and when representations are likely to be activated and brought to bear on the tasks of perceiving and judging the social world (e.g., Higgins, 1996; Wyer & Srull, 1986). Considerable progress has also been made regarding the role of mental representations in guiding social behavior (e.g., Bargh, 1990, 1997). Where do we go from here? This chapter examines the content, structure, and use of mental representations. It focuses on the idea, long recognized on a theoretical level yet often absent from empirical work, that perceivers' stored mental representations of the social world may include “theories” – that is, explanatory or causal forms of knowledge. If we have not been examining such knowledge all along, then what have we been examining? Even a quick mental survey would reveal that in much social-cognitive research, representations are treated as if they were composed, essentially, of lists of features or attributes. Further, their activation and use are thought to depend, in part, on the fit or overlap between the features of a stimulus and features associated with perceivers' representations (Bruner, 1957a; Higgins, 1996; see also Bargh, 1997; Hardin & Rothman, 1997). For instance, the activation and use of a racial stereotype to interpret a stimulus person are often thought to occur by virtue of the fit between racial features of the person and ones stored as part of a perceiver's stereotype representation. Or, observing a stimulus person engage in a helpful behavior, such as helping an elderly person across a street, is thought to elicit the use of a perceiver's representation of the trait helpful due to the overlap between a feature of the stimulus and one stored as part of the perceiver's trait representation. In recent years, however, challenges to such a feature-based approach have arisen in the social-cognitive literature (e.g., Wittenbrink, Gist, & Hilton, 1997), paralleling trends in cognitive and developmental work (e.g., Murphy & Medin, 1985). These challenges argue that theories are basic to mental representations and their use. Unlike features, which are often treated as unrelated units of knowledge, theories are seen as embodied in explanatory relations linking bits of knowledge about an entity. To illustrate the potential value of theory-based views, this chapter presents a theory-based approach to mental representations of significant others. It is an approach that brings together theory-based ideas in the cognitive, developmental, and social-cognitive domains with the growing body of evidence for the role of significant-other representations in interpersonal perception and relations. The role that representations of significant others play in interpersonal perception and relations reflects transference, originally a clinical notion (Freud, 1912/1958; Sullivan, 1953), but one that also has been conceptualized in socialcognitive terms (Andersen & Glassman, 1996). In these terms, transference refers to the phenomenon whereby something about a new person activates a perceiver's representation of a significant other, leading him or her to interpret and respond to the person in ways derived from prior experiences with the significant other. Applying theory-based ideas to research on significant-other representations and transference, the main thrust of the theorybased approach proposed in this chapter is that perceivers are especially likely to have theories stored about their significant others, which implies that the content and structure of significant-other representations involve such theories, and that these theories play an important role in the transference

< previous page

page_125

next page >

< previous page

page_126

next page >

Page 126 phenomenon. A BRIEF HISTORY The content, structure, and use of mental representations are long-standing topics in social cognition (e.g., Lingle, Altom, & Medin, 1984; Wyer & Gordon, 1984) and ones that continue to be debated (e.g., E.R. Smith, 1998). Because many of the issues surrounding them relate to ones discussed in the cognitive literature (e.g., Keil, 1989; Rosch & Mervis, 1975; E.E. Smith & Medin, 1981; Tversky, 1977), a brief history of relevant cognitive work is presented first, followed by a closer look at relevant social-cognitive work. Features as the Building Blocks: Cognitive Views The general assumption that features are the primary building blocks of representations has a long history in cognitive views on categories and categorization (for reviews, see Komatsu, 1992; Medin, 1989; Rosch, 1978; E.E. Smith & Medin, 1981). The classical view is often discussed as the starting point of this history (e.g., E.E. Smith & Medin, 1981). This view maintains that categories are defined by a set of necessary and sufficient features. For example, the features orange and round might be a subset of the features that define one's ‘‘basketball” category. According to this view, category membership is all-or-none. If an object possesses all of the necessary and sufficient features of a category, it is a category member; otherwise, it is not. The most immediate criticisms of the classical view took the form of prototype or family resemblance approaches (e.g., Rosch, 1975, 1978; Tversky, 1977; Wittgenstein, 1953), which contend that categories are organized around a prototypical member whose features reflect the central tendency of the category or summary abstracted from various experiences with category examples. In such approaches, category membership is seen as probabilistic or “fuzzy”, rather than all-or-none. Objects are members if they possess some, but not necessarily all, of the features associated with the category. Greater featural similarity or overlap with the prototype implies being a more prototypical category member or, put another way, having greater family resemblance. Although classical and prototype perspectives differ fundamentally in their definitions of categories and their views on categorization, they nonetheless share the general view that features are the main building blocks of categories. Prototype approaches have been challenged by exemplar ones, which argue, in their strongest form, that representations of single instances, called exemplars, are stored in memory rather than prototypes abstracted from experience with various instances (e.g., Medin & Schaffer, 1976; E.E. Smith & Medin, 1981). From an exemplar perspective, then, decisions about category membership depend on the similarity or overlap between the features of an object and those associated with stored exemplars, rather than with a prototype. Although prototype and exemplar views differ in several other basic ways, these differences are beyond the present scope (for recent discussions, see Fiske & Taylor, 1991; Komatsu, 1992; E.R. Smith, 1998). Of key relevance is the fact that, by and large, features have been treated as the basic building blocks of categories in exemplar approaches, just as in classical and prototype ones. Although assumptions about representational structure have tended to be less explicit than content ones, some inferences might be drawn about the former from the latter. In particular, feature-based approaches have often treated the featural content of representations as unrelated units of knowledge, which is a view that arguably suggests little or rather simple structure, at least in terms of structure that lies in linkages between features (see Medin, 1989). However, the notion of correlated features has long been discussed (e.g., Chapman & Chapman, 1969; Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976) and is one that arguably suggests a greater degree of structure. Nonetheless, neither the principles that account for why features correlations may exist, nor the nature of the linkages between features that such correlations imply, have been spelled out with much precision (for a review, see Murphy & Medin, 1985). Relative to assumptions about structure, feature-based views on the use of representations have tended to be more explicit. Specifically, feature-based ideas about content and structure have often come hand in hand with the view that categorization, or the likelihood that a representation is activated and used to interpret a stimulus, depends on the number of matching and mismatching features between a stimulus and a stored category (e.g., Tversky, 1977). The greater the number of matching features and the fewer the number of mismatching ones, the higher the likelihood that the stimulus is classified as a category member. Said differently, categorization has often been conceptualized as reflecting some function of feature-based similarity between a stimulus and a category (for reviews, see Goldstone, 1994; Medin, Goldstone, & Gentner, 1993; see also Bruner, 1957a). Over the last decade or so, several lines of cognitive and developmental research have offered alternatives to feature-based approaches. In broad strokes, these alternatives contend that the building blocks of representations

< previous page

page_126

next page >

< previous page

page_127

next page >

Page 127 may extend beyond features. Before considering this work, featural assumptions in social-cognitive views on mental representations and their use are examined. Features as the Building Blocks: Social-Cognitive Views Featural assumptions have been pervasive not only in social-cognitive work on mental representations and their use (e.g., Bruner, 1957b; Higgins, 1989, 1996; Lingle et al., 1984; Sedikides & Skowronski, 1991; Wyer & Srull, 1986), but also in social-cognitive views of personality (e.g., Cantor & Kihlstrom, 1987, 1989). In these social domains, features have been widely defined as trait attributes (see Zuroff, 1986). For example, representations of professors might be treated as if they include educated and absent-minded as features, whereas representations of Asians might include shy and hardworking . Although it is likely that most social cognition researchers would not, on a theoretical level, endorse simple feature-based definitions of representational content and structure, featural assumptions are often built right into the research methodologies used. For example, the stereotyping literature is composed largely of research in which stereotypic features were presented as stimuli, and participants' stereotypes were assessed in feature-based terms. Nonetheless, tracing the history of social-cognitive work on mental representations and their use reveals that, although features (i.e., traits) have been widely treated, especially on an empirical level, as the basic units of perceivers' conceptions of others, there have been hints of forms of knowledge extending beyond such features from the start. Elemental and Holistic Approaches . The beginnings of social-cognitive work on mental representations might be traced back to work on impression formation. Early on in this work, an important distinction was drawn between elemental and holistic approaches (e.g., Fiske & Taylor, 1991; Leyens & Fiske, 1994). Elemental approaches emphasize the elements or pieces of information that presumably underlie perceivers' impressions of others (e.g., N.H. Anderson, 1974). These elements are treated as separate entities, ones that generally do not exert any influence on one another. In this regard, elemental approaches are fairly akin to feature-based ones in the cognitive literature, discussed earlier, that view the content and structure of categories in terms of lists of features and that tend to treat the features associated with a given category as independent entities. In contrast, the essence of holistic approaches is that perceivers' impressions of others should be viewed in terms of wholes rather than separate elements. Said differently, impressions reflect coherent configurations of individual elements. Moreover, unlike elemental approaches, holistic ones emphasize the subjective nature of perceivers' construction of reality, thus arguing that perceivers subjectively construe coherent meanings among separate elements. Gestalt psychology, which reflects this holistic view, was a major influence in early social-cognitive work. From a Gestalt perspective, individual elements (e.g., traits) are thought to influence one another, implying that relations exist among these elements such that each element only has meaning in the context of others. Asch's (1946) classic experiments clearly demonstrated the major propositions of the holistic approach, as did subsequent work in the same tradition (Asch & Zukier, 1984; see also Heider, 1958). For example, emerging from Asch's early work was the concept of central traits , which refer to traits that provide organizing themes around which the individual elements (i.e., traits) of one's knowledge about another person are interpreted and understood. These themes allow the presence of traits related to a central trait to be inferred, and the result is unified and coherent impressions reflecting these themes. Implicit Personality Theories. Compared with elemental views, then, holistic approaches hinted early on that any view of features as the sole building blocks of representations was likely to fall short. The concept of implicit personality theory has a similar flavor in that it refers to the very idea that relations exist among features (for reviews, see Leyens & Fiske, 1994; Schneider, 1973). Indeed, implicit personality theories are thought to be embodied in the linkages between traits. The concept was introduced by Bruner and Taguiri (1954) in part to account for how perceivers are able to form unified impressions around central themes and to infer the presence of traits from the presence of others to which they are related. It appears, then, that even some of the earliest findings from research reflecting a holistic approach to impression formation suggested, if considered in the language of mental representations, that features may not be the sole units of analysis for representational content, structure, and use. Indeed, as social-cognitive theorizing on representations developed, with the introduction of the schema concept, as well as the importation of the cognitive concepts of prototypes and exemplars, holistic themes persisted. Schemas, Prototypes, and Exemplars . The schema concept emerged in part from holistic notions. A schema is a “cognitive structure that represents knowledge about a concept or type of stimulus, including its attributes and

< previous page

page_127

next page >

< previous page

page_128

next page >

Page 128 relations among those attributes” (Fiske & Taylor, 1991, p. 98; see also Brewer & Nakamura, 1984; Carlston & E.R. Smith, 1996; Fiske & Linville, 1980; Markus & Zajonc, 1985; E.R. Smith, 1998). On the one hand, it is important to note in this definition that features/attributes are, once again, treated as the basic building blocks of representations. Yet, on the other hand, there is also an explicit recognition of relations among features, just as there is in the notion of implicit personality theory, implying the existence of internal organization and structure. Indeed, schemas have often been likened to theories that perceivers have about a given concept or stimulus, which harks back to the Gestalt notion of unifying themes or configurations (e.g., Markus & Zajonc, 1985). The term prototype has often been used interchangeably with schema in the social-cognitive literature (see Fiske & Taylor, 1991; E.R. Smith, 1998), which suits the present purposes. As in the cognitive literature, prototypes in socialcognitive work have referred to categories organized around a summary, central member or prototype, abstracted from experience with various instances (e.g., Cantor & Mischel, 1977, 1979). Likewise, this prototype has been characterized by a “fuzzy” set of features, and category membership has thus been seen as probabilistic rather than all-or-none. Also paralleling trends in cognitive work, exemplar approaches arose in the social domain as an alternative to prototype or schema views, suggesting that representations of people, groups, and social situations can be highly specific, designating individual instances rather than abstracted summaries of multiple instances (e.g., Lewicki, 1985; Linville & Fischer, 1993; E.R. Smith & Zarate, 1990, 1992). In the social realm, as in the cognitive arena of object categories, the assumptions of schema (or prototype) and exemplar views differ fundamentally in various ways (for a discussion of these differences, see E.R. Smith, 1998). Of central relevance here, however, are the generally converging assumptions these approaches make about how perceivers represent the social world, which are discussed next. So Where Do Social Cognition Researchers Stand ? Taken as a whole, theories in social cognition have tended to treat features, typically in the form of trait attributes, as the basic building blocks of perceivers' impressions of others and, presumably, of the representations that correspond to them. Indeed, in line with featural assumptions about content and structure, prevailing social-cognitive views on the use of representations have also tended to be featurebased. The relevant literature here is that concerned with principles of knowledge accessibility and use. In this body of work, it is widely agreed that, in addition to chronic and transient sources of knowledge accessibility, the activation and use of stored knowledge depend on applicability, which is defined as activation arising from the featural match or overlap perceived between the attended-to features of a stimulus and featural knowledge stored about a knowledge construct (Higgins, 1989, 1996; see also Bargh, 1997; Bruner, 1957b; Hardin & Rothman, 1997; Sedikides & Skowronski, 1991; Wyer & Srull, 1986). By serving as a basis for the contribution of applicability to knowledge activation and use, features have been accorded a central role in governing the use of representations in social perception (see Chen, Andersen, & Hinkley, 1999). In fact, paralleling feature-based views of categorization in the cognitive literature, described earlier, it has been explicitly argued that “the greater the number of shared features [between a stimulus and construct], the more the stimulus would contribute to a construct's activation level” (Higgins, 1996, p. 156). Yet as noted, there has long been the recognition that perceivers construe, and presumably store in memory, more meaning than implied by a list of independent trait attributes. This can be seen in Asch's (1946) early research on impression formation, in later work in the same Gestalt spirit (e.g., Heider, 1958; Lewin, 1951), such as work on implicit personality theories, and still later in the vast literature on schemas and prototypes. Each of these lines of thinking seems to suggest, in some manner, that organizing themes or configural meanings – indeed, theories – emerge out of individual attributes. Such emergent meanings imply that attributes are in some way related to one another and, in this regard, that the content, structure, and use of mental representations are likely to involve more than simple lists of unrelated features. Exemplar approaches in the social-cognitive literature, in their defining emphasis on separate instances, also suggest that relations between attributes may exist in being able to account for feature clusters or correlations. Before drawing the conclusion that social-cognitive work departs from prevailing feature-based approaches in the cognitive literature, it is critical to note several aspects of this work that suggest this conclusion is not yet warranted, at least on empirical grounds. First, as noted, the bulk of social-cognitive research has focused almost exclusively on trait attributes, despite considerable evidence that perceivers' conceptions of others also include knowledge about others' attitudes, thoughts, goals, feelings, and so on (e.g., Andersen, Glassman, & Gold, 1998; Fiske & Cox, 1979; Karniol, 1986; Park, 1986; Prentice, 1990; Read, Jones, & Miller, 1990; Trzebinski, 1989). Thus, if representational content,

< previous page

page_128

next page >

< previous page

page_129

next page >

Page 129 structure, and use involve themes or meanings that extend beyond trait attributes, the social-cognitive literature cannot speak very extensively about these yet. Second, in most research in the tradition of implicit personality theories, the focus has been limited to trait attributes, and theories have been conceptualized simply in terms of trait covariations or a small number of dimensions (for reviews, see C.A. Anderson & Sedikides, 1991; Schneider, 1973). That pairs or bundles of traits may covary does little to elaborate on, in a substantive way, the view that features (i.e., trait attributes) are the main building blocks of representations. The same holds true even when viewing implicit personality theories in terms of dimensions because such a view is still based on the notion of trait covariations, with traits that fall within or along the same dimension expected to covary. Overall, then, although long-standing social-cognitive theorizing in the spirit of holistic approaches – such as Gestalt views, implicit personality theories, as well as schemas, prototypes, and exemplars – clearly hinted at the notion that features are not the only building blocks of representations, in most existing socialcognitive research, alternatives to a feature-based approach to representational content, structure, and use have not been fully realized. BEYOND FEATURES: THE ROLE OF THEORY-BASED FORMS OF KNOWLEDGE Having provided some historical context, recent work on mental representations and their use is examined. The focus is on theory and research which propose, in one form or another, that beyond featural units of knowledge the content, structure, and use of representations involve theoretical, explanatory, or causal forms of knowledge. As a group, these bodies of work are referred to as theory-based approaches despite variations among them. After briefly touching on relevant cognitive and developmental work, emphasis is given to social-cognitive work on theories and theory-based processes so as to lay the groundwork for describing a theory-based approach to significant-other representations and transference. Cognitive and Developmental Work Despite their pervasive influence, alternatives to feature-based approaches to categories and categorization have arisen in the cognitive and developmental domains, challenging various aspects of feature-based views (e.g., Armstrong, Gleitman, & Gleitman, 1983; Murphy & Medin, 1985; Rips & Collins, 1993). Although an examination of the full range of these challenges is beyond the scope of this chapter (for discussions, see Goldstone, 1994; Medin et al., 1993), a few of the most widely questioned assumptions of feature-based approaches are considered next. Challenging Key Assumptions of Feature-Based Approaches . Perhaps first and foremost, feature-based approaches have typically assumed that category members cohere due to their featural similarity to one another, and that classification of a new instance as a category member is thought to depend on the degree to which the instance is featurally similar to a category prototype or to category exemplars. In short, feature-based similarity serves as an organizing principle in many feature-based views on categories and categorization. Closely related to this assumption is the view that categories essentially reflect lists of features, thus implying that features are independent entities, as noted previously. The assumption that feature-based similarity grounds categories and categorization has been called into question on numerous grounds, most notably by research showing its flexibility (for reviews, see Goldstone, 1994; Medin et al., 1993). For instance, research has shown that variations in the directionality of similarity comparisons may alter perceptions of featural similarity (e.g., Holyoak & Gordon, 1983;Srull & Gaelick, 1983; see also Medin et al., 1993). Research on Tversky's (1977) contrast model, which views similarity between two objects as a weighted function of their common and distinctive features, has also shown the fluidity of judged similarity as a function of factors such as the nature of the task and stimuli being judged, as well as the scope of the stimulus context (e.g., Gati & Tversky, 1984). Several cognitive and social-cognitive models of categorization similarly recognize that featural similarity is not fixed (e.g., Nosofsky, 1987; E.R. Smith & Zarate, 1992; see also Higgins, 1996). These models posit that similarity depends on perceivers' allocation of attention to particular stimulus dimensions, which in turn depends on, for example, the context. The malleability of feature-based similarity suggests that features may not be the sole building blocks of categories, nor the sole basis of categorization (e.g., Medin, 1989; Murphy & Medin, 1985; see also Goldstone, 1994). How so? If the degree to which members of a category are featurally similar and the degree to which a new instance is featurally similar to members of a category are not fixed, this implies that, under any given set of conditions, there must be knowledge extending beyond features that help provide coherence to categories and that serve as a basis for categorization, respectively. Such knowledge helps to define or constrain the basis for

< previous page

page_129

next page >

< previous page

page_130

next page >

Page 130 determining what makes members of a particular category similar to one another, and what instances should be classified as members of a category (Murphy & Medin, 1985). Indeed, a critical problem with featural similarity as an organizing principle is that it is unclear how the particular set of features that makes members similar, or that determines categorization, is decided (Medin, 1989; Murphy & Medin, 1985; Medin et al., 1993). There is an infinite number of features that make two stimuli similar or dissimilar to one another, or a new instance similar or dissimilar to the members of a category. As such, the assumption that categories reflect feature lists, and that categorization is based on some function of the number of shared and unshared features between an instance and a category, falls short because knowledge beyond features is needed to constrain what features are relevant to category membership (e.g., Murphy & Medin, 1985). Perhaps the clearest examples of this shortcoming are categories that group together instances that share little or even no obvious featural similarity, such as the category for ‘‘things to take out of one's house in case of a fire,” which might include pets, money, jewelry, and photo albums as members (example cited in Murphy & Medin, 1985; see also Barsalou, 1983; Medin, 1989). Feature-based approaches, and the assumption often inherent to them that feature-based similarity grounds categories and categorization, have also been challenged by research documenting the nonredundance of ratings of similarity and categorization judgments (Rips & Collins, 1993; for a review, see Goldstone, 1994; cf. Smith & Sloman, 1994). This work has shown that judgments of how similar an instance is to a category versus judgments of the likelihood that the instance is a member of the category are differentially influenced by variations in a given stimulus dimension. If similarity ratings and categorization judgments diverge, this implies that similarity, defined in featural terms, is not likely to be the only basis for categorization. If it were, judgments of how similar an instance is to a category prototype or exemplar should go hand in hand with judgments of the likelihood that the instance is a member of the category. In essence, challenges to feature-based similarity as the sole organizing principle of categories and categorization counter the idea that feature lists adequately capture the content and structure of categories. This idea has, in fact, been explicitly called into question by various strands of theorizing and research in both the cognitive and developmental literatures (e.g., Armstrong et al., 1983; Barsalou, 1983; Carey, 1985; Keil, 1989; Medin, 1989; Murphy & Medin, 1985; Rosch, 1978). For example, Armstrong et al. (1983) argued that knowledge about a given entity (e.g., bird) is likely to reflect not only a list of features associated with the entity (e.g., has wings, able to fly), but also relationships between these features (e.g., birds can fly because they have wings). Taken as a whole, challenges to feature-based approaches tend to suggest that the glue that holds category members together may involve more than shared features among them, and that the basis on which categorization judgments are made may involve more than the matching and mismatching of such features. In fact, the mounting evidence against features as the sole basis of categories and categorization coincides with the growing recognition of a potential role for theory-based forms of knowledge in cognitive and developmental work (for recent reviews, see Komatsu, 1992; Medin, 1989). Theory Based Approaches as an Alternative. Although there is variation in the precision with which theories are defined, and in how directly theory-based processes in categorization are addressed, theory-based approaches in the cognitive and developmental domains maintain that people's lay theories about the world are fundamental to how categories are represented and used. Unlike feature-based views, which tend to treat features as unrelated units, such approaches argue that explanatory relations among category features exist and are what make a category coherent (e.g., Medin, 1989; Murphy & Medin, 1985; Rips, 1991). Although some have argued for merely treating theoretical relations as kinds of features (e.g., Goldstone, 1994), such a resolution continues to imply that categories are grounded simply in features, rather than also, potentially, in the theoretical mortar binding the featural bricks of categories together (Armstrong et al., 1983; Medin, Wattenmaker, & Hampson, 1987; see also Gentner, 1989; Medin et al., 1993; Markman & Gentner, 1993). Put differently, it is argued that recognizing that theoretical relations may exist is “to concede that attributes may have a complex internal structure … this internal structure means that one is working with more than a list of simple attributes and that constraints and explanatory power will derive from this richer structure” (Murphy & Medin, 1985, p. 297). Support for theory-based approaches cuts across a wide range of domains and has been discussed in detail elsewhere (see Murphy & Medin, 1985). As alluded to before, particularly notable is evidence for the fluidity of featural similarity, which various researchers suggest implies that perceivers must bring knowledge extending beyond features to bear on judgments of similarity and categorization, and thus that categories and categorization are not simply feature-based (e.g., Goldstone, 1994; Medin et al., 1993). It is worth noting that much of the

< previous page

page_130

next page >

< previous page

page_131

next page >

Page 131 evidence for the malleability of featural similarity has involved judgments of the similarity between a pair of stimuli, rather than judgments of the featural similarity between a stimulus and stored category, with the latter speaking more directly to the possible role of theory-based forms of knowledge in categorization. However, some direct evidence for the influence of theories on categorization judgments also exists. For instance, developmental research has shown that children's theories about category membership may at times override the perceptual (i.e., featural) similarity of a stimulus to stored, category information as the primary basis of their categorization judgments (Gelman & Markman, 1986; for reviews, see Carey, 1985; Keil, 1989; cf. Jones & Smith, 1993). Research on category learning also points to a role for theories in categorization (Wisniewski & Medin, 1994). In this work, participants were shown two sets of children's drawings, which were either given neutral labels (“Group 1” vs. “Group 2”) or theoretical labels (“drawn by creative children” vs. “drawn by noncreative children’’). They were then asked to categorize a set of new drawings, providing a rule for each of their categorization decisions. Category learning varied depending on whether a theory had been activated. For instance, whereas the categorization rules of the neutral-label group referred mainly to concrete features, those of the theoretical-label group referred to relations between these features and more abstract ones that reflected previously assessed lay theories about what drawings done by creative versus noncreative children should look like. As a final example, cognitive work on analogical reasoning fits the notion that categorization may at times be theorybased (for recent reviews, see Gentner & Markman, 1997; Holyoak & Thagard, 1995). Specifically, some researchers have argued that relational (vs. attribute) similarity between stored knowledge and a stimulus may trigger the retrieval and use of the knowledge to interpret the stimulus (Genther & Markman, 1997; see also Gick & Holyoak, 1980, 1983; Holyoak & Thagard, 1995; Markman & Gentner, 1993). In other words, structural “parallels” rather than simply attribute-based matches may at times fuel categorization judgments (Gentner & Markman, 1997). Social-Cognitive Work Social-cognitive support for theory-based approaches also takes various forms. First, as described, some of the earliest contributions to social cognition suggested that perceivers' impressions of others reflect unified configurations or Gestalts rather than sets of separate elements or features (Asch, 1946). As noted, a major idea emerging from this work is that meaningful relations exist among the features (i.e., trait attributes) associated with an entity, a notion also embodied in the concept of implicit personality theory (e.g., Bruner & Taguiri, 1954; Schneider, 1973). Beyond this, early work on attribution theory also reflected Gestalt influences, as seen quite clearly in Heiderian views on the formation of causal units composed of persons and acts (Heider, 1958). More recent work in the Gestalt tradition has explicitly conceptualized relations between traits as theoretical in nature. Specifically, research has shown that one way people make sense of incongruent information about others is to impose a theory – that is, call on knowledge to serve expressly as an explanation – for why the incongruence exists (Asch & Zukier, 1984; see also Hastie, Schroeder, & Weber, 1990; Kunda, Miller, & Claire, 1990; Park, DeKay, & Kraus, 1994). Along the same lines, it has been argued that perceivers tend to integrate inconsistencies in their experiences into internally coherent narratives or stories, which similarly implies that perceivers may think in terms of explanatory relations (e.g., Baumeister & Newman, 1994; Murray & Holmes, 1993, 1994, 1996; Pennington & Hastie, 1992, 1993; Read & Miller, 1993; Thagard, 1989; Zukier, 1986). Recent work on person models also has the flavor of a theory-based approach (Park et al., 1994; see also Wittenbrink, Park, & Judd, 1998). These researchers argue that perceivers' impressions of others resemble stories or narratives consisting of a central concept that serves as an organizing structure, in which pieces of information about others are linked in ways consistent with the central concept. More specifically, they demonstrate that perceivers may extract different central concepts from an identical pool of target information, which leads to the formation of distinct person models that, while having certain target attributes or features in common, nonetheless imply, overall, very different portraits of the same target (Park et al., 1994). Thus, this work fits the idea that conceptions of others may include linkages among the attributes associated with them, some of which may be explanatory in nature, that transform the meaning of attributes considered in isolation. In the same spirit is recent research conceptualizing implicit personality theories as person types (Sedikides & C.A. Anderson, 1994; see also Andersen & Klatzky, 1987; C.A. Anderson & Sedikides, 1991; Cantor & Mischel, 1977, 1979). The main argument of this research is that perceivers often conceptualize others in terms of person types, which are composed of causally linked traits. While acknowledging the possibility of

< previous page

page_131

next page >

< previous page

page_132

next page >

Page 132 various other types of glue bonding traits together (e.g., evaluative), this research has explicitly shown that causal connections are one type of link between traits. In this regard, this work speaks quite directly to the idea that representational content and structure involves causal or explanatory forms of knowledge. Social-cognitive work on analogy also exists. This research offers rather direct support for theory-based approaches by demonstrating that explanations are stored in memory, and are capable of being retrieved and used (Read & Cesa, 1991; see also Schank, 1982, 1986). Specifically, it found that retrieval of a previously generated explanation was facilitated when participants were faced with the task of explaining a new, unexpected event similar in causal structure to the one for which the explanation was originally constructed (Read & Cesa, 1991). Thus, stimulus cues reflecting causal relations helped trigger and render applicable the stored explanation for the prior event. In a similar vein, recent research has argued that stereotype representations include knowledge about cause-effect relations, and that this knowledge may influence the processing of stereotype-relevant information (Wittenbrink et al., 1997; Wittenbrink, Hilton, & Gist, 1998; see also Kunda & Thagard, 1996; Shoda, Mischel, & Wright, 1989). This work focused on perceivers' theories about the cause of African-Americans' socioeconomic disadvantage (Wittenbrink et al., 1997). Participants scoring high on the Modern Racism Scale (McConahay, Hardee, & Batts, 1981) were thought to hold the cause-effect belief that African-Americans are responsible for their disadvantage, whereas those scoring low were thought to hold the belief that African-Americans are the victims of structural disadvantages. In two studies, the two participant groups varied in their perceptions of responsibility (i.e., causality) in a dispute involving an African-American target in a manner reflecting the nature of the cause-effect belief each held (Wittenbrink et al., 1997, Experiments 1 & 2). In a third study, the impact of participants' cause-effect beliefs on perceptions of causality was found only when the pool of to-be-interpreted information contained cause-effect cues that matched these beliefs – that is, when the information was presented in a way that facilitated versus inhibited construing causal structure among them (Wittenbrink et al., 1997, Experiment 3). Overall, this research suggests that matches between cause-effect relations (i.e., theories) stored in memory about a stereotype with cues in the stimulus environment that embody these relations contributed to the activation and use of the stereotype (Wittenbrink et al., 1997). In a somewhat different vein, theory-based notions can also be seen in connectionist models of social perception (Read, Vanman, & Miller, 1997; see also Miller & Read, 1991; Read & Miller, 1993; cf. Kunda & Thagard, 1996; E.R. Smith, 1996, 1998). Reflecting basic connectionist assumptions, these approaches generally argue that mental models of persons and social groups contain interconnected pieces of knowledge about others' traits, goals, beliefs, and so on, some of which are explanatory. Reminiscent of Gestalt principles, these approaches view social perception as conditional – that is, people perceive the characteristics, emotions, and behaviors of others in the context of one another, as a configuration, with the aim of achieving a coherent understanding of the social world. That is, each element or piece of knowledge about a person or group only has meaning in the context of all the other elements to which it is connected (Read et al., 1997). Conceiving of individuals in such conditionalized terms lies at the heart of the long-standing cognitive-social model of personality (Mischel, 1968, 1973, 1990; Mischel & Shoda, 1995). Countering widely held assumptions about the stable and traitlike nature of personality, this model takes an interactionist approach, positing IF-THEN relations as the basic units of personality. IFs refer to situations or conditions, and THENs refer to the responses that are exhibited in them. The notion that personality is composed of IF-THEN relations raises the possibility that perceivers may conceive of other individuals and social groups in terms of such IF-THEN relations, recognizing that what others think, feel, and do may be contingent on the situation (Shoda & Mischel, 1993). The idea that perceivers may possess knowledge about the IF-THEN relations of others served as a primary basis for the theory-based approach to significant-other representations and transference presented later. To provide a conceptual framework for this approach, the social-cognitive model of transference is first described (Andersen & Glassman, 1996; Chen & Andersen, 1999). The main thrust of this model is that the phenomenon of transference can be understood in terms of the activation and use of significant-other representations in interpersonal perception and relations. THE SOCIAL-COGNITIVE MODEL OF TRANSFERENCE The concept of transference refers to the idea that past experiences with and assumptions about significant others may resurface in relations with new others (Freud, 1912/1958). Said differently, transference entails the reemergence of aspects of past relationships in present

< previous page

page_132

next page >

< previous page

page_133

next page >

Page 133 ones. This general notion can be seen in various psychological domains, among them the clinical, personality, developmental, and social literatures (e.g., Baldwin, 1992; Bartholomew & Horowitz, 1991; Bowlby, 1969; Bugental, 1992; Feeney & Noller, 1990; Greenberg & Mitchell, 1983; Guidano & Liotti, 1983; Hazan & Shaver, 1987; Homey, 1939; Horowitz, 1991; Luborsky & Crits-Christoph, 1990; Rogers, 1951; Safran & Segal, 1990; Shaver & Rubenstein, 1980; Sullivan, 1953), and forms the basis of the social-cognitive model of transference (Andersen & Glassman, 1996). Basic Assumptions According to this model, mental representations of significant others are stored in memory, housing considerable knowledge, including characteristics of significant others, as well as feelings, motivations, and roles experienced in relation to them (for a review, see Chen & Andersen, 1999). Relative to representations of stereotypes, traits, and nonsignificant others, significant-other representations have been shown to be associatively richer, in that perceivers spontaneously list a greater number of idiographic features to characterize them (Andersen & Cole, 1990, Studies 1 & 2). Further, significant-other features have been shown to be distinctive compared with features of related stereotypes and traits, meaning they are less likely to be associated with stereotypes and traits compared with the reverse (Andersen & Cole, 1990, Study 2). Featural richness and distinctiveness suggest the greater structural complexity of significant-other representations relative to the other representations (Tversky, 1977; see also Andersen & Klatzky, 1987; Srull & Gaelick, 1983). Significant-other representations designate single individuals, rather than categories or types of individuals, and are thus exemplars (e.g., E.R. Smith & Zarate, 1990, 1992). As alluded to in prior sections, research has shown that exemplars are activated and used in social perception, just as are other constructs designating, for example, traits or stereotypes (e.g., Andersen, Klatzky, & Murray, 1990; Bargh & Thein, 1985; Higgins, Rholes, & Jones, 1977). As such, transference can be understood in terms of the activation and use of a significant-other representation to interpret and respond to a new person in representation-derived ways. Chronic and Priming Sources of Accessibility In viewing transference in terms of the activation and use of significant-other representations, it is important to consider principles of knowledge accessibility and use. In line with prevailing social-cognitive models (e.g., Bargh, 1997; Higgins, 1989, 1996), the triggering of significant-other representations depends on chronic and transient sources of accessibility. Chronic sources arise from the frequent past activation of a representation. They are often referred to as the chronic accessibility of a representation, or its chronic readiness to be brought to bear on social perception. Research has documented the chronic accessibility of significant-other representations by showing that participants made stronger inferences about a target person based on a significant-other representation, compared with various control representations, even in the absence of transient sources of accessibility (Andersen, Glassman, Chen, & Cole, 1995). One possible transient source of accessibility lies in cues in the environment that temporarily increase accessibility levels. This source is widely referred to as priming (e.g., Higgins et al., 1977; Srull & Wyer, 1979, 1980). Research has shown that priming contributes to the activation and use of significant-other representations (Andersen et al., 1995, Study 1). Specifically, although participants made stronger inferences about a target person based on a significant-other representation relative to various control representations, even in the absence of priming, inferences derived from participants' significant-other representations were still greater when these representations were primed. This finding is similar to ones from research on trait constructs showing that chronic and transient sources of accessibility may combine to trigger the use of trait constructs, as seen in the enhanced strength of traitconsistent inferences about new others in the presence of both sources versus either source alone (e.g., Bargh, Bond, Lombardi, & Tota, 1986). Applicability Sources of Accessibility The contribution of chronic accessibility and priming to the activation and use of a stored representation to interpret and respond to a stimulus have typically been examined using stimuli to which the representation is somehow applicable or relevant (for reviews, see Hardin & Rothman, 1997; Higgins, 1996). For example, research has used ambiguous targets of perception who have features that are equally similar to an activated representation, as they are to some alternative one (e.g., Higgins et al., 1977). Other research has used vague targets who have enough features to render a representation's use appropriate but not inevitable (e.g., Srull & Wyer, 1979). Beyond this, the potential for a to-be-interpreted stimulus to contribute to the activation and use of a representation has not been widely discussed (for exceptions, see

< previous page

page_133

next page >

< previous page

page_134

next page >

Page 134 Bruner, 1957b; Bargh, 1997; Hardin & Rothman, 1997; Higgins, 1989, 1996). Nonetheless, as discussed in prior sections, considerable research indicates that activation can also arise from applicability, defined as the overlap – that is, the match or fit – between the attended-to features of the stimulus and those associated with a stored representation (e.g., Higgins & Brendl, 1995; Higgins et al., 1977; Srull & Wyer, 1979, 1980).1 In terms of transference, cues in a new person that match knowledge stored about a significant other should enhance the activation and use of the representation of this other to interpret and respond to the person. Research has found direct evidence for such applicabilitybased triggering of transference, with applicability arising from a target person's featural resemblance to a perceiver's significant other (Andersen et al., 1995, Study 2). Specifically, participants' inferences about the target based on a significant-other representation were stronger in the presence of applicability relative to in its absence. The Basic Transference Paradigm To empirically examine the transference phenomenon, a combination of idiographic and nomothetic methods have been used. In the typical research paradigm, participants idiographically generate features to describe a significantother representation, as well as various control representations, in a pretest session. For example, they may generate descriptive features such as “Ramona is smart” or “The nerd studies hard.” In an ostensibly unrelated session, which takes place at least 2 weeks later, participants' idiosyncratic information is used as stimuli in a nomothetic experimental design. In the learning phase of the experimental session, participants ‘‘meet” one or several fictional target persons by being presented with a series of descriptive features about them. For each target, some of these features are ones that participants had generated in the pretest session to describe a given representation. One target is always characterized in part by features that participants had generated earlier to describe a significant other. Thus, applicability-based triggering of significant-other representations is operationalized in terms of feature-based cues in a target person that overlap with features associated with a significant-other representation. These cues serve as feature-based triggers for the activation and use of the representation to interpret the target, and are viewed as an analog for the cues that perceivers confront in the context of actual, face-to-face encounters with others (Chen et al., 1999). After the learning phase, participants are asked to complete a recognition-memory test about each target; they are presented with a series of descriptive sentences and asked to rate their confidence about having learned each one earlier (see Cantor & Mischel, 1977). Some of these are ones that were actually presented earlier to characterize a given target, and some are ones that were not actually presented earlier but that were derived from the representation to which the target corresponds. For the target corresponding to the participant's own significant other, recognition-memory confidence ratings for representation-derived descriptors that were not presented earlier serve as an index of transference – that is, of inferences about the target that are derived from the participant's representation of a significant other (e.g., Andersen & Baum, 1994; Andersen & Cole, 1990; Andersen et al., 1995, Andersen, Reznik, & Manzella, 1996; Baum & Andersen, 1999; Chen et al., 1999; Hinkley & Andersen, 1996). Overall, a growing body of evidence supports the social-cognitive model of transference. However, as in the broader social-cognitive literature, featural assumptions have been embedded in research on transference. Yet such assumptions may not pertain equally to all mental representations; instead, theories and theory-based processes may better characterize some representations and their use. In particular, this may be the case for representations of significant others, which is the central argument of the theory-based approach to significantother representations and transference presented next. A THEORY-BASED APPROACH TO SIGNIFICANT-OTHER REPRESENTATIONS, AND TRANSFERENCE The present approach draws in part from the cognitive-social model of personality and social perception (Mischel & Shoda, 1995; Shoda & Mischel, 1993). As noted, this model takes an interactionist approach, arguing that IF-THEN relations are basic to both personality and social perception. IFs refer to situations and THENs refer to the responses exhibited in them. In terms of personality, the model contends that the nature of IFs depends on the acquired meaning of a given situation for an individual (Mischel & Shoda, 1995) – that is, it depends on the individual's construal of the situation. Thus, for example, the IF designating a given situation may be construed as “at a party” for one person, whereas for another it is seen as “when meeting new people.” Regardless of how a situation is construed, the model maintains that situations activate particular sets of cognitive-affective units (e.g., feelings, goals, expectancies) that represent the psychological mediators of the IF-THEN relations that constitute personality.

< previous page

page_134

next page >

< previous page

page_135

next page >

Page 135 Applied to social perception, the model suggests that perceivers may perceive others in terms of IF-THEN relations, viewing others' feelings, thoughts, and behaviors as contingent on the situation (Shoda & Mischel, 1993). The present approach extends this basic idea by proposing that knowledge about the IF-THENs of others may at times represent theories – that is, IFs may serve expressly as explanations for THENs. What kind of IF-THEN knowledge might be considered a theory? It is argued that knowledge about the observable situations (e.g., “in school”) in which another person exhibits particular responses is not especially explanatory because it simply reflects prior empirical observations of the person. At times, however, perceivers may take an explanatory step beyond simply making observations about another person to making inferences about the psychological situations that produce his or her responses – that is, the psychological contingencies of these responses. Psychological situations refer to internal, private states such as feelings, wishes, or motives. In the present approach, IF-THENs in which the IFs refer to the psychological situations that elicit a person's THENs are seen as theories because, essentially, they reflect knowledge about the most proximal causes of the person's responses. That is, they reflect conceiving of the IFs of the person not simply in terms of the observable situations that, according to the cognitive-social model, activate psychological mediating variables, but rather directly in terms of the psychological mediators linking observable IFs and THENs. The central argument of the current approach is that perceivers are especially likely to conceptualize the IFs of their significant others at the level at which they perceive these situations are construed by these important others. In terms of the previous example, rather than merely viewing the IF as an observable situation such as “at a party” or “when meeting new people,” for a significant other, perceivers are particularly likely to view the situation in terms of the psychological situation of the person (e.g., “when feeling inadequate”). Thus, rather than simply conceiving of IFs in terms of the observable situations that presumably activate particular psychological situations, which in turn elicit particular THENs, perceivers are especially likely to understand the IFs of significant others directly in terms of these psychological situations. In summary, the present approach defines theories in terms of IF-THENs, in which the IFs represent inferences about the psychological situations of another person believed to produce his or her responses or THENs (e.g., “IF Ramona feels insecure, THEN she doesn't talk in class’’). Put another way, it proposes that perceivers' conceptions of others may include not only unrelated units of featural knowledge, but also psychological-state theories that take the form of short, explanatory IF-THENs, each composed of a psychological IF and a THEN to which it is causally linked. The overriding argument is that perceivers are especially likely to make sense of significant others in terms of such psychological-state theories. This implies that the content and structure of significant-other representations, as well as the use of these representations in transference, are likely to involve such theories. These assertions are bolstered by various strands of existing evidence, which are considered next. The Case for Theories about Significant Others As reviewed earlier, a growing body of social-cognitive work suggests that perceivers' conceptions of others involve explanatory or causal knowledge (e.g., Sedikides & C.A. Anderson, 1994; Wittenbrink et al., 1997). Such knowledge implies prior causal thinking on the part of perceivers. But why would perceivers be especially likely to engage in such thinking about their significant others? Causal Thinking About Significant Others. A wide range of theory and evidence suggest that, given the relevance of significant others to one's emotional and motivational outcomes, perceivers are likely to be especially willing to exert effort to make sense of these important individuals (see Andersen et al., 1997, 1998). Indeed, it is widely agreed that having one's outcomes dependent on others, such as is the case with significant others, engenders more effortful processing of information about these others than if there were no such outcome dependency (e.g., Erber & Fiske, 1984; Fiske & Neuberg, 1990). This desire to know significant others is likely to involve coming up with explanations for the behaviors of significant others (e.g., Fletcher & Fincham, 1991). Such explanations should enable perceivers to better regulate their own thoughts, feelings, and behaviors (Andersen et al., 1997, 1998). Said differently, figuring out why significant others think, feel, and act as they do is likely to serve a self-regulatory function, informing perceivers about how to feel, what goals to pursue, what to expect and hope for, and so on in relationships with these important others. Indeed, considerable evidence supports the basic idea that significant others are relevant to self-regulation, essentially demonstrating that the self is experienced in part in relation to significant others (e.g., Aron, Aron, Tudor, & Nelson, 1991; Baldwin, 1992; Downey & Feldman, 1996; Hinkley & Andersen, 1996; Moretti & Higgins, 1999). Direct evidence that perceivers engage in causal

< previous page

page_135

next page >

< previous page

page_136

next page >

Page 136 thinking about significant others can be found in close relationships work (for a review, see Chen & Andersen, 1999). For example, research examining attributional tendencies and relationship satisfaction indicates that relationship partners engage in attributional thinking about each other and their relationship (e.g., Fletcher, Fincham, Cramer, & Heron, 1987; Miller & Read, 1991; Read & Miller, 1989), and that the types of attributions that partners make about each other may be associated with satisfaction in the relationship (for a review, see Bradbury & Fincham, 1990). The association found between attribution type and satisfaction has been found both when attributions are assessed using direct solicitation procedures, as well as when they are assessed spontaneously, suggesting that attributional thinking in relationships may reflect controlled, conscious processing, as well as relatively automatic processing (see Fletcher & Fincham, 1991). Building on the attributions and satisfaction work is research examining the association between attributional tendencies and attachment style (Collins, 1996). This work suggests that individuals with different attachment styles are inclined to interpret and explain relationship events in ways consistent with their particular working models of self and other. Stated another way, it proposes that working models of the self and others entail explanatory tendencies, developed in part on the basis of past experiences with significant others (see also Downey & Feldman, 1996). For instance, this research has found that individuals with a preoccupied attachment style tend to formulate explanations for relationship events reflecting a more negative interpretation, as well as more negative views of relationship partners, than do securely attached individuals (Collins, 1996). In turn, such variations in explanatory tendencies have been shown to influence emotional responding, as well as self-reported behavioral intentions. Still more evidence that perceivers engage in causal thinking about significant others lies in research showing that perceivers construct theories about significant others to construct and maintain positive illusions about these individuals, thus bolstering feelings of confidence in relationships with them (e.g., Murray & Holmes, 1993, 1994, 1996; Murray, Holmes, & Griffin, 1996; see also Holmes & Rempel, 1989). In more concrete terms, this research has found that, when faced with a romantic partner's fault, people formulate theories to “encapsulate or defuse the negativity” of the fault (Murray & Holmes, 1993, 1994). For instance, they reconstrue the fault into a virtue, in this way forging protective, explanatory linkages between the faults and virtues of their significant others. Analogous data have emerged in research on transference (Andersen et al., 1996). Specifically, one study found that, on encountering a target person characterized by some negative descriptors associated with a positively evaluated significant other, participants responded to these descriptors with especially positive facial affect while encoding them. The inconsistency of the negative features in relation to the overall positive affective tone of the significant-other representation apparently led perceivers to transform the negative into a positive response – arguably a reflection of the influence of their positive theories about their significant others. The construction of theories that cast significant others in a favorable light may reflect a basic motivation to remain attached to these important others (Andersen et al., 1997). That is, they may reflect motivated attempts to bolster a perceiver's sense that a significant other loves him or her (see also Murray & Holmes, 1993, 1994, 1996). Overall, considerable research indicates that perceivers engage in causal thinking about significant others. Such thinking suggests that significant-other representations contain more than unrelated featural units of knowledge; they may also include knowledge reflecting causal relations – essentially, theories. But what evidence is there that such knowledge is especially likely to reflect psychological-state theories? Psychological-State Theories About Significant Others. On perhaps the most fundamental level, several lines of research have shown that representations of significant others contain more knowledge about internal, private states (e.g., thoughts, feelings) than do representations of various others, such as nonsignificant others (e.g., Andersen et al., 1998; Prentice, 1990). This finding bolsters the argument that perceivers are particularly likely to have psychological-state theories stored about their significant others insofar as it indicates that they have an especially large pool of psychological-state knowledge to draw from to construct such theories for these important individuals. Also relevant to this argument is recent research in which close relationships have been defined in terms of the inclusion of the other in the self (Aron & Aron, 1996; Aron et al., 1991). This research suggests that, to the extent a significant other is incorporated as part of the self, there should be greater similarity in the nature of cognition about the self and this other (Aron & Aron, 1996). For example, actor/observer differences in perspective are likely to be diminished (e.g., Andersen et al., 1998; Aron et al., 1991), resulting in greater similarity in the attributions made for one's own behaviors and those made for the other's behaviors (Nisbett, Caputo,

< previous page

page_136

next page >

< previous page

page_137

next page >

Page 137 Legant, & Marecek, 1973; see also Prentice, 1990; Sande, Goethals, & Radloff, 1988; Zuroff, 1982; cf. Taylor & Koivumaki, 1976). Similarity in cognition about the self and significant others implies some degree of similarity in representations of self and significant others. Although recent work has shown that self-representations contain more knowledge about private aspects (e,g., thoughts, motives) than do significant-other representations, it has also shown that significant-other representations resemble self-representations more so than nonsignificant-other representations in several ways (Andersen et al., 1998). Specifically, significant-other representations are more similar to the self than to nonsignificant-other representations in terms of the distinctiveness of knowledge about private aspects relative to public aspects (e.g., behaviors), as well as in terms of the accessibility of knowledge about private aspects. The notion that there is some degree of similarity in knowledge about the private aspects of the self and significant others is pertinent because, coupled with research indicating that perceivers generally view cognitive/affective information to be more informative than behavioral information and that this is especially true for the self versus others (Andersen & Ross, 1984), it suggests that information about significant others' internal, private experiences is likely to be seen as especially informative, just as for such information about the self. Overall, this coheres with the research noted earlier suggesting that perceivers are likely to be especially motivated to engage in effortful thinking about their significant others (e.g., Erber & Fiske, 1984), and extends it by suggesting that inferences about others' inner, psychological states are particularly likely to satisfy perceivers' desire to know their significant others, just as one's own inner workings and experiences are thought to best reflect one's “true” self (Andersen & Ross, 1984; see also Andersen et al., 1998; Johnson & Boyd, 1995). Overall, the bulk of existing evidence suggests that the knowledge perceivers store about the psychological states of significant others is substantial. Although it is less than what is stored about the self, it is nonetheless comparable in various ways to such knowledge about the self and is more available than such knowledge stored about nonsignificant others. Together, the clear availability of psychological-state knowledge about significant others, and the informational value perceivers place on such knowledge, along with evidence suggesting that perceivers are especially motivated to know their significant others, suggest that perceivers are particularly likely to draw on psychological-state knowledge to explain the behaviors of their significant others. In short, although it may often be sufficient to know that an acquaintance or a “nerd’’ is smart and skinny, it is unlikely that such simple, feature-based conceptions would suffice in making sense of a significant other. Instead, the present approach proposes that perceivers are especially likely to know their significant others in terms of psychological-state theories. Evidence for a Theory-Based Approach to Significant-Other Representation and Transference Research examining this approach has focused on three main hypotheses. First, the argument that perceivers are especially likely to conceptualize their significant others in terms of psychological-state theories suggests that such theories should be more available in memory about significant others relative to various control targets. Second, it suggests that these theories should be especially accessible from memory. Finally, if significant-other representations are especially likely to involve psychological-state theories, then cues in a new person that reflect such theories should be especially likely to trigger transference. Thus, it was hypothesized that applicability-based triggering involving psychological-state theories would serve as a particularly powerful source for the activation and use of significant-other representations to interpret and respond to new others – that is, transference. Three studies have been conducted to examine these hypotheses. In each, “theories” referred to participants' inferences or beliefs about the psychological situations (i.e., IFs) that explain the responses (i.e., THENs) of others. Also, in each study, participants were asked to idiographically identify a significant other (e.g., romantic partner) and, as control conditions, a stereotype (e.g., jock) and nonsignificant other (e.g., friend of a friend). Theory-Based Content in Significant-Other Representations. A first study tested the hypothesis that psychologicalstate theories are especially likely to be stored as part of significant-other representations (Chen, 1997, Study 1). To do so, this study examined two kinds of IF-THEN knowledge. In the first, the IFs referred to observable situations (e.g., “at a party”). To assess this knowledge, participants were given a list of sentence prompts for each representation. Each prompt designated an observable situation (i.e., IF), and participants were asked to complete the sentence if they felt they knew the most likely response (i.e., THEN) of the significant other, stereotype, or nonsignificant other in the observable situation. Thus, these IF-THENs assessed the number of empirical observations participants have stored about

< previous page

page_137

next page >

< previous page

page_138

next page >

Page 138 each representation (e.g., “IF at a party, THEN Amy talks to everyone”). In the second kind of IF-THEN knowledge, the IFs referred to psychological situations. To elicit such IF-THENs, participants were asked to consider each IFTHEN observation they had listed and provide an explanation for the observation if they felt they had one. More specifically, they were asked to provide a BECAUSE phrase (i.e., explanation) following each observation they felt they could explain (e.g., “IF at a party, THEN Amy talks to everyone BECAUSE…”). Participants' explanations were then coded as ones reflecting psychological states, such as thoughts, feelings, or goals (e.g., “…because he wants to get ahead” or “…because she thinks others don't like her”) or ones that did not reflect such states (e.g., ‘‘…because he's in school” or “…because it's her job). BECAUSE phrases referring to psychological states served as an indirect measure of psychological-state theories, in that they reflect participants' inferences about the psychological situations that elicit the responses exhibited in particular, observable situations. Put another way, these BECAUSE phrases reflect what participants perceive as the psychological mediators between observable situations (i.e., observable IFs) and the responses (i.e., THENs) that occur in them. In this regard, these psychological-state explanations represent participants' perceptions of the most proximal IFs of THENs. Participants were expected to provide more psychological-state theories to explain the IF-THEN observations listed for their significant other relative to their stereotype and nonsignificant other. This would support the hypothesis that perceivers are especially likely to go beyond simply making IF-THEN observations about their significant others to formulating theories that reflect beliefs about the psychological situations that mediate these observable, IF-THEN relations. Although the difference between the significant-other representation versus the control representations was expected to be largest in comparisons involving such theories, participants were also expected to list more BECAUSE phrases overall for their significant other relative to control representations, regardless of whether these phrases were based on information about psychological states. This prediction reflects the general idea that perceivers are especially likely to engage in causal thinking about their significant others. Several findings emerged from this study. First, participants listed more IF-THEN observations for their significant other relative to the control targets, a finding that fits prior evidence indicating that perceivers are more likely to view familiar versus less familiar others in situation-specific terms (e.g., Prentice, 1990; Zuroff, 1982). The situations examined in the earlier work were also observable ones, such as “in class” or “with family” (Prentice, 1990, p. 383). Extending this earlier work, a greater proportion of participants' IF-THEN observations about their significant others relative to those about the control targets was backed by an explanation (i.e., BECAUSE phrase). Moreover, the explanations given for IF-THEN observations about significant others were especially likely to refer to psychological states. In short, participants were especially likely to provide explanations about their significant others, and these explanations were particularly likely to reflect psychological-state theories. Despite strong support for this study's availability hypothesis, one limitation of this study was the nomothetic nature of the sentence prompts that were used. These prompts were used to measure the number of IF-THEN observations participants have stored, controlling for the observable situations considered. This also allowed an assessment of differences across representations in the degree to which participants' IF-THEN observations were backed by an explanation – based on the same set of situations. To address this limitation, a second study conceptually replicated the first using a fully idiographic procedure in which participants were asked to freely generate psychological-state theories for each representation (Chen, 1997, Study 2). The findings of this study converged with those of the first. Namely, participants spontaneously generated more psychological-state theories (i.e., IF-THENs with psychological IFs) to describe their significant other versus their stereotype and nonsignificant other. Taken together, these data clearly suggest that the content of significant-other representations is particularly likely to include psychological-state theories defined in IF-THEN terms. Theory-Based Structure in Significant-Other Representations. If perceivers tend to think about their significant others in terms of psychological-state theories, it was hypothesized that the linkages connecting stored knowledge about the psychological IFs and THENs of these individuals should be especially strong – that is, activation should spread relatively quickly across them. Thus, the component parts of psychological-state theories should be retrieved from memory in tandem quite readily. Put another way, psychological-state theories about significant others should be particularly accessible, suggesting that perceivers should find it especially easy to retrieve the psychological IFs that explain the THENs of these important others. To test this hypothesis, Chen (1997, Study 2) had participants generate psychological-state theories to describe each of the three representations on a computer,

< previous page

page_138

next page >

< previous page

page_139

next page >

Page 139 which recorded their generation latencies. More specifically, participants generated each theory by first typing in a response (i.e., THEN) true of the individual (significant other or nonsignificant other) or stereotype being described, followed by the psychological situation (i.e., psychological IF) that they believe produces the response. As predicted, participants showed significantly greater ease in coming up with psychological IFs to explain THENs (i.e., shorter average latency within IF-THEN units) about their significant other relative to the control representations. This was true even when controlling for the average latency between IF-THEN units –verifying the distinctiveness of the average latency within IF-THEN units, and bolstering the interpretation that the latter reflects the existence of IF-THEN units in memory that, for significant-other representations, possess particularly strong linkages. These results support the accessibility hypothesis and suggest that psychological-state theories about significant others take the form of especially tightly linked IF-THEN units, each composed of a psychological state and the piece of data it explains. Said differently, they indicate that the theory-based content of significant-other representations has structural implications. A Theory-Based Approach to Applicability-Based Triggering in Transference . Applicability's role in the activation and use of a representation can be understood in learning terms (Bruner, 1957b; Higgins, 1989, 1996; see also Bargh, 1997; Hardin & Rothman, 1997; see also Lewicki, 1986; E.R. Smith, 1990). Specifically, a representation's applicability to a stimulus depends on the stimulus patterns that have most frequently led to the activation of the representation in the past. Thus, applicability is based in part on a representation's activation history. For instance, some research has found that the effect of priming gender-stereotypic traits on ratings of male and female targets on these trait dimensions was greater when there was stereotype-consistent fit between the primed trait and the gender of the target (Banaji, Hardin, & Rothman, 1993). Presumably, target gender enhanced the applicability of the corresponding gender stereotype because of the gender-consistent nature of the activation history of gender stereotypes (see also Higgins et al., 1977; Leyens, Yzerbyt, & Corneille, 1996; Oakes, Turner, & Haslam, 1991). By and large, the stimulus patterns or cues thought to contribute to the activation and use of a representation have been conceptualized in featural terms. That is, applicability has been widely defined in terms of the match between features in a stimulus and feature-based knowledge stored as part of a representation, with greater match corresponding to greater applicability (Higgins, 1989, 1996; see also Chen et al., 1999), as noted in prior sections. But if perceivers conceive of their significant others in terms of psychological-state theories, the stimulus cues that should render a significant-other representation most applicable are ones that embody these theories. In other words, psychological-state theories about significant others should serve as a powerful basis for applicability's role in triggering transference. For example, observing the IF-THEN relation “when criticized, a target person gets angry” should activate a perceiver's representation of a significant other because the IF-THEN embodies the psychologicalstate theory that her significant other lashes out at others if feeling inadequate. In short, the main basis of applicability-based triggering should reflect the level at which perceivers tend to think about the entity designated by the representation (Hardin & Rothman, 1997). A third study examined the hypothesis that psychological-state theories about significant others play a distinctly powerful role in triggering transference (Chen, 1997, Study 3). This study used a variant of the basic transference paradigm and compared the contribution of theory-based applicability to that of feature-based applicability to the activation and use of participants' significant-other representations. More specifically, participants in this study learned about several target persons, one of whom was described by descriptors that participants had generated in a pretest session to describe a significant other. The other targets resembled various control representations. Thus, as in all research on transference, applicability-based triggering of participants' representation of a significant other took the form of a target's resemblance to this important other. The critical difference in this study was that applicability was based on either features or theories associated with the significant other. In the feature-based applicability condition, participants saw feature-based stimuli in the learning phase (e.g., “Terry is an intelligent person”). In contrast, participants in the theory-based condition were presented with stimuli reflecting psychological-state theories (e.g., “Terry is nice to others when he wants to get ahead”). In both conditions, participants were given a recognition-memory test about each target in which they were presented with a series of descriptors and were asked to rate their confidence about having learned each one earlier (see Cantor & Mischel, 1977). Some of these descriptors had been presented earlier to describe the target, and some had not actually been presented but were derived from the representation that the target resembled. Confidence ratings for representation-derived

< previous page

page_139

next page >

< previous page

page_140

next page >

Page 140 descriptors that were not presented earlier were averaged as a measure of the degree to which representationderived inferences about the target were made based on the corresponding representation, Inferences based on a significant-other representation served as an index of transference, as in prior research. Regardless of applicability condition, participants made stronger representation-derived inferences about the target resembling their own significant other relative to the control targets – an effect that constitutes evidence for transference. In the feature-based condition, this effect replicates prior research in which applicability-based triggering of transference has always taken the form of feature-based cues in a target that matches features associated with participants' significant other. However, if perceivers tend to represent their significant others in terms of psychological-state theories, then theory-based cues in a new person should produce a stronger transference effect. Although some deviations from this strict prediction emerged, the transference effect – defined in terms of greater representation-derived inferences about the significant-other target relative to the control targets – was in fact significantly stronger in the theory-based applicability condition relative to the feature-based one. This finding supports this study's key hypothesis that theory-based applicability plays an especially powerful role in triggering transference. Taken together, the three studies described earlier provide support for a theory-based approach to significant-other representations and transference. The content, structure, and use of representations of significant others appear to be especially likely to involve theories, defined specifically as perceivers' IF-THEN knowledge in which the IFs refer to the psychological states that elicit the THENs. ISSUES AND FUTURE DIRECTIONS On one level, the theory-based approach presented earlier clearly resonates with the recurrent idea in the socialcognitive literature that perceivers strive for coherence in social perception, thus viewing others in terms of configurations, organized around central themes, rather than in terms of separate elements or features. Asch's notion of central traits, and later concepts such as implicit personality theories and schemas, embody this basic idea. It is an idea that leaves room for the possibility that theoretical relations exist among features and are basic to representational content, structure, and use. On several other levels, however, the present approach extends existing social-cognitive concepts. Among them, it goes beyond viewing features simply in terms of trait attributes, which are the focus of most social-cognitive research. More precisely, it posits not only that representational content and structure may include explanatory relations, but also that these relations may involve inferences about others' feelings, motives, thoughts, and so forth. On a more basic level, the current approach offers a concrete view of the form that theories might take in the social realm, which is a level of specificity that is absent in most social-cognitive work consistent with, or even explicitly espousing, theory-based notions. On a different note, the present approach is also distinct in arguing that representational content, structure, and use are likely to vary as a function of the significance of the “other’’ designated by representations. More specifically, it suggests that to understand the content and structure of perceivers' representations of others, as well as the processes that govern the role of these representations in interpersonal perception and relations, it is critical to consider the nature of perceivers' relations with these others (see also Hardin & Conley, chap. 1, this volume). Put differently, how the self is related to others is likely to influence how perceivers go about making sense of these others, and, as a result, how knowledge about these others is mentally represented and how and when it is likely to be used. By proposing that the nature of perceivers' relations with others influences mental representations and their use, the current approach is part of an expanding literature reflecting a cross-fertilization of social-cognitive theory and close relationships work. Finally, in a related vein, the present approach extends existing work on the principles that govern knowledge activation and use by suggesting that applicability may at times be primarily feature-based, and at other times theory-based, depending on how perceivers conceive of the entity designated by a given representation (see Chen et al., 1999). Of course, many issues have yet to be addressed. Some of these are considered next, followed by a discussion of some future directions they suggest or compel. What Is Meant By “Theories?” Existing theory-based approaches vary considerably in terms of how and the precision with which theories have been defined. One dimension onto which many variations fall is the globality of theories. Whereas some theory-based approaches appear to define theories as single, global, or all-encompassing explanatory principles, others suggest a lower level of analysis, at which multiple

< previous page

page_140

next page >

< previous page

page_141

next page >

Page 141 theories about a single entity exist, embodied in linkages connecting bits of featural knowledge. For example, global theories often seem to be implied in cognitive and developmental work, perhaps because this work tends to deal with theories about objects and other nonhuman entities (e.g., lawnmowers) which focus on the function of these objects or some other singular essence. Of course, some social-cognitive conceptions of theories, which generally deal with human entities (e.g., individuals, social groups), also seem to suggest a global view of theories. For instance, the notions of central traits, implicit personality theories, and schemas often seem to imply a single organizing principle or theory. Although the theory-based approach proposed in this chapter concurs with relatively global views of theories in the basic argument that representations involve explanatory knowledge, it is worth reiterating that theories were defined in the present approach as knowledge that explains other stored knowledge, rather than knowledge that fully accounts for who or what an entity is (see Read, 1987, with the latter implied by more global definitions of theories). Although perceivers may well possess single, global theories about objects, people, and groups alike, research testing the present approach has thus far only examined theories at a lower level of analysis, namely, in the form of IF-THEN units. Defining theories in such IF-THEN terms fits squarely with the increasing recognition in the socialcognitive literature that perceivers may at times view the social world in interactionist, IF-THEN terms (e.g., Bargh, 1997; Prentice, 1990; Shoda & Mischel, 1993). What Is Meant By Theory-Based Processes? In terms of what is meant by theory-based processes, theory-based approaches espousing relatively global definitions of theories often tend to focus on the notion that theory-based knowledge may be invoked in various judgmental tasks. For example, perceivers may draw on a global theory about what makes a given object what it is in deciding whether an instance is a member of a given category, or in making theory-derived inferences about the instance (see Murphy & Medin, 1985). The latter type of theory-based process is often what is implied in the vast social-cognitive literature on schema-driven processing. Schemas have often been likened to theories that guide topdown processing, whereby perceivers make schema-derived inferences about a stimulus person. In contrast, the present approach, which defines theories in terms of explanatory relations among featural bits of knowledge, is specifically concerned with the role of theory-based processes in knowledge activation and use. That is, as noted, the focus is on the match between cues in the stimulus environment with stored, theory-based units of knowledge. Put another way, theory-based processes have to do with the mapping of theory-based cues in the stimulus environment to theory-based content and structure, which should enhance a representation's likelihood of activation and use (see also Gentner & Markman, 1997; Wittenbrink et al., 1997, 1998). Overall, future work needs to be precise about what is meant by theory-based content, structure, and processes because variations are likely to carry distinct implications for the role of theories in interpersonal perception and relations. Some Further Avenues to Explore Beyond addressing what is meant by theories and theory-based processes, future research might consider more closely the notion that self and significant-other representations are related. In the research presented in this chapter, the focus was on perceivers' conceptions of who their significant others are – specifically, on their psychological-state theories reflecting knowledge that is entirely about their significant others. However, another potentially critical kind of IF-THEN is that in which perceivers themselves are the IFs of their significant others' THENs. Just as perceivers are likely to be motivated to figure out the psychological situations that elicit their significant others' responses, they may also be motivated to understand the relational dynamics that elicit these responses. Examples of such relational IF-THENs might be “When I am feeling frustrated, my partner soothes me” or “When I am weak, my mother criticizes me” (see also Baldwin, 1992; Baldwin, Fehr, Keedian, Seidel, & Thomson, 1993). The idea that linkages between the self and significant others exist coheres well with various lines of research (e.g., Baldwin, 1992; Bugental, 1992; Hinkley & Andersen, 1996). For example, research outside of the transference realm has shown that activation of a significant-other representation leads to changes in perceivers' self-evaluations (Baldwin, Carrell, & Lopez, 1990). It is argued that such self-evaluative changes occur by virtue of self/significantother linkages stored in memory as part of perceivers' “relational schemas” (for a review, see Baldwin, 1992). In the transference domain, research has shown that when a significant-other representation is activated and used in an encounter with a target person, stored expectations regarding the significant other's acceptance or rejection are likely to color relations with the target such that the perceiver comes to expect the target to accept or reject him or her just as his or her significant other does (Andersen et al., 1996).

< previous page

page_141

next page >

< previous page

page_142

next page >

Page 142 Similarly, stored motivations to approach or distance oneself from a significant other have been shown to come into play in relations with new others when transference occurs. Expectations and motives associated with significant others reflect knowledge about relational patterns between the self and significant others, just as relational IF-THENs do. Beyond this, research has shown shifts in self-definition and self-evaluation in the context of transference (Hinkley & Andersen, 1996). That is, activation of a significant-other representation accordingly activates aspects of a perceiver's self that are associated with the relevant significant other, leading the perceiver to become the self he or she is when with the other – that is, the “self-with-significant-other’’ (see also Ogilvie & Ashmore, 1991). Thus far, research on transference has conceptualized self/significant-other linkages in terms of what they imply for the nature of interpersonal perception and relations. In line with the current theory-based approach, however, future research might consider the role of relational IF-THENs in triggering transference. For example, having one's concerns about a particular issue belittled or dismissed by a new acquaintance may be a relational cue that reminds one of a pattern of interaction often experienced with a significant other, thereby triggering the activation and use of the relevant significant-other representation. The idea that relational IF-THENs may serve as triggers of transference raises the questions of when and for whom different kinds of IF-THEN cues might elicit transference. Cross-cultural differences in social perception are interesting to consider in this regard. For example, research examining the distinct ways in which Americans and Hindus explain others' behaviors suggests that Hindus may be less likely to come up with explanations that refer to internal or psychological aspects of others (Miller, 1984). In general terms, this research found that Hindus were more likely than Americans to generate contextual explanations for an actor's behavior, such as explanations referring to social norms, social roles, or interpersonal relationships (e.g., “Because she is his aunt”), rather than explanations referring to something internal about the actor (e.g., “He was interested only in himself”). Such differences in the kinds of attributions made by Americans compared to Hindus were interpreted as reflecting the emphasis placed in Indian culture on “locating a person, object, or event in relation to someone or something else” (Miller, 1984, p. 968). This work suggests that evidence for the pronounced availability and accessibility of theories about significant others, defined specifically in terms of IF-THENs with psychological Ifs, and the especially powerful role of such psychological-state theories in triggering transference, may hold more or less strongly among perceivers across different cultures. CONCLUDING REMARKS Because much of everyday social cognition involves significant others, it is critical to consider how perceivers go about making sense of these individuals. The theory-based approach presented in this chapter, and the findings it has yielded thus far, suggest that perceivers are especially likely to make sense of significant others in terms of psychological-state theories about them. This is reflected in both the content and structure of their representations of these important others, as well as in the triggering of these representations in encounters with new others. On a broader level, this work suggests the need to explicitly incorporate a role for theories in social-cognitive work on mental representations and the processes that govern their use in making sense of the social world. Social cognition may often be theory-based, and potentially in qualitatively different yet systematic ways, about the different persons and groups that make up the perceiver's social world. ENDNOTES 1. Although applicability has been theoretically defined in terms of overlap between a stored representation and cues in a to-be-interpreted stimulus (e.g., Hardin & Rothman, 1997; Higgins, 1996; see also Srull & Wyer, 1979, 1980), applicability has not been examined strictly in these terms in the small body of research examining its contribution to the activation and use of stored representations. In this work, much of which has focused on the applicability of a stored, trait construct to a target stimulus, the degree of applicability Or match of the trait construct to the stimulus has most typically been operationalized on the basis of pilot participants' ratings of how much the stimulus implies the trait (e.g., Higgins & Brendl, 1995). Nonetheless, research on transference has taken a rather strict, featurebased approach in defining applicability, guided by theorizing on feature-based approaches from the social-cognitive domain (e.g., Higgins, 1996), as well as from the cognitive and developmental literatures (for recent reviews, see Komatsu, 1992; Medin et al., 1993). As noted, these approaches generally maintain that categorization reflects in part the degree of feature-based matches and mismatches between stored knowledge about a category and a to-beclassified stimulus.

< previous page

page_142

next page >

< previous page

page_143

next page >

Page 143 9 What We Theorize When We Theorize That We Theorize: Examining the “Implicit Theory” Construct From a Cross-Disciplinary Perspective Michael W. Morris Daniel R. Ames Eric D. Knowles Stanford University University of California, Berkeley The human ability to make sense of events through generating explanations has long fascinated philosophers, psychologists, and other scholars. Particularly impressive is the facility with which people generate explanations for others' behaviors. Starting at a surprisingly early age, children seem to instantly interpret observed action in terms of what an agent is thinking, wanting, and planning. They also trace these fleeting mental states to origins in enduring personal dispositions and situational forces. This penchant for making sense of others as individuals with particular beliefs, desires, goals, and traits is essential to how we navigate the social environment. Our inferences about an actor's dispositions tell us whom to avoid, whom to trust, whom to blame, and so forth. Moreover, our perceptions of situational forces tell us which positions to seek out and which predicaments to avoid. Many scholars have described the lay person's process of generating explanations by analogy to the scientist's process of reasoning from theories. The role of theories in scientific explanations is seen very clearly in the field of psychology because the factors driving behavior cannot be directly observed and hence have to be posited on the basis of a theory. Areas of psychology that settle on a cogent theory often move forward quickly as researchers easily generate explanations for a variety of phenomena; in areas lacking a theory, generating explanations is more difficult. Hence, in accounting for the facility of lay people in generating explanations for behavior, it has been tempting to suggest that they possess theories. Although the differences between lay and scientific explanation are apparent, a host of scholars have drawn the analogy to capture similarities between the lay person and the scientist. This account – sometimes called the “theory theory” – has developed in several different academic disciplines. In philosophy, scholars have described the role of folk psychology (e.g., Dennett, 1991). In developmental psychology, researchers have proposed that children have an implicit ‘‘theory of mind” (e.g., Wellman, 1990). In social psychology, researchers have argued that individuals understand behavior in part by invoking implicit theories (e.g., Bruner & Tagiuri, 1954; Heider, 1958; Ross & Nisbett, 1980; Wegner & Vallacher, 1977). In psychological anthropology and cultural psychology, a growing body of work examines the roles of cultural theories in judgment and explanation (e.g., D'Andrade, 1995; Peng, Ames, & Knowles, 2000). The essence of these accounts is that lay people work with knowledge structures that have the same structure, function, and relationship to evidence as that of scientific theories. Proponents of the theory theory differ in how far they push the analogy between lay and scientific theories. In this chapter, we take the analogy seriously by identifying the criteria for identifying a piece of knowledge as a theory. This distinguishes the theory account from rival, alternative accounts that posit different kinds of knowledge as the mechanism for lay social understanding. One class of alternatives posits other kinds of learned, structured knowledge such as narratives (e.g., Bruner, 1990) or metaphors (e.g., Lakoff, 1987). Other alternatives posit knowledge structures that are not learned: innate modular cognitive mechanisms for social understanding (Fodor, 1986). Elsewhere, accounts posit learned knowledge that is not structured into theory-like forms, such as knowledge that exists in an associative network (e.g., Read & Miller, 1993) or knowledge derived from imaginative simulation of another person's state of mind (Goldman, 1993). All of these proposals are plausible and may capture part of people's social understanding competence. Our goal, however, is to review the evidence that supports the theory theory relative to these other proposals. This does not require that the alternative approaches be wrong, just incomplete. We do so by reviewing theory theory scholarship in four fields – phil-

< previous page

page_143

next page >

< previous page

page_144

next page >

Page 144 osophy, developmental psychology, social psychology, and cultural psychology. We begin by identifying defining characteristics of theories. WHAT IS A THEORY? In claiming that people employ lay theories, it is meant that they work with mental structures in a way that is analogous in important respects with the way scientific theories are developed and applied (Gopnik & Wellman, 1994; Ross & Nisbett, 1980). To distinguish theories from other types of knowledge, we review key criteria that have been distilled in philosophy of science. The criteria considered are widely accepted by philosophers as the defining characteristics of scientific theories (although there is by no means complete consensus) and accordingly provide a promising starting point in clarifying the claims involved in positing that lay people possess and use implicit theories. Structural Features Several structural features distinguish theories from many other kinds of knowledge. First, theories involve an ontology – a set of assumptions about the kinds of things that exist. To reason with a theory means treating these often unobservable entities as if they exist (Boyd, 1991). For example, genes were posited theoretically long before it was possible to observe them directly. An example from social psychology is the state of cognitive dissonance that was posited based on theory rather than on direct measurement. Second, theories propose causal mechanisms – principles that account for regularities in concrete phenomena (Cartwright, 1991). The fact that most Icelandic people have blue eyes is explained, for example, in terms of the causal role of the blue-eye allele and its preponderance in the Icelandic gene pool. Third, theories comprise a set of coherent principles, not merely a list of unrelated causal beliefs. Constructs and principles interrelate in lawful ways, working in conjunction to explain observable phenomena (Glymour, 1980). For instance, certain genes are posited to explain blue eye color, whereas other genes are posited to explain brown eyes, and there are principles describing their interaction – for example, that the dominant brown gene, if present, will block expression of the recessive blue gene. A final structural property, resulting from the first three, is that theories are expressed in abstractions rather than in the vocabulary of concrete observations (Gopnik & Meltzoff, 1997). The property of abstractness sets theories apart from other knowledge structures. A theoretical statement is more abstract than an empirical generalization, which merely infers a regularity (e.g., “Most Icelandic people have blue eyes”) from a set of observations (“Most Icelandic people I have seen have blue eyes”) and are still phrased in the language of the evidence. The abstractness of theories distinguishes them from other belief structures such as analogies and stories that preserve the concreteness of observation. Functional Features Another distinguishing characteristic is that theories can be used in explaining, predicting, and interpreting phenomena. Like many knowledge structures, theories afford explanations – that is, accounts of how an outcome was caused. Although simple attributions of an outcome to a cause could be produced from a simple cause-effect belief, more complex explanations such as those involving multiple plausible causes or counterfactual conjectures would mark the existence of a theory about how different causes interact. Next, the abstractness of theories means they are less tightly yoked to past experience (than are other knowledge structures) and hence enable prediction about a wider range of phenomena – even phenomena unknown at the time of a theory's construction. Finally, theories also aid in interpretation of evidence – that is, in deciding which evidence is relevant to a particular problem and which observations may be ignored or attributed to methodological failings. Empirical Defeasibility Perhaps the most central issue in analyses of how science differs from religion or purely formal disciplines is the relation between theory and empirical evidence. Philosophers of science have debated how scientific theories depend on, and change in light of, counterevidence. The influential empiricist school (Carnap, 1966; Popper, 1959) proposed a simple model in which predictions are deduced from a theory and subsequently compared to empirical observations. According to this account, when the predictions conflict with the observations, the theory is falsified and rejected. This model has itself been rejected based on arguments of Kuhn (1977) and others about resilience of theories against initial counterevidence. The history of science reveals that theories are fortified with ad hoc auxiliary hypotheses that explain away conflicting data for quite some time until the structure becomes so unwieldy that scientists find it unattractive and rally around a new paradigm (Kuhn, 1977). A full answer to the question of how theories are empirically defeasible requires distinguishing types of

< previous page

page_144

next page >

< previous page

page_145

next page >

Page 145 theories at different levels of abstraction or generality. Following philosophers of science (e.g., Lakatos, 1970; Lauden, 1977), we can distinguish the general frameworks that define a discipline or paradigm (e.g., trait psychology, the framework of tracing differences in behavior to dispositions) from the more specific models that guide research within a content area (e.g., the authoritarian personality model of prejudice). A theory's level of abstractness is related to its resistance to counterevidence and, consequently, to its ability to resist change. Because more specific scientific theories are more tightly yoked to particular patterns of evidence, they are the more easily disconfirmed. Likewise, lay theories that are more specific may be more likely to change based on the lay person's observations than more general framework theories (Wellman, 1990). The commitments of the theory theory stance can now be seen more clearly. Lay people possess knowledge structures that (a) abstractly represent the kinds of things that exist in a domain and the configuration of causal relations among these things; (b) are used in explaining, predicting, and interpreting; and (c) are affected by empirical evidence. Few scholars would claim that lay theorizing parallels that of science in a more extended sense (i.e., lay theorists do not submit their conclusions for anonymous peer review). Rather, theory theory claims rest on the three basic defining features of theories that are enumerated earlier.1 THEORY THEORIES IN PHILOSOPHY Whereas philosophy of science has provided our criteria for theory, philosophy of psychology provides our first literature in which the theory theory of social understanding has been advanced. A long tradition of philosophers concerned with the problem of how lay people understand each others' minds have assumed that people interpret observed behavior according to a body of commonsense knowledge about how people's situations, mental states, and behavior are related. This structured knowledge, seen as analogous to the knowledge garnered through scientific psychology, has been referred to as folk psychology (see e.g., Carruthers & Smith, 1996). This knowledge is theory-like in its structure in that it specifies an ontology (positing kinds of mental states, such as beliefs, desires, and goals), describes causal mechanisms for mental states to direct behavior (reactions, deliberate acts, etc.), and integrates these mechanisms at an abstract level. In addition, this knowledge is theory-like in its function in that people use it to interpret, explain, and predict others' behavior. Moreover, the presumption is that individuals adjust their folk psychological beliefs in light of empirical evidence. Among philosophers of psychology who acknowledge the existence of folk psychology, there remains disagreement over the issue of whether these lay people's theories mirror the actual causes of behavior. Those in the realist camp think folk psychological notions about mental states such as intentions, beliefs, and desires are largely accurate; these folk-theorized entities are real (Searle, 1983). Those in the antirealist camp compare such folk notions to debunked historical ideas about impetus and phlogiston: Although widespread, they are ultimately wrong (Ryle, 1962) and do not correspond to the real causes of behavior in the brain (Churchland, 1988). In between the positions that folk psychology is largely correct and utterly invalid lies the position of Dennett and others, that folk psychology should be evaluated for its practical value rather than its veracity. Folk psychology works not necessarily because it is true, but because it allows social coordination. According to Dennett (1986), beliefs and desires are not real things so much as abstracta – calculated entities like centers of gravity or vectors that are useful in prediction. Right or wrong, folk psychology is an interpretive stance that gives people an ability to deal with others. As we shall see, this question of whether lay theories lead to accurate understandings is an important one across disciplines. The important point for now is that proposing theories as the mechanism for social understanding does not require one to believe these theories posit constructs that have a real existence. Indeed, the inaccuracies that come from overreliance on theories is an important source of evidence for the theory theory. Challenges The position in philosophy that social understanding relies on a folk psychology is challenged from two directions. A nativist challenge by Fodor (1986) and others holds that evolution has provided us with an innate capacity to understand other minds by hardwiring modules for detecting intentions and other crucial mental states from their signatures in others' observable behavior. These perceptual modules meet some criteria for a theory. For example, they serve the function of explanation. However, they fail other criteria, such as coherence and empirical defeasibility, respectively, because they are encapsulated and indefeasible. Hence, this position differs from the position that lay people develop theories of other minds. A second challenge emphasizes the primacy of introspection, arguing that people's other-understanding relies on a simulation of one's self in their position. Gold-

< previous page

page_145

next page >

< previous page

page_146

next page >

Page 146 man (1993) argued that, ‘‘One can predict another person's choices or mental states by first imaging himself in the other person's situation and then determining what he himself would do or how he would feel. … You do not need to know any psychological laws.” In other words, folk psychology does not hinge on something people know (a knowledge structure), but rather on something they do (the process of introspection). The evidence adduced for a simulation account over a theory account is the seeming superiority of self-knowledge of mental states. Philosophical proponents of the theory theory perspective have not fully explained why people feel a stronger understanding of their own minds than of other people's minds. These two basic challenges to the theory theory – nativism and simulationism – are seen in other disciplines as well. In the philosophical literature, it seems there have been no deciding arguments by theory theory proponents that have not been answered by the nativist and simulationist camps. As we shall see, however, empirical work in areas of psychology has provided more incisive evidence. THEORY THEORIES IN DEVELOPMENTAL PSYCHOLOGY Debate about implicit theories and social understanding in the developmental psychology literature centers on some of the same issues as the philosophical literature. Yet developmentalists bring a different kind of evidence to these questions: Developmentalists track age-related changes in children's performance that might suggest what kinds of structures or processes underlie that performance. The history of lay-theory constructs in developmental psychology begins with Piaget, who proposed that children develop causal theories, or schemata,2 in order to understand the physical (Piaget, 1960) and social worlds (Piaget, 1965). Piaget assumed that these theories are associated with distinct ontological kinds – mechanical events and human behavior, respectively – and that they capture the associated causal principles. Also, Piaget is explicit that such theories are abstract enough to be cognitively reversible to allow both prediction and explanation. The dependence of the child's theories on the child's empirical observation and experimentation is also clearly assumed in Piaget's writing. The weakness of Piaget's account of the child's understanding is its own relation to empirical evidence: Piaget's clinical method of interviewing children did not allow him to observe the sophistication of children's understanding. Subsequent research with methods that minimize communicative demands on the child has refuted many of Piaget's claims about the limitations of social and physical understanding in older children. Correspondingly, there has been a theoretical shift toward the view that a child's empirical learning complements innate capacities. For example, researchers of children's understanding of physical and biological concepts find evidence that children are prepared to carve the world of experience in particular ways. Even young infants seem to work with a basic ontology and associated causal principles to interpret, categorize, and predict events (see Carey, 1985; Keil, 1989). This work was reinforced by contemporaneous arguments that the coherence of categories used by adults derives not merely from the similarity of shared features, but from the general theories in which categories are embedded (Murphy & Medin, 1985). The most encompassing theory theory of children's understanding was presented by Gopnik and Meltzoff (1997). These researchers claimed that children develop a number of theories in their first years of life, starting with initial innate notions and revising them into increasingly elaborate learned theories. Children are seen (like scientists) as tinkerers and testers, sometimes experimenting and sometimes (again, like scientists) misinterpreting and ignoring counterevidence to their current theories. Gopnik and Meltzoff (1997) pointed to evidence that children are born with rudimentary theories. The fact that newborn infants imitate other humans – but not machines and objects – indicates an awareness of the human kind. The fact that they look for effects of their actions in their local environments suggests an understanding that the nexus of mechanical causation is physical contiguity. At around 9 months, children begin to act on principles of physical causality, through regulating their physical contact, and on principles of social causality, through communicative attempts toward people. They begin to develop different notions of causality in two domains: people and objects. By 18 months, their actions take into account specific properties. An understanding of physical principles becomes apparent in their manipulation of objects and an understanding of specific properties of individuals apparent in their interactions with people. Eventually, through empirical observation and theory change, children become able to understand that other persons have distinct beliefs that are representations of reality. To illustrate how developmentalists investigate children's understanding of beliefs, let us consider how children's errors provide evidence for their use of a theory. Piaget noticed that children fail to appreciate the extent to which beliefs differ between self and other, and suggested that children conceive of thoughts like objects

< previous page

page_146

next page >

< previous page

page_147

next page >

Page 147 – simply a part of the physical environment. Children's understanding of belief has been greatly illuminated in recent years by studies with Wimmer and Perner's (1983) “false-belief” task. In its original form, a young subject learns about Maxi, a character who places some chocolate in a particular location in a room. After Maxi leaves, the chocolate is moved to another location. The child is then asked where Maxi will look for the chocolate on return. To succeed, the child must think that Maxi thinks that the chocolate is in the original location rather than in what the child knows to be its true current location. In other words, to predict Maxi's behavior, the child must entertain the notion that this other person has a belief that does not correspond to current reality: a false belief. Young children fail this task, indicating that they cannot conceptualize and reason about false beliefs. Since Wimmer and Perner's (1983) work, a nearly unbelievable number of false-belief studies have been conducted and have converged on the finding that children undergo a dramatic shift in the ability to solve the problem at around age 3 (see Astington et al. 1988; Carruthers & Smith, 1996). Shifts in the child's understanding of others, such as the shift in understanding of belief around age 3, are described as analogous to paradigm shifts in science by Wellman (1990). Contrary to Piaget, Wellman contended that even early childhood theories make an ontological distinction between mental things, like thoughts and dreams, and physical things, like balls and clocks. Yet where these theories run aground is that they fail to recognize that beliefs are representations of reality mediated by perception. Wellman proposed that the child first develops a desire theory, which posits that individuals have unique wants and generates explanations of behavior in those terms. Next, the child develops a conception of individual beliefs and eventually a belief-desire theory that allows for actions to be explained in terms of either beliefs or desires. When relying on this theory (shown in Fig. 9.1), children generate explanations that reference either an actor's desire or belief. They can conceptualize distinct individual beliefs and hence can solve the false-belief task. Finally, the next major change is toward a theory in which belief and desires are assumed to be jointly necessary for intentions and intentions to be the proximal cause of action. Children relying on this theory generate explanations more like those of adults (see Fig. 9.2). The evidence for these theory changes comes from age-related tendencies in children's performance on prediction tasks and from an analysis of the content of their spontaneous explanations (Bartsch & Wellman, 1995). Gopnik and Wellman concentrated on rendering the Fig. 9.1. A child's lay theory of mind (Wellman, 1990).

major changes in children's abstract framework theories rather than more concrete specific theories. These researchers have relied on data about systematic age-related change in several kinds of performances: success in predicting behavior of others in situations that require sophisticated understanding of mental states, and the content in children's spontaneous explanations. As in the philosophical literature, the emphasis in this research has been on lay reasoning about mental states of other individuals. Challenges The rivals to the theory theory in developmental psychology closely parallel those in philosophy. First, from a nativist perspective, Leslie (1994) argued that the knowledge structures guiding children's understanding of causes are innate modules dedicated to particular kinds of causation. The primary evidence for modules as opposed to theories comes from demonstrations that infants are sensitive to highly specific stimulus patterns that mark types of causation (e.g., mechanical events, intentional actions) before the relevance of such patterns could have been induced from experience (Leslie, 1988, 1994; cf. White, 1990). Also taken as evidence for modules are studies of autistic children that suggest highly specific impediments in understanding mental states. Baron-Cohen and others have argued that this mind-blindness is best explained by a genetically based deficit in the module for understanding intentional acts, although alternative accounts can be framed in terms of the theory theory (see Currie, 1996, for a recent review). In summary, the evidence raised by nativist researchers points to some role of innate mechanisms. However, whether this role excludes a role of empirically defeasible theories is an open question. Many theory theorists posit a role for innate structures as well as theories. Although parsimony suffers from positing both kinds of processes, it is also the case that purely modular accounts do not afford parsimonious accounts of

< previous page

page_147

next page >

< previous page

page_148

next page >

Page 148 Fig. 9.2. The adult's lay theory of mind according to Wellman (1990).

qualitative shifts in processing with age, needing to resort to competence/performance distinctions or ad hoc arguments about when innate mechanisms come on-line in development. Hence, parsimony does not clearly favor one camp over the other. A natural question for the theory theory raised by nativists is why children converge on the same kinds of understandings at around the same times; experimenting scientists surely do not seem to do as much. Gopnik (1996) argued that children have more in common with each other than do scientists: Given that children share the same innate concepts as a point of departure and have similar early experiences (i.e., parental caretaking), it is no surprise that development is largely homogenous. Astington (1996) and others emphasized a Vygotskian perspective that social experience and acculturation is a major source of theories. Consistent with this perspective, a number of researchers have found that social experience variables associated with increased acculturation – contact with parents and siblings – lead to earlier success at false-belief tasks (Dunn et al., 1991; Perner et al., 1994). In a second form of challenge to theory accounts, Harris (1989, 1991) adopted the perspective that we simulate others through our self-understanding. He argued that empathetic, imaginative projection – rather than conceptual understanding – is at the heart of how children comprehend others' minds. Even very young children, according to Harris, can predict some behaviors given wants and, given behaviors, ascribe some wants as explanations, all through simulating what it would be like to be the person involved. All of this, Harris claims, can be done without explicit knowledge of belief/desire or theory concepts. Errors like those in the false-belief task might result from projected self-understanding rather than from a limited theory of action. Another important source of evidence for this position is the developmental priority of mental state self-ascriptions as opposed to other-ascriptions. Nevertheless, defenders of the theory theory reply that the facility with self-understanding may simply reflect greater expertise with self than any other given person (Gopnik, 1993). In summary, developmental psychology has produced a rich tradition of theory about lay theories that change with age. However, many of the central pieces of empirical evidence for and against this proposal are vulnerable to alternative inter-

< previous page

page_148

next page >

< previous page

page_149

next page >

Page 149 pretations. THEORY THEORIES IN SOCIAL PSYCHOLOGY Although the idea that adults understand themselves and others by applying theories dates back to James (1890), the first elaborate accounts of social understanding in terms of lay theories came from the mid-century scholars who applied Gestalt theory to social perception. This literature is immense and so we only review contributions where the analogy to scientific theories is taken seriously. To organize our review, we draw on the contrast between more general framework theories and more specific domain theories. An early and influential contribution was Ichheiser's (1943) account of recurrent misunderstandings in social perception. Ichheiser noted a tendency “to interpret in our everyday life the behavior of individuals in terms of specific personal qualities rather than in terms of specific situations…based on the [ideological] presupposition of personal determination of behavior (as opposed to the situational or social determination of behavior)” (p. 151). In his famous analysis of social perception, Heider (1958) drew on Ichheiser's ideas for his proposal about the general framework theories that guide the lay person's understanding of action. Heider imputed to the lay person a framework theory much like Lewin's field theory, in which action results from the dynamic interaction of personal and environmental forces. In the path diagram with which he expressed the proposed lay theory, causal interactions are seen in the convergence of paths (Fig. 9.3). Heider proposed not only this broad framework theory, but also, embedded within it, specific models for particular kinds of behavior, such as achievements or emotions. With regard to evidence, the framework theory was defended by allusion to general patterns in experimental findings; the more specific proposals were supported by more direct evidence of how these conceptions were represented in folk tales, literature, and ordinary language. Proposals About General Framework Theories A great deal of research energy has been devoted in the intervening decades to testing consequences of general framework theories. The bias to see “behavior [as] caused primarily by the enduring and consistent dispositions of the actor, as opposed to the particular characteristics of the situation to which the actors responds” (Nisbett & Ross, 1980, p. 31) has been called the fundamental attribution error (Ross, 1977). However, whereas Fig. 9.3. The lay person's theory of action (Heider, 1958).

Ichheiser asserted that the bias derived from ideologically tinged framework theories, Heider did not. Rather, he emphasized Gestalt perceptual processes that draw attention to factors internal to the actor. Close in spirit to Heider's account, most influential reviews since have emphasized the role of judgment heuristics (Gilbert & Malone, 1995; Ross, 1977). However, cross-cultural comparisons (see Krull, chap. 13, this volume) have recently reawoken interest in the contribution of theories to the fundamental attribution error. Before turning to proposals about more specific theories, it is worth noting in proposals about general framework theories two points of contrast between social psychology, on the one hand, and developmental psychology and philosophy, on the other hand. A key principle for social psychologists is that the perceiver strives to attribute behavior of another person to stable, distal properties. As Heider put it, “Man grasps reality, and can predict and control it, by referring transient and variable behavior and events to relatively unchanging underlying conditions, the so-called dispositional properties of his world” (p. 79). Whereas developmentalists portray a perceiver who attends to the psychological states proximal to another's action, social psychologists portray a perceiver who tries to read something from the actor's behavior about the actor's enduring dispositions or the stable properties of the social situation. In part, this difference may reflect that adults make different inferences owing to different theories. Wellman (1990) proposed that the adult's lay theory is far more complex in its ontology and causal principles than the child's (compare Fig. 9.2 to Fig. 9.1), such that inferences about proximal causes of behavior such as the actor's beliefs and desires are traced back to ultimate causes such as traits. As Wellman (1990) wrote “Specific beliefs and desires are often explained as springing from such larger traits (the desire to wear

< previous page

page_149

next page >

< previous page

page_150

next page >

Page 150 Fig. 9.4. Distinction between theories of intentional action and unintentional behavior (Malle, 1998).

pretty clothes may spring from a vanity or, perhaps, from an insecurity about one's attractiveness). Because of this grounding of specific beliefs and desires in traits, traits figure prominently in causal explanations of actions’’ (p. 114). This emphasis on learning about stable properties may reflect a more general orientation of adults to be more pragmatically driven rather than curiosity driven in the inferences from another person's behavior. Perhaps the converse of social psychologists' emphasis on the inference of stable traits is their lack of attention to inferences about fleeting intentions. A core idea among philosophers and developmentalists is that perceptions of intention are pivotal in social understanding. Yet social psychologists have only focused intermittently on inferences about the actor's intentions. An influential model by Jones and Davis (1965) accorded a central role to ascriptions of intent, but the specific claims of this model did not fare well empirically (see Jones, 1979). For the most part, inferences about intent have been prominent in social psychology only in discussions of responsibility and culpability assignment (Shaver, 1985). Similarly, there has been intermittent attention to the notion that a distinction should be drawn between responses to intentional actions as opposed to mere occurrences of behavior (Buss, 1978; Kruglanski, 1975). This argument has been elaborated in recent work by Malle and Knobe (1997a, 1997b), which proposed that perceivers have distinct theories for explaining intentional actions as opposed to unintentional behaviors. A sketch of this framework can be seen in Fig. 9.4. Although the tests of predictions from Malle's distinction are as yet few, it appears to be a promising way to reconcile the tradition in social psychology that behaviors are directly attributed to traits and the tradition in other scholarship that actions are understood first in terms of intentions. Proposals About Domain-Specific Theories A broad characterization of the history of lay theory research in social psychology might be that theorists have tried to winnow and formalize the ideas in Heider's eclectic and sprawling treatise. In an influential article, Kelley (1972a) provided precise proposals of specific implicit theories in a matrix formulation.3 An assumption of multiple sufficient causes (MSC) was posited as

< previous page

page_150

next page >

< previous page

page_151

next page >

Page 151 Fig. 9.5. Lay theory that interpersonal behaviors have multiple sufficient causes (after Kelley, 1972a).

underlying the understanding of interpersonal behavior (Fig. 9.5 presents this implicit theory in matrix formation and in a causal path formulation). Briefly, perceivers may view behavior as having several sufficient explanations, with each acting independently, and only one of which is needed to produce the behavioral effect. For instance, a person's nervous behavior may be caused by a situation (a high-pressure job interview) or a disposition (an anxious person); both are not necessary to explain the behavior. Kelley proposed a list of implicit theories at this level of generalization, described patterns of inference that should follow from each, and linked each to a domain of behavior. The path diagram formulation makes clear that the lay person purportedly assumes that each factor is sufficient, to cause the behavior. This marks a difference from Heider's insistence that lay perceivers assume factors interact to produce actions (Fig. 9.3). One area of evidence for the MSC theory concerns the discounting principle , which holds that when one gains evidence for one causal factor being present, this reduces one's confidence in the presence of an alternative causal factor. If one learns a nervousacting person is on the way to a job interview, there is less need to posit dispositional anxiety. An interesting twist in Kelley's work is that the analogy to scientists is used both descriptively and normatively. The discounting principle is presented as a rational inference akin to those that would be made by “a good scientist” reasoning from evidence and theory. Following Kelley, a number of researchers have worked on the specific theories that guide inferences in particular content domains. The work of Weiner and colleagues (for a review, see Weiner, 1986) has greatly clarified the set of assumptions that guide attributions for achievement. Reeder and colleagues (Reeder, 1993; Reeder & Brewer, 1979) have closely analyzed the inferences drawn from schemata for domains such as achievement and morality. This work highlights asymmetries in the evidence needed to draw inferences about traits on different ends of a dimension (e.g., a single act of cheating can lead to an inference of dishonesty, whereas a single act of rule following is unlikely to lead to an honesty inference). Other researchers have tried to understand individual differences in social understanding in terms of differences in domain theories. Research by Dweck and colleagues (see Dweck & Leggett, 1988; Dweck et al., 1995, for reviews) focused on a divide between two ways of thinking about a particular domain of behavior. Starting with the domain of intelligence, Dweck argued that lay people fall into two basic groups: entity theorists, who hold that intelligence is fixed, and incremental theorists, who hold that intelligence is malleable. Dweck and colleagues showed that this individual difference in belief affects both teachers' and students' judgments and actions in the classroom (Dweck & Leggett, 1988). Related research showed that perceivers differ in theories of moral behavior, and these assumptions carry with them patterns of reasoning about moral behavior (Dweck,

< previous page

page_151

next page >

< previous page

page_152

next page >

Page 152 Chiu, & Hong, 1995; Hong et al., 1997; Chiu et al., 1997). Elsewhere among the social psychologists studying organizations, the construct of implicit theories of organization has proved a useful way to understand how organizations differ in the assumptions that members share (Downey & Brief, 1986). As with theories about personal characteristics, such as intelligence and morality, these theories of group or organizational characteristics have been empirically linked to consequences in perceivers' thoughts and actions (Baron, Hannan, & Burton, 1999). Proposals About Highly-Specific Theories Thus far, we have discussed highly general framework theories and more specific theories that apply to a particular kind of behavior or characteristic. The lay theory construct has also been used to refer to even more highly specific knowledge structures delimited to particular actions, social groups, or persons (for a review, see Wegner & Vallacher, 1977). The earliest use of the construct at this level were studies identifying the ideographic theories of personality held by particular perceivers by Rosenberg (1972) and others. This research is covered only briefly here because it is central to another chapter in this volume (see Chen, chap. 8). A seminal research program on highly specific theories was the work of Schank and Abelson (1977) on the scripts used to understand routine activities, such as having a restaurant meal. Although more concrete than general theories, these bear a resemblance to theories because they are abstracted from aggregations of explanations for observed instances of the behavior (Read & Cesa, 1991). Others have proposed that stereotypes are not simply lists of features, but theories about the structure of factors causing behavior by a particular group (Wittenbrink, Gist, & Hilton, 1997). More specific yet, some scholars have proposed that perceivers rely on lay theories of particular other individuals (see Park et al., 1994). Some researchers of knowledge structures representing particular events rather than classes of events or objects (e.g., Pennington & Hastie, 1993) have described them as stories rather than theories. Can the criteria that we have distilled help us distinguish highly specific lay theories from stories? The functional criteria may be of little help because stories and beliefs can serve functions such as explanation, prediction, and interpretation. However, structural criteria provide some distinctions. Accounts of scripts (Schank & Abelson, 1977) and action stereotypes (Wittenbrink et al., 1997) are abstracted from observations; by describing interactions of causes, they bring coherence to one's understanding of the mechanisms producing a particular behavior. Stories, however, are not abstracted; they are in the same language as the evidence at hand. Challenges to Theory Theories As in philosophy and developmental psychology, accounts of social understanding in terms of lay theories have been challenged by proponents of alternative accounts. Again these challenges come from two fronts. On one side are researchers who argue that the lay perceiver has no need to acquire elaborate learned theories because social understanding proceeds through innate capacities to perceive relevant social cues. For example, the Gibsonian approach of Zebrowitz and others (for a review, see Zebrowitz, 1998) has focused on inferences that are drawn based on properties of faces. However, although some dispositional judgments may follow directly from facial evidence without the need of folk theories, it is doubtful that other kinds of inferences (e.g., judgments about a target's beliefs) could be explained in this way. Elsewhere, researchers argue that, instead of developing theories, perceivers construct social understandings through associational processes. The connectionist approach to social cognition (Miller & Read, 1991; Read & Miller, 1993) describes judgment as a process distributed throughout a network of associations shaped by continuous input and output. Given a set of inputs, the network settles on an output solution through a process of selective activation and feedback. Such a process is implicit and requires no semantics or knowledge structures such as theories. Like simulation accounts, connectionist judgments are not a matter of knowing but a matter of doing. In some sense, this poses an alternative to lay theory descriptions of decision making. However, little research has been done to directly compare these accounts for given domains of judgment. Although these rival accounts of social understanding have been raised in social psychology, they do not present a serious threat to the widely accepted accounts of social understanding in terms of lay theories. The greatest challenge has come from critics who have pointed out that, despite the wide acceptance of some proposals concerning causal theories, the evidence from empirical tests is quite weak. In other words, skeptics demand more detailed evidence that people possess and use knowledge structures in this form. For example, one of the most influential proposals is that lay people possess and widely apply the specific theory that interpersonal behaviors have multiple sufficient causes

< previous page

page_152

next page >

< previous page

page_153

next page >

Page 153 (Kelley, 1972a). This theory that purportedly underlies people's tendency to discount one causal factor given the presence of an alternative factor, and researchers have examined patterns of attributional inferences to look for evidence of the underlying implicit theory. Yet such tests have not converged in support of Kelley's proposal (for reviews, see Shaklee, 1983; Surber, 1981) with some research programs concluding that participants err by discounting less than would be rational (Jones, 1979) or more than would be rational (Calder & Staw, 1975; Lepper, Greene, & Nisbett, 1973). A critique of research on lay causal theories by Fiedler (1982) suggested that “first inferring a causal schema from judges' modal attributions of an event or action and then explaining the same attributions by the same schema” (p. 1003) is circular and that progress requires that the theories be measured directly and independently of attributions. The problem of whether implicit theories can be measured directly in themselves rather than merely detected indirectly in their consequences raises a number of interesting questions. Many researchers have assumed that theories are implicit – beyond the lay person's declarative knowledge. This has been acknowledged, in part, by developmental researchers who use paradigms such as the false-belief task (rather than asking children, ‘‘Do you believe in beliefs?”). There is the danger that simple self-reports of one's beliefs might tap metacognitions that differ from people's actual working theories. Also, as stereotype researchers have long been aware, social desirability concerns often lead people's espoused theories to diverge from their implicit theories. However, some researchers have tried to measure an individual's theory independently of the attributions purportedly based on the theory. Approaches to Direct Measurement of Lay Theories The predominant approach among researchers who have attempted to directly measure lay theories is to design belief or attitude inventories. Dweck, Chiu, and Hong (1995) developed measures of this sort. Although not measuring an individual's theory in all its nuances, they suffice to sort individuals into the general categories of entity versus incremental theorists about a given domain. Developmentalists have designed tests of whether children have acquired the capacity to make decisions premised on an understanding that interpersonal behaviors have multiple causes that operate independently (Morgan, 1981). Although these techniques allow for a less circular research strategy than the one that Fiedler critiqued, they do not measure lay theories in enough detail to test Kelley's (1972a) argument that people draw the discounting inferences that follow rationally from the theory triggered in a domain. In another approach, Morris and Larrick (1995) suggested that Kelley's proposals about implicit theories can be formulated in terms of subjective probability and, by interrogating individuals about a series of specific probabilities, a researcher can elicit the implicit theory that the individual applies to a behavioral domain. In addition to affording a way to measure theories, the probability formulation has two other important advantages: First, it is more descriptively flexible than a matrix or path-diagram notation. Second, it enables an explicit normative analysis of the discounting inferences implied by a particular theory about how the multiple causal factors in a domain interact to produce a behavioral outcome. For example, the lay theory of multiple sufficient causes comprises the following assumptions: The sufficiency of two causal factors, C1 and C2, for an effect, E, which can be represented as the conditional probability of the effect given each cause, P(E | C1) and P(E | C2). The independence of the causal factors, which can be represented in terms of whether the conjoint probability of the two factors equals the product of their base rates or prior probabilities as opposed to equaling a higher quantity (in the case of positively associated factors) or a lower quantity (in the case of negatively associated factors). The number and prevalence of other, third causes for this effect or, in other words, the exhaustiveness of this pair of causes, which can be represented by the difference between the total probability of E minus the probability of E given C1 or C2. Morris and Larrick (1995) showed analytically that discounting follows from MSC and that the amount of discounting depends on the prior probabilities of C1 or C2. They analyzed the discounting inferences from lay theories comprising all permutations of assumptions on these dimensions to uncover the boundary conditions of where discounting is logically implied and where it is not. Morris and Larrick (1995) suggested that variations of lay theories may account for behavioral domains where slight or extreme discounting inferences have been observed. They tested this in a few cases by measuring each participant's lay theory through a series of questions. For example, the strength of a causal mechanism or level of sufficiency of cause C for effect E can be measured with a question such as, “What is the likelihood that E would happen, given that C is present?” or more

< previous page

page_153

next page >

< previous page

page_154

next page >

Page 154 extentionally, “Of 100 cases where C is present, in how many of these cases would E be present?” Morris and Larrick (1995, Study 1) examined the lay theories participants applied to the behavior of writing an essay in favor of an unusual political position, and found that participants tended not to assume that external instructions to write such an essay are a sufficient cause of doing so. Consistent with previous studies, when participants were later asked to make an attribution for a particular case of essay writing, they tended to attribute it to the student's attitude, and this attribution was not entirely discounted on receiving information about the presence of the situational pressure of instructions from a professor to write the essay. However, Morris and Larrick's participants did not fail to reason rationally; their level of discounting matched that which normatively followed from their idiographically measured theories (see Wegener & Petty, 1997, for a discussion of similar approaches). Their tendency to attribute to the dispositional factor even in the presence of the situational factor did not result from an inferential bias, such as anchoring and insufficient adjustment. Rather, it resulted from the application of a theory that assumed the situational cause was not by itself perfectly sufficient – an assumption that is incorrect about this particular experimental setting although it may be generally correct about most contexts in which essays are written. In initial studies, Morris and Larrick (1995) found perceivers' attributions correlated highly with what would be normatively appropriate given their assumptions about the sufficiency and independence of factors. In subsequent studies, Morris and Larrick (1998) found that people's inferences are highly sensitive to some normatively relevant parameters, such as the association between two causal factors, C1 and C2, and less sensitive to other parameters, such as the magnitude of other, third causes of an effect. This suggests that people's theories or schemata may not be well elaborated with regard to the latter dimension. In other words, schemata may focus on the relationship and properties of two focal causes. That is, people focus on probability of the effect owing to the two focus causes and not on its complement, the area of the effect accounted for by other causes. In any case, the technique of representing and measuring schemata in terms of subjective probabilities holds promise for more nuanced descriptions of the packages of assumptions people bring to bear on particular domains of behavior and on the inferences drawn from each. Summary As the foregoing review makes clear, the lay theory construct has played a long-standing and central role in the social-psychological literature. Our review yields several summary insights. First, it seems useful to distinguish levels of abstraction or generality in lay theory proposals. Just as framework theories in science are less likely to be overturned by empirical observations, the same may be true for framework theories in lay understandings. Also, it may be the case that researchers should expect different kinds of evidence for lay theories at different levels of abstraction. Just as framework theories have less direct contact with empirical experience, they may be less accessible to introspection. Proposals about general framework theories may have to rely on much more indirect evidence than proposals about specific domain theories. Importantly, proposals about knowledge structures at different levels of abstraction need not be rival accounts, but rather may be complementary. Second, against the backdrop of scholarship on lay theories in other disciplines, the social-psychological approach reveals some distinctive, perhaps arbitrary, assumptions. This lends support to recent calls for questioning the assumption that lay perceivers single-mindedly strive to infer personality traits (Kihlstrom & Hastie, 1997). The idea that perceivers distinguish beliefs, desires, and intentions of actors may prove useful in resolving a number of social psychological questions (Ames, 1998). If, as philosophers and developmentalists propose, lay perceivers work within the framework of a belief-desire folk psychology, this calls for a reconsideration of core concepts such as traits and situations. The social perception of traits may be best seen not as detecting some essence that manifests itself in behavior, but as a set of mental events, of thoughts and wants, that perceivers impute to others. To understand how perceivers reason about the effects of situations on others (e.g., the fundamental attribution error), we can consider how perceivers see targets' beliefs and desires as situated in particular contexts. Finally, there is a benefit to considering how the notation in which lay theory proposals have been instantiated has encumbered scholarship. When Kelley's proposals are converted from matrix notation to path-diagram or probabilistic notation, some of the assumptions of the model become more explicit. For example, the feature that two causal factors are each independently sufficient to produce a behavior (represented by two independent arrows leading to the behavior) differs from the notion in Heider's thinking that lay perceivers understand behavior as emerging from an interaction of factors. Morris and Larrick (1995) found that this feature of Kelley's model does not square with their participant's

< previous page

page_154

next page >

< previous page

page_155

next page >

Page 155 actual theories, in that situational factors were not seen as perfectly sufficient, in isolation, to produce the behavior. To name another example, Kelley's (1972a) matrix formulation, drawn from his (Kelley, 1967) earlier ANOVA cube model of induction, carries the assumption that the two factors are statistically independent without making this parameter explicit. A subjective probability notation makes this assumption and its consequences transparent (Morris & Larrick, 1995). The lesson is not that one notation is superior, but that theorists should guard against the tendency to inherit or ignore particular assumptions as a function of whether the notation used makes those assumptions convenient. IMPLICIT THEORIES AND CULTURAL DIFFERENCES Whereas early work on cultural differences in social psychology relied heavily on claims about differences in values, researchers have increasingly posited that many cultural differences are better understood in terms of lay theories. We focus on the research most relevant to questions of social understanding that have been central to debates about the theory theory in the disciplines we have reviewed. A first wave of such research has focused on the extent to which the lay theory underlying the error of overattributing personal dispositions is a product of the individualism in North American culture. Lay Theories About Persons Several recent ethnographies have proposed that lay categories of person radically differ in their most basic ontological assumptions. Rosaldo (1980) studied the Illongot people in the Philippines and found different notions about internal states that cause action, such as rinawa, which translates as heart. Rinawa is considered the center of thought and can be found in plants and other aspects of nature; it leaves the body during sleep and gradually disappears with aging. Elsewhere, Lutz's (1985) work on the Ifaluk of Micronesia turns up other conceptions, with the center of thought located in the gut. D'Andrade (1987, 1995) used ethnographic interviews to examine American adults' lay theory of mind. When appraising the folk theory that he depicted, D'Andrade (1995) emphasized the striking extent to which it holds thinking to be visible to consciousness and under an individual's volitional control. Lillard (1998) reviewed these and other “ethnopsychologies” and concluded that there is substantial variety across cultures in conceptions of person and mind, as well as the casual relations to action. However, researchers using more structured comparative methods have questioned whether the espoused lay theories uncovered by ethnographic methods actually direct cognitions and emotions (Mesquita et al., 1996). As yet there is little psychological evidence to support claims about differences in the most general folk framework theories of person. Far more research has focused on whether cultures differ in the more specific lay theories that guide social understanding in particular domains. The notion that cultures differ in folk theories of a particular domain of life has been a long-standing idea in ethnography. These claims focus, for the most part, not on the basic ontology of causes of behavior (i.e., beliefs, desires, social pressures, etc.), but on the mechanistic beliefs about the strength and configuration of these factors in producing particular kinds of actions. For example, ethnographers who have studied the accounts given for everyday behavior in Chinese society have described a tendency to refer to an actor's social context, rather than to personality characteristics, that reflects a society-centered conception of social behavior (Hsu, 1953). Ethnographic (Selby, 1974) and comparative (Miller, 1984) studies of everyday explanations in other non-Western societies have similarly found that there is a greater tendency to account for an individual's deviance in terms of the individual's social context rather than in terms of the individuals internal, stable personality traits. Morris and Peng (1994) argued that lay perceivers in the individualistic North American cultural tradition and the more collectivist, Confucian tradition differ in their implicit theories and consequently in their attributional biases. In the United States, a theory centered on the autonomous person channels attributions to an individual actor's internal, stable properties; in China, a theory centered on situations and groups directs attributions to the actor's embeddedness in groups and roles. A number of studies have been conducted in the last decade to investigate the extent to which these differences in the ethnographic record actually reflect theories. To assess this, it is helpful to consider the criteria that distinguish theories from other kinds of knowledge structures. The structural criteria that we have suggested for a lay theory is that it specify an ontology, define causal mechanisms, describe their coherence, and do so at a level abstracted from any particular case. Perhaps the only structural criterion that is directly tested in the cross-cultural literature is abstraction. Researchers have striven to prove that cultures differ not merely in particular beliefs about the causes of a specific action, but rather in more abstract, general beliefs. The standard method has been to demonstrate that a pattern – the greater emphasis on personal traits rather than social situations

< previous page

page_155

next page >

< previous page

page_156

next page >

Page 156 by North Americans compared with East Asians – occurs across many different examples of behavior (Choi & Nisbett, 1998; Miller, 1984; Morris, Knowles, Chiu, & Hong, 1998; Norenzayan, Choi, & Nisbett, 1998). The hypothesized pattern should be as broad as the purported underlying theory but not broader. For example, Morris and Peng (1994) showed that the influence of culturally varying social theories is seen in perceivers' interpretations of the behavior of social animals but not the behavior of mechanical objects.4 Moving from the structural to the functional criteria for a theory, the evidence is much more direct. A knowledge structure functions as a theory if it enables prediction, explanation, and interpretation. Findings with all three kinds of tasks supported the hypothesis that North Americans and East Asians work with the differing lay theories of person proposed earlier. In a prediction task, Norenzayan et al. (1998) found across many specific behaviors that American participants were less likely than Korean participants to predict that a behavior would follow from a salient situational cause. A number of studies measuring patterns of explanation have found predicted differences. Morris and Peng (1994) found that different explanations for the same crimes are offered in English-language as opposed to Chineselanguage newspapers: Regardless of whether the criminal is American or Chinese, the mainstream American paper dwells more on personality dispositions than does the Chinese-language paper. To eliminate alternative explanations, Morris and Peng followed up this content analysis of archival data with a survey of Chinese and American graduate students at an American university, all of whom responded in English. Again, North American respondents were more likely to trace criminal acts to the criminal's internal dispositions, whereas Chinese perceivers were more likely to point to the criminal's social context. Several studies have employed paradigms based on that of Jones and Harris (1967) to test whether the well-known tendency of North American perceivers to be relatively insensitive to situational constraints on behavior is less marked among East Asian perceivers. Studies with East Asian participants in typical procedures manipulating the presence of subtle information about situational constraints replicate the typical pattern (Krull et al., 1996; Masuda & Kitayama, 1998). However, when the situational factor is made more salient, Choi and Nisbett (1998) found that Korean perceivers are more sensitive than Americans to its presence versus absence. In a similar study with a different design and participant population, Morris, Knowles, Chiu, and Hong (1998) found that American participants' tendency toward dispositional attributions was virtually unchanged by the introduction of information about situational constraints on the actor, whereas for Hong Kong participants, situational cues dramatically reduced dispositionalism. Finally, other studies have reported findings suggesting that interpretation of perceived events differs across cultures in the way that would be expected if perceivers in the two cultures were relying on different theories. Interpretation, in this sense, largely involves deciding what deserves attention and what can be ignored. Studies that have compared how persons are described reveal that North Americans are more likely to refer to internal personality traits when describing themselves (Cousins, 1989) or when describing others known to them (Shweder & Bourne, 1982). Other researchers have found that cultural individualism is associated with a tendency to make spontaneous inferences to personality traits from observations of behavior (Duff & Newman, 1997; Newman, 1993). The argument is that acculturation into individualistic theories results in automatized habits of interpreting behavior through personality trait inference. Many social psychologists have argued that trait inference is automatized on the basis of findings (with North American participants) that perceivers under cognitive load show more dispositional interpretations of behavior (e.g., Gilbert, Pelham, & Krull, 1988). Apparently, perceivers burdened with performing unrelated cognitive tasks at the same time as a social perception task have fewer resources for drawing nonautomatic inferences to the situational factors influencing an actor's behavior. Yet it has long been known that automatic inferences to situational factors are also possible (Krull, 1996; Quattrone, 1982). Knowles, Morris, Chiu, and Hong (1998) argued that the impact of cognitive load should be qualified by culture because the automatized procedures of Chinese perceivers is more balanced between personal and situational causes. That is, busy Americans become more dispositional because they fall back on the automatized trait inference processes that they have been socialized into (see Uleman, Newman, & Moskowitz, 1996), yet busy Chinese perceivers will fall back onto a mixed set of processes. In several studies adapting the Gilbert et al. (1988) paradigm, Knowles et al. (1998) found that the dispositionalism of Americans increases with cognitive load, whereas the dispositionalism of Chinese participants is unaffected by cognitive load. A final set of studies makes particularly clear the role of theories in guiding the interpretation of observed events. Most evidence concerning the consequence of culture relies on correlational or quasi-experimental methods, in which an individual's dispositional value

< previous page

page_156

next page >

< previous page

page_157

next page >

Page 157 Table 9.1: Functional Criteria for Theories and Cross-Cultural Evidence Understanding personsUnderstanding groupsPredictionCompared to Americans, East Asians makestronger predictions of actions given salientsituation causes (Norenzayan, Choi, &Nisbett, 1998)Compared to Americans, East Asians make strongerpredictions of group responsibility (Morris, 1993)ExplanationIn explaining individual deviance, NorthAmericans place greater weight than Asianson personal dispositions as opposed toexternal factors (Peng & Morris, 1998)North Americans are less sensitive to thepresence of salient situational constraintsthan are East Asians (Choi & Nisbett, 1998;Morris, Knowles, Chiu, & Hong, 1998)In explaining group transgressions, East Asians placemore weight than Americans on group dispositions asopposed to external factors (Menon et al., 1998)InterpretationBicultural participants show a bias towardindividual attribution when primed withAmerican cultural symbols and a bias towardgroup attribution when primed with Chinesecultural symbols (Hong, Morris, Chiu, & Benet,1998)East Asians make more reference than Americans tosocial groups when describing themselves (Peng &Morris, 1998) and when describing others (Miller, 1987)In describing ambiguous negative outcomes inorganizations, North Americans are more likely toconstrue an individual as the key actor while EastAsians are more likely to construe a group as the keyactor (Menon et al, 1998)Need for closure increases Chinese bias toward groupdispositions and American bias toward individualdispositions (Chiu et al, 1998) orientation or an individual's nationality is the independent variable. Along with the limitations of these methods, research of this sort has led to a limited conceptualization of the relation of individuals to cultures, positing that each person has a unitary and static cultural orientation. Moving beyond these methodologies and conceptualizations, Hong, Morris, Chiu, and Benet (1998) used a method that effectively manipulates the operative culture in bicultural participants. Participants selected for their acculturation into both traditions in Hong Kong (Study 1) and California (Study 2) were randomly assigned to be primed with icons of either American culture (Mickey Mouse, the Capitol building, etc.) or Chinese culture (a Dragon, the Great Wall, etc.). Results show that this manipulation shifted perceivers' interpretation of an event entirely unrelated to the objects in the primes, indicating that an abstract theory is triggered by the prime and becomes a lens for interpreting the subsequently encountered event. In summary, as Table 9.1 shows, evidence on cultural differences in understanding of persons meets the criteria for positing the involvement of lay theories. Lay Theories About Groups A recent wave of studies has shifted the focus from how perceivers understand individuals to how they understand groups. Consistent with the arguments about collectivism and social context reviewed earlier, the notion is that Confucian E. Asian cultures regard groups as having stable properties that confer to them agency or autonomy. Some historical evidence for this is seen in that collective responsibility is a distinguishing feature of traditional Chinese law. Starting from 746 BC, the system of yuan zuo (holding offenders' superordinates, kinsmen, and neighbors responsible for their crime simply because they are related to the offenders) was widely practiced in China (see Nishida, 1985; Zhang, 1984, for documentation of the prevalence of yuan zuo ). The rationale underlying this practice was the belief that the would-be

< previous page

page_157

next page >

< previous page

page_158

next page >

Page 158 Fig. 9.6. References to individuals and groups in newspapers explanations for negative outcomes in organizations (reprinted from Menon et al., 1998).

offender's ingroup had the obligation to monitor his or her behavior, and therefore should have been able to prevent the crime. In studies comparing the social judgments about groups made by East Asian as opposed to North American perceivers, a pattern of greater attribution to dispositions of the group appears across a wide variety of particular cases (Chiu & Hong, 1992; Morris, 1993). A recent line of research has examined cultural differences in attributions to individuals and groups with the goal of demonstrating that the East-West difference in attributions to individual dispositions reflects a theory of person rather than a domain-general difference in goals or style of social perception. Across a range of particular outcomes (Menon, Morris, Chiu, & Hong, 1998; Morris, Menon, Chiu, & Hong, 1998), in analyses of newspaper articles and in surveys, individual attributions are endorsed more by Americans and group attributions are endorsed more by Chinese (see Fig. 9.6). The pattern of greater Asian dispositionalism with regard to groups has been observed even in interpretations of the behavior of social animals (Menon et al., 1998). In summary, there is good reason to conclude that the theory underlying these differences is abstracted from any particular content example. We can move on to review how the emerging evidence concerning culturally varying theories of groups fares with regard to functional criteria for theories. The earliest studies of culture and group perception involved prediction tasks. For example, Morris (1993) asked American and Chinese survey respondents to predict whether particular groups would feel responsible after a negative outcome involving one of their members; Chinese respondents predicted more group feelings of responsibility. Although this implies that a theory of group dispositions is used to make predictions, a test of the more direct predictive use of theories would be a useful step for future research. That is, it should be that when predicting a group action from a group disposition or from a factor in the group's environment, Chinese participants are more influenced by the group disposition than are Americans. Most of the evidence for culturally varying theories of groups comes from research with explanation tasks. In a vignette study by Menon et al. (1998, Exp. 3), negative outcomes (e.g., designing an unfair compensation system) were described as following the action of either an individual or group actor. For each vignette, North American and Hong Kong participants rated a series of possible causal factors, including internal stable factors (greed), internal unstable factors (haste), external stable factors (difficulty of the task), and external unstable factors (chance). Analyses focused on the extent to which internal stable factors (i.e., dispositions) were favored relative to other kinds of factors. Results show an interaction effect of perceiver culture x level of actor: American participants were more likely to endorse dispositional factors in the individual condition than the group condition, whereas Chinese participants were more likely to endorse dispositional factors in the group condition.

< previous page

page_158

next page >

< previous page

page_159

next page >

Page 159 There is also compelling evidence that lay theories of groups guide interpretation of social outcomes. One way to see this is in what a perceiver's attention is drawn to when the actor behind an outcome is ambiguous. The consistent finding from several such studies (experiments 1 and 2; Menon et al., 1998) is that American observers pay more attention to individuals associated with the outcome and Chinese observers to groups associated with the outcome. Another way to see the role of theories in interpretation is to contrast individuals identified as high versus low in need for closure – a dimension of cognitive style that can be identified by inventories (Kruglanski, 1996; see also Thompson, Naccarato, Parker, & Moskowitz, chap. 2, this volume). High Need for Closure (NFC) is associated with greater reliance on stereotypes and other ready-made explanations to filter the facts of perception (Kruglanski, 1996). In a striking result, Chiu, Hong, Morris, and Menon (1998) found that among North American participants NFC is associated with more dispositional attributions for individual acts and not associated with any particular type of attribution for group acts, yet among Hong Kong participants, NFC is not associated with any particular type of attribution for individual acts and is associated with dispositional attributions for group acts. In summary, as Table 9.1 displays, there is evidence for culturally varying theories of groups that satisfies the basic criteria for a theory. Challenges Although we stated that cross-cultural research benefits from the construct of lay theories that developmentalists and other “theory theory” advocates have done much to refine, it is equally true that the theory theory position benefits greatly from cross-cultural evidence. One reason for this is that the standard rivals to the theory theory do not account as well for cultural differences in social understanding as they do for developmental differences. Consider first the nativist account of social understanding. It is hard to build an account of cultural differences in social understanding in terms of genetically prepared perceptual modules. If innate adaptations for social understanding exist, they have been tailored by prehistoric environments that bear little relation to current North American or Chinese cultures. Although nativist accounts cannot explain cultural differences, some researchers have pointed out that cross-cultural data do not necessarily indicate any differences in general lay theories, but only in the strength of specific causal beliefs in specific domains. In a useful commentary, Wellman (1998) suggested that cultural differences may exist in specific content theories but not in underlying framework theories. Just as framework theories in science are less susceptible to empirical data, lay framework theories may be less likely to change as a function of differences in an individual's socialization experiences. Likewise, a simulationist account of cultural differences in social understanding is unsatisfying. To explain cultural differences in other -understanding in terms of differences in self-understanding merely begs the question of why cultures differ in self-understanding. That is, one needs to explain why particular ways of interpreting self (and other actors) differ between populations sharing distinct cultural traditions. The best explanation seems to be that lay theories are represented in teachings and practices and are transmitted from one generation to the next. Approaches to Direct Measurement of Lay Theories The primary challenge to the advocates of the theory construct in cultural psychology is to clarify two questions: Are theories held implicitly or explicitly? Moreover, how specific are belief structures that differ across cultures? Researchers who conceptualize theories as explicitly held knowledge, like explicit attitudes, have attempted to develop scales like those that measure explicit belief structures such as locus of control. Results thus far have met only partial success. A pioneering study developed a scale to measure the general belief that attitudes cause behavior and found that individualist Australians were higher than collectivist Japanese participants (Kashima et al., 1992). Similarly, recent work by Nisbett and colleagues measured endorsement of dispositionalist versus interactionist accounts of behavior. They found that Koreans are more likely than Americans to reject statements expressing a purely dispositionalist view of behavior and to endorse statements expressing the view that the behavior arises from the interaction of dispositional and situational influences (Norenzayan et al., 1998). Although these fairly general measures of theories have identified clear patterns of cultural difference, they have not statistically linked these theories to attributions. Hence, the validity of these measures remains unproved. Other researchers have focused on narrower social beliefs. Menon et al. (1998) assessed beliefs about the autonomy of individuals and organized groups and found that Americans tended to endorse the autonomy of individuals but not groups, whereas the opposite was true for Chinese participants. Chiu, Dweck, Tong, and Fu (1997) assessed beliefs about the temporal stability of moral characteristics of both individuals and social

< previous page

page_159

next page >

< previous page

page_160

next page >

Page 160 institutions. Of particular interest were respondents who saw moral characteristics at one level to be fixed and at the other level to be fluid. Americans were more likely to assume that individual moral character is fixed and the social world is fluid, whereas Chinese participants tended to assume individual moral character is fluid and the social world fixed (Chiu, et al., 1997). Although these more specific measures of social beliefs showed modest correlations to attributional outcomes, the goal of statistically mediating country differences in social understanding with an explicit measure of theories remains elusive. One interpretation of this state of affairs is that general theoretical beliefs fail to predict attributions for the same reason that general attitudes fail to predict behavior, and that researchers should measure highly specific theoretical beliefs in order to predict specific attributional outcomes (see Ajzen & Fishbein, 1980). One method that may work in this sort of approach is the technique of Morris and Larrick (1995, 1998), which measures the particular theory-based expectancies that perceivers impose on an event. Another interpretation of the foregoing pattern of results is that the actual theories shared within a culture are held implicitly. Because individuals lack direct introspective access to the theories they hold, these theories are at best weakly tapped by explicit measures. To the extent that researchers believe that theories are implicit, then the best approach to establishing the causal role of theories may be manipulating theories through priming experiments rather than attempting to measure them. Work in the psychology of categorization has demonstrated that priming a theory can facilitate category learning by directing attention (Wisniewski & Medin, 1994). Priming studies have been used in a similar way to establish the role of theories in producing the patterns found in cross-cultural studies (Hong et al., 1998). The emerging dynamic constructivist approach to culture and cognition centers on this notion that cultural theories impact social understanding only to the extent that they are operative – accessible and applicable to the task; mere possession of cultural knowledge does not ensure its influence. Summary We have already noted the importance of theories as knowledge structures in guiding everyday judgments such as explanations and predictions. To the extent that we are concerned with cultural differences in such conclusions, we are well served to articulate and measure the folk theories that seem likely to underlie them. We might even go so far as to suggest that what it is to have a culture is to have a set of shared lay theories for understanding and explaining the world. If that is the case, cultural researchers should feel obliged to identify and measure important folk theories and gauge their relationship with everyday judgments. Cross-cultural findings also provide further support to some of our recommendations about future social-psychological work with the lay theory construct. We have argued that researchers have imputed an overly simple notion of traits to lay perceivers, whereas the lay understanding of factors causing behavior may be more nuanced. Research in East Asian cultures makes it clear that perceivers form impressions of personal dispositions that are context-bound and situated. Similarly, the conventional assumption about the perception of traits is that it should only occur in the case of individual actors. However, research with East Asian participants has made it clear that, to these lay perceivers, group-level traits are as plausible as individual-level traits. This is surprising to many researchers, but should it be? Group-level dispositions are no different than individual level traits: They are unobservable latent constructs underneath the regularities in observable outcomes. The lack of attention to perceived group dispositions in social perception research may reflect the extent to which individualistic lay theories have shaped our scientific theories. Studying lay theories across cultures helps us recognize the extent to which our lay theories have limited our scientific imaginations. CONCLUSION We have reviewed arguments and evidence that the ability of humans to understand each other owes to the fact that we possess and use theories. Social psychologists and cognitive scientists have discussed a variety of kinds of knowledge structures, such as associative networks, models, scripts, expectancies, and beliefs. Clearer definitions help one avoid artificial debates between proponents of constructs that should not be construed as alternative proposals. The definition we have provided clarifies that some of these constructs are synonymous with theories, some are alternatives, and some are components of theories. Our review of arguments and evidence from different fields reveals a variety of support for the theory theory position. In particular, evidence from developmental psychology helps remind adult researchers that, at one time, they too would have failed a false-belief task, having an incomplete understanding of the nature of belief. By watching children tinker with evidence, change their theories, and make predictions, the role of folk theories – perhaps taken for granted by adults – becomes

< previous page

page_160

next page >

< previous page

page_161

next page >

Page 161 more apparent. Meanwhile, cross-cultural researchers have found a pattern of consistent differences in the social judgments adult perceivers make about persons and groups. These patterns are highly consistent with the idea that specific domain theories differ across cultures, being shaped by acculturation into belief systems as well as by the ecology of behavior and causes in which the perceiver is embedded. There is no easy way to account for such differences in social understanding in terms of innate capacities or in terms of simulations. In summary, there is increasing evidence, across a variety of disciplines and topics, for the stance that we understand the social world through inferences based on specific theories about behavioral domains nested within a general framework for understanding social actors. ENDNOTES 1. The foregoing discussion of criteria from philosophy science draws in part on similar arguments by Gopnik and Meltzoff (1997) and Wellman (1990). 2. The construct of causal schemata, as discussed in development by Piaget and adult causal reasoning by Kelley, is equivalent to the construct of implicit theories. That is, schemata are abstract expectations about cause-effect relationships that are organized into a system and used for interpretation, explanation, and prediction. Further, they are subject to change. Thus, schemata are more like theories than they are like beliefs, attitudes, narratives, metaphors, or scripts. Although Piaget, Kelley, and other scholars originally described their ideas with the term schemata , we use the label theory for consistency with our argument. 3. The matrix formulation was carried over from Kelley's (1967) previous model of causal induction because he saw schemata as the residue of previous inductive activity in a domain. Perhaps because induction model is context general, secondary treatments of the schema or lay theory model have often misconstrued the extent to which Kelley claimed that one or two given schemata are used generally. In his statement of this model, Kelley (1972a) linked specific schemata to specific domains, such as achievement, and he also discussed sub-types of each. 4. This same study also tested predictions from proposals about more specific theories that would affect the extent to which American and Chinese perceivers encoded displays as social interactions as opposed to non-contingent actions of an individual near a group. These predictions were also supported: Displays involving an individual becoming connected to a group were more likely to be seen as social than displays of an individual becoming separate from a group. Displays of an individual harmonizing its movement to that of a group were more likely to be seen as social interactions by Chinese perceivers than were displays of discordant movements between the individual and the group.

< previous page

page_161

next page >

< previous page

page_162

next page >

page_162

next page >

Page 162 This page is intentionally left blank.

< previous page

< previous page

page_163

next page >

Page 163 10 From Cognition to Culture: The Origins of Stereotypes That Really Matter Mark Schaller Lucian Gideon Conway, III University of British Columbia Stereotypes matter because they have consequences. Stereotypes matter because they have specific consequences. The specific consequences result from the particular contents of those stereotypes. If you believe that folks from the American South are fond of a good slab of barbecued meat, then you may be more likely to serve smoked pork shoulder when you have your new Alabamaborn co-worker Billy over for a welcoming dinner, but this particular stereotype of Southerners is not likely to make you reconsider the rest of your guest list. Alternatively, if you believe that Southerners are racist bigots, then this stereotype probably will not influence the menu, but may make you decide not to invite the Nwazuruokeh's from next door to join you as well. Stereotypes matter because they have general consequences. The general consequences result from the fact that stereotypes are, to some degree, shared. If you are the only person who believes Southerners to be barbecue lovers or racist bigots, then the consequences of this stereotype would be minimal indeed, affecting the lives of no one but you and the Southerners whom you encounter. You might insensitively invite a finicky vegetarian over for a rack of ribs, or your coworker Billy might miss the opportunity to meet your Nigerian neighbors, but these incidents hardly comprise a social problem. However, when these stereotypic beliefs are shared by many, then the consequences of the stereotype are broad, affecting the actions of all those who hold the stereotype and the lives of the much larger set of Southerners that these many individuals encounter. None of this is news to anyone who has either a personal or professional interest in stereotypes. In commenting on several stereotypes that have historically mattered – such as stereotypes of Jews – Haslam (1997, p. 119) noted that ‘‘the stereotypes became potent because, within the groups who held them, they came to reflect and express a particular world view which dictated that group's collective behavior towards particular targets of oppression.” But, although social scientists may readily acknowledge why stereotypes matter, we do not know very much about how those stereotypes that matter come to be what they are. If we wanted to predict which exact traits and characteristics would be central to a shared stereotype of some group, we would find that the existing socialpsychological literature offers only a rough set of rules on which to offer a prediction. One challenge for the future, then, is this: We need to nourish a scientific literature that more completely offers explanations for the origins of stereotypes that really matter. How do emerging, shared stereotypes arise that have the specific contents that they do, rather than some other contents? Through what processes can we understand and predict how some individual cognitions rather than others become socially shared cultural structures? WHAT WE KNOW Prior to the cognitive revolution in social psychology, content and consensus were the explicit foci of many studies of group stereotypes. Consider the well-known “Princeton trilogy” of studies examining ethnic stereotypes (Gilbert, 1951; Katz & Braly, 1933; Karlins, Coffman, & Waters, 1969). The questions these studies addressed were explicitly: “What are the contents of stereotypes about various ethnic groups?” and “How much consensus is there in the perception of these stereotypes?” However, just as these research questions revealed the focus on stereotypes that really mattered, the results also reveal the limitations of the accompanying sociometric methods: The studies were descriptive, not explanatory. However impressively they described the content and consensus of ethnic stereotypes, the results offer nothing by way of explanation. Why these particular contents and not others? Why are these contents consensually shared? To answer questions such as these, it is necessary that stereotypes be defined at the individual level of analysis. Within contemporary psychological frame-

< previous page

page_163

next page >

< previous page

page_164

next page >

Page 164 works, a stereotype refers to some sort of cognitive representation consisting of the knowledge/information that is associatively linked to a group label – the information that is especially likely to come to mind when one perceives or otherwise thinks about the group or group members. These associative knowledge structures may include many types of information, including traits that are judged to be diagnostic of group membership, images of typical group members, script-like expectations about typical group behaviors in particular situations, and so forth (see Stangor & Schaller, 1996). With this definition in mind, important questions about content and consensus can be framed more explicitly. How is it that particular stereotypes come to be defined by particular traits and prototypes and expectations, rather than by others? And how is it that the contents of these individual-level cognitive representations come to be so similar across large populations? Why Do Stereotypes Have the Content They Do? Whereas early sociometric studies were weak on explanation, a focus on explanation has been the strong suit of social-cognitive approaches to the origins of stereotypes. Results revealing the various causal processes that explain the formation of stereotypes are too numerous now to be easily summarized within any standard review article (for partial reviews, see chapters from Macrae, Stangor, & Hewstone, 1996; Sedikides, Schopler, & Insko, 1997; Spears, Oakes, Ellemers, & Haslam, 1997). Answers to questions about content, however, are relatively rare within the social cognition literature (Schneider, 1996). It is not impossible to find answers to the questions about content, but one does have to read between the lines a bit. Esteem-Related Goals Lead to Stereotype Content That Is Negative As has been amply demonstrated, stereotypes are handy tools to have in our cognitive toolboxes (e.g., Allport, 1954; Gilbert & Hixon, 1991; Macrae, Milne, & Bodenhausen, 1994). The fact that stereotypes are functional explains why we have stereotypes at all. Of course, some stereotypes are more functional than others, and this obvious statement offers several general answers to the question of content. People want to feel good about themselves (see Blanton, chap. 5, this volume; Pomerantz, Saxon, & Kenney, chap. 4, this volume, for discussions of the self-esteem motive). Toward that goal, people desire to perceive groups to which they belong in a positive light. Evaluative perceptions of ingroups are most meaningful within a comparative context, in which ingroups are judged relative to relevant outgroups. Therefore, according to the language of social identity theory, individuals seek to establish positive distinctiveness of ingroups relative to outgroups (Tajfel & Turner, 1986). The goal of positive distinctiveness may contribute to predictable effects on the contents of stereotypes. In general, outgroup stereotypes are likely to be more negative than those about ingroups. Consistent with this line of thinking, descriptive studies on the evaluative contents of ethnic stereotypes have shown exactly this pattern – although the tendency varies depending on the social context within which stereotypes are measured, and the sensitivity of measurements to that context (e.g., Devine & Elliot, 1995; Karlins et al., 1969). More clearly, mere categorization into groups leads individuals to form beliefs about groups according to which they differentially evaluate ingroup and outgroup members on evaluatively laden traits (Brewer, 1979; Jost, chap. 6, this volume). Indeed, the power of the ingroup/outgroup context is such that it not only affects the outcomes of the information integration processes through which people form impressions of groups (e.g., Schaller, 1992a), but can actually change the very nature of those processes. For instance, Schaller and Maass (1989) found that impartial observers of groups based their eventual impressions of the groups on recalled bits of information. In contrast, recall for specific information played little role in the eventual impressions formed by people who believed themselves to be members of those groups. Instead, categorization into groups led perceivers to engage in more effortful, integrative on-line processing of group-relevant information. Aside from the general prediction about the relative negativity of outgroup stereotypes, social identity considerations suggest that whatever traits most efficiently and effectively contribute to positive distinctiveness are most likely to become stereotypic. The extent to which any particular trait may contribute to that goal is influenced by a variety of other variables relating to idiosyncratic aspects of perceivers and social contexts under which groups are perceived (see e.g., Oakes, Haslam, & Turner, 1994, for a review). If we want to predict which particular traits are likely to become central to the stereotype of some group, we might narrow down the range of possible traits along evaluative lines by first attending to the perceiver's own group membership. And then we would have to make certain ad hoc predictions on the basis of a deeper consideration of the personal and social context.

< previous page

page_164

next page >

< previous page

page_165

next page >

Page 165 Epistemic Goals Lead to Stereotype Content That Is Accurate People want to predict their social environment (see Thompson, Naccarato, Parker, & Moskowitz, chap. 2, this volume). Stereotypes provide us with a handy, albeit imperfect, basis on which to predict the behavior of individuals in the absence of any information about those individuals other than their membership in some group. Of course, if stereotypes are to serve the function of predicting individuals' behaviors, then the stereotypes must be somewhat accurate. A consideration of this epistemic need leads to the prediction that the specific characteristics that emerge as central to individuals' stereotypes are likely to be those characteristics that do, in fact, accurately convey information about a group's behavioral tendencies. This general assertion is supported by abundant research evidence indicating that stereotypic beliefs of groups are, at least to some extent, based on reality. Individuals' judgments about differences between European and African Americans in scholastic performance correlate positively with actual differences (Clarke & Campbell, 1955). Judgments about the differences between men's and women's behavior do correlate positively with the actual differences between men's and women's behavior (Swim, 1994). Across all kinds of different target groups, there is plenty of evidence that individuals' judgments about group differences are never entirely divorced from objective reality (Jussim & Eccles, 1995; McCauley, 1995). Of course, the mere observation of a correspondence between stereotypes and reality cannot distinguish between the effect of reality on stereotypes and the reverse causal effect of stereotypes on reality (self-fulfilling prophecies; cf. Jussim, 1986). Perhaps more persuasive in demonstrating the role of objective accuracy in determining the emergence of stereotype content are results from research by Ford and Stangor (1992). In an experimental investigation into the formation of new group stereotypes, they found that those traits that were most objectively diagnostic in distinguishing between two target groups (as consequences of intergroup mean differences and intragroup variabilities) were most likely to emerge as central to a set of stereotypic beliefs about groups. It is worth noting that this investigation did not merely measure ad hoc judgments about groups, but also used a spontaneous, free-response format; as such, it more truly assessed the contents of those traits that were cognitively linked to group membership. Thus, there is abundant evidence that grains of truth underlie the content of many group stereotypes. The content of stereotypes is influenced by those traits that are diagnostic in distinguishing groups from other groups. If we want to predict which particular traits are likely to become central to the stereotype of some group, we would be advised to identify those traits that really do accurately describe a group and distinguish members of that group from others. Epistemic Goals Lead to Stereotype Content That Is Inaccurate Just because stereotypes serve individuals' needs does not mean that these needs always exert the intended effects on stereotype contents. The epistemic need to accurately predict our social environment does not always guarantee that we accurately explain our social environment. People are imperfect observers, and so what we perceive as accurate may not actually be objectively so. In addition, even if our explanations accurately predict behavior, they may not be based on accurate theories or explanations about why those behaviors occur. Indeed, research into several different cognitive processes in the formation of stereotypes implies that accurate recognition of behavioral differences between groups in no way guarantees accurate inferences about the causes for those observed differences. One reason for the fallibility of inferences about underlying group differences lies in the fact that groups differ not only in terms of their overt behaviors, but also in the circumstances under which one perceives group members. The circumstantial differences between groups offer clues as to the traits that are likely to emerge in group stereotypes. For instance, across virtually every human culture, men and women tend to occupy different social roles. Men tend to occupy agentic roles that require the performance of instrumental behaviors, whereas women tend to occupy communal roles that require the performance of nurturant behaviors (Eagly, 1987). Does this mean that men and women are inherently different in their capacities for instrumental and nurturant behaviors? No. But, not surprisingly, instrumental and nurturant traits figure prominently in stereotypes of men and women, whereas traits that are far more accurately diagnostic of the social categories of men and women (e.g., bodily restlessness; Hall, 1984) tend to be more peripheral to the stereotypes. Two complementary processes help explain the emergence of these particular stereotypes. First, although people may accurately recognize that group membership and behavior are both correlated with some social role (or other situational constraint), they often fail to take this third variable into account and so fail to recognize

< previous page

page_165

next page >

< previous page

page_166

next page >

Page 166 that an observed relation between group and behavior might be entirely spurious. This failure to intuitively partial out the effects of confounding third variables can contribute to the formation of new group stereotypes (Schaller, 1992a, 1994; Schaller & O'Brien, 1992). Second, people may draw erroneous causal inferences to explain the observed correlation between social roles and group membership (cf. Yzerbyt, Rocher, & Schadron, 1997). Although the reasons that men and women have differentially occupied agentic and communal social roles almost certainly owes more to differences in physiology than to differences in personality, perceivers are inclined to assume that the choice of roles was dictated by role-congruent personality. This explanatory process also contributes to the formation of new group stereotypes (Hoffman & Hurst, 1990). Thus, the content of stereotypes is determined by those traits that most accurately describe the meaningful social circumstances in which groups find themselves. If we want to predict which particular traits are likely to become central to the stereotype of some group, we would be advised to identify those traits that describe that group's physical and social environment. Goal-less Cognitive Processes Lead to Stereotype Content That Is Inaccurate One of the most important messages emerging from the cognitive approach to stereotype phenomena is that cognitive processes entirely hidden from phenomenological awareness may contribute to the formation of group stereotypes. Although these processes almost certainly emerged in response to evolutionarily functional concerns, they proceed in the absence of any explicit stereotype-relevant goal being active in working memory. Several lines of research reveal that certain stereotypes may emerge simply as unintended side effects of these cognitive processes. Ironically, one process has its roots in relatively sophisticated statistical reasoning. People intuitively appreciate the problems of drawing confident inferences from small samples (Kunda & Nisbett, 1986; Schaller, 1992b). Consequently, the smaller the sample of behavior about any particular group, there is a greater regression phenomenon wherein observed differences between normative (frequent) and counternormative (infrequent) behaviors are judged to be less diagnostic of true tendencies toward normative or counternormative behavior. This phenomenon has important implications when drawing inferences about the characteristics of majority and minority groups. Given that there is a smaller sample of evidence from minority groups than from majority groups, the regression phenomenon is more powerful in judgments about minority groups, and so the impact of counternormative behavior on emerging stereotypes of minority groups ends up being relatively greater than its impact on emerging stereotypes of majority groups. Consequently, even in the absence of any objective correlation between group membership and behavioral tendencies, people tend to form erroneous stereotypes in which minority groups are believed to be more likely to engage in counternormative behavior (Fiedler, 1991; Fiedler, Russer, & Gramm, 1993). A second process results from the tendency to make judgments or form impressions on the basis of evidence that is most accessible from memory (Tversky & Kahneman, 1974). When forming impressions about groups, for instance, the most accessible bits of evidence are likely to be those that received the most attention at the time of encoding. It is often the rarest events that attract the most attention, and within minority/majority contexts the rarest events are those that match minority group members with counternormative (infrequent) behaviors. Thus, to the extent to which group stereotypes are based on accessible information, those stereotypes will most likely be based on a subset of information that overrepresents incidents in which minority group members engaged in counternormative behaviors. So, again, even in the absence of an objective correlation between group membership and behavioral tendencies, people come to believe that minority groups are more likely to engage in counternormative behavior (Hamilton, Dugan, & Trolier, 1985; Johnson & Mullen, 1994; McConnell, Sherman, & Hamilton, 1994a). These phenomena suggest that some of the roots of stereotype content lie in the broad social norms that govern the behavior of all people within a culture. If we want to predict which particular traits are likely to become central to the stereotype of some group, we might predict that traits corresponding to counternormative behaviors are likely to be stereotypic of minority groups. Why are Stereotypes Shared? Given that the cognitive approach to stereotypes focuses on the ways in which individuals process information, the question of consensus is answered only indirectly. The existence of stereotype consensus is, at best, tacitly attributed to some sort of common informational input (Haslam, 1997). The explanation rests on two assumptions. First, it is assumed that the information on which stereotypes are based is available to most perceivers. Second, it is assumed that the motivational and cognitive

< previous page

page_166

next page >

< previous page

page_167

next page >

Page 167 processes operating on that information are common across most perceivers (at least by those who are outgroup members). If roughly the same information is integrated in roughly the same way by roughly all relevant perceivers, then it is inevitable that some degree of consensus will emerge among those perceivers. Of course, given the power of the social context to influence the cognitive processing of information, that degree of consensus will be far from perfect. The logic underlying this common information input model is compelling. The implications of that logic fit with empirical observations concerning stereotype consensus. Stereotype consensus in the real world is far from perfect. In some descriptive studies, the measured degree of consensus for some stereotypes has been surprisingly low. Indeed, it may be tempting to argue that stereotype consensus is not sufficiently high to warrant additional explanation beyond the appealing account implied by common information input. This approach to the consensus question is not only logically compelling, but it reassures us that the reductionist tendencies of the social cognition approach are scientifically fruitful. If all the subtle cognitive processes that contribute to stereotype development indirectly contribute to the emergence of some consensus, then we have gained something but lost nothing when we put the question of consensus on the intellectual back-burner and shifted our empirical attentions specifically toward questions of process. Even better: If a purely individual-level cognitive analysis is sufficient to explain a cultural-level phenomenon, then this analysis is powerful indeed. These comforting assurances, however, are best resisted. There are some clear explanatory limits to purely individual-level explanations of stereotype consensus. First, empirical research on other elements of shared knowledge has revealed that purely individual-level perceptual and cognitive processes fail to fully account for whatever level of consensus is present within a culture (Boster, 1991). This is almost certainly the case with stereotypes. Stangor and Jost (1997) pointed out that common informational input models “are clearly limited, both because individuals develop stereotypes about groups with which they have had no direct contact, and because contact is notoriously ineffective at changing stereotypes” (P. 34). Second, the appealing parsimony of the cognitive approach is an illusion: Although a common informational input process can account for some level of stereotype consensus all by itself, the fact is that other processes operate on individuals' beliefs as well, and these other processes also contribute to the emergence of consensus. These other processes demand that we attend to the psychologically complex and messy – but very real – consequences of interpersonal interaction. Consensus Goals Lead to Consensus Just as the contents of stereotypes may be direct consequences of individuals' goals that are logically and phenomenologically relevant to those contents, so too consensus may be directly influenced by individuals' psychological needs for consensus. People achieve consensus because they try to do so. There are a lot of reasons that people might seek consensus. People may seek consensus because it implies social approval. Social approval is emotionally rewarding and may serve as an important buffer against existential anxiety (as specified by terror management theory; Greenberg, Solomon, & Pyszczynski, 1997). In addition, people may seek consensus because it provides a form of social validation of their own perceptions and beliefs (Hardin & Conley, chap. 1, this volume; Sherif, 1936). In the absence of objective criteria against which to judge the accuracy of one's subjective perceptions, people attempt to validate these perceptions through comparisons with others' perceptions. Consensus provides validation: If lots of other people believe something, then attributional logic implies that an objective external reality provides a firm foundation for subjective perceptions and beliefs. Finally, and perhaps most important, the desire for consensus helps satisfy the psychological need for a feeling of belongingness (Baumeister & Leary, 1995; Guisinger & Blatt, 1994). Consensus can remove feelings of isolation because we fit in with those whose beliefs we share. Given these multiple underlying psychological needs that are met through perceptions of shared reality, people often experience the goal of achieving consensus. The desire to obtain consensus does indeed lead to the achievement of a perception of consensus (Hardin & Higgins, 1996). And it often leads to actual consensus. Although it is difficult to find research explicitly examining the role of consensus goals in the formation of group stereotypes, there is supportive research examining the application of existing stereotypes. People are more likely to form similar stereotype-based impressions about individuals under conditions in which they are explicitly motivated to seek consensus (Ruscher, Hammer, & Hammer, 1996). Social Influence Goals Lead to Consensus Circumstances within which people are explicitly attempting to create consensus are probably relatively

< previous page

page_167

next page >

< previous page

page_168

next page >

Page 168 infrequent in comparison with circumstances in which other goals are active in working memory. Although these goals may not be experienced phenomenologically as desires for consensus per se, the successful attainment of these goals ordains that consensus must occur. Some of these goals may be experienced as desires to influence others – to convince others to hold the same perceptions, impressions, and beliefs as oneself. Although only a tiny percentage of the voluminous literature on persuasion pertains explicitly to beliefs about social groups, the psychology of persuasion operates on stereotypic beliefs just as it does on any other belief. One implicit end product of successful persuasion is increased consensus on the target belief (for examples relevant to beliefs about groups, see Clark & Maass, 1988; Myers & Bishop, 1970). Other goals may be experienced as desires that make one vulnerable to influence by others. For instance, under conditions in which needs of understanding or needs for belongingness are threatened, individuals are especially likely to be influenced by the beliefs of others (Sherif, 1936; Snyder & Fromkin, 1980). These psychological pressures toward conformity and/or conversion have consequences on Stereotype consensus: The more vulnerable individuals are to conformity, the more likely consensus is to emerge as a consequence. The perception of consensus influences vulnerability to conform – suggesting that consensus breeds greater consensus (Allen & Levine, 1969; Prentice & Miller, 1996). Recent research reveals this effect on shared stereotypes: Perceptions about the degree of consensus in stereotypic beliefs among others influences individuals' own endorsement of those stereotypes (Stangor, Sechrist, Jost, & Bar-Tal, 1997). These general processes of social influence may have specific consequences within ingroup/outgroup contexts. After all, a feeling of belongingness typically requires that an individual identify with and feel accepted by a set of individuals who are perceived to differ in some meaningful way from other individuals (Haslam, 1997; Turner, Hogg, Oakes, Reicher, & Wetherall, 1987). The social influence processes that enhance consensus are most persuasive among ingroup members (Mackie, 1986). This has implications for understanding changes in preexisting shared stereotypes. When people are members of meaningful group categories, their desire for a coherent and positive social identity increases their desire for consensual perceptions of ingroups and outgroups. This motivation, coupled with the opportunity for social influence through communication, influences consensus in perceptions of groups and affects the overall favorability of the impressions of ingroups and outgroups (Haslam, 1997). Communication Goals Lead to Consensus In addition to offering a useful means to persuasive ends, effective and efficient communication represents a desired end. Toward fulfilling that goal, in conversational contexts, people establish and maintain shared reference points and consensual sets of perceptions (e.g., Clark & Wilkes-Gibbs, 1986; Fussell & Krauss, 1992; Garrod & Anderson, 1987; Isaacs & Clark, 1987). Given the important role that social categories play in individuals' impressions of themselves and others, people regularly engage in communication about groups and group members. It follows that these communicating individuals will seek to establish common beliefs about these groups to facilitate efficient communication (Gardner, 1994). Empirical evidence is consistent with this analysis. For instance, Gardner, Kirby, and Findlay (1973) observed that high school students identified stereotyped groups more accurately and efficiently when those stereotype traits were more widely shared – results suggesting that shared stereotypes, like other common perceptions, contribute to efficient understanding of interpersonal communications. Goal-Less Communication Leads to Consensus Sometimes communication operates in the service of particular social goals, but not always. As anyone who travels by airplane knows, sometimes people just talk. But even idle chatter has consequences for those subjected to it. Communication always has some content, and that content – like a flu virus – stands a good chance of being spread when people open their mouths. Stereotype consensus may emerge simply as a side effect of interpersonal interaction. The most persuasive evidence of consensus emerging as a consequence of mere communication is revealed by research on dynamic social impact theory (Latané, 1996). Simply as a result of the many communications between the many individuals who comprise a population, there emerge stable clusters of consensually shared attitudes and beliefs. These clusters emerge in part because people are most highly influenced by those close to them in physical space and because certain individual differences – both demographic (e.g., wealth, status) and psychological (e.g., credibility, extroversion) – lead some persons to be consistently more influential than others. The simple coincidence of communication and individual differences predicts both that some consensus will emerge and that consensus will rarely be complete

< previous page

page_168

next page >

< previous page

page_169

next page >

Page 169 across an entire social landscape. Instead, clusters of consensus emerge that are limited to certain bounded subpopulations. This occurs not only because people within those subpopulations are more likely to communicate within than without their group boundaries (e.g., liberals are more likely to talk to liberals, and conservatives are more likely to talk to conservatives); it occurs because the boundaries of those subpopulations are established by the existence of the shared beliefs (e.g., liberals and conservatives are defined by the existence of shared ideologies). Communication is not merely influenced by cultural constraints, but culture emerges as a result of the interpersonal dynamics of communication. The specific implications of dynamic social impact for stereotype consensus remain largely unexplored (Schaller & Latané 1996). But, as the cognitive representation of stereotypes is governed by the same psychological processes that govern the representation of other elements of social knowledge (Fazio, Jackson, Dunton, & Williams, 1995), this much seems inevitable: The act of communication compels the emergence of clusters of shared stereotypes. Summing Up A joint consideration of the cognitive processes influencing stereotype content and the interpersonal processes influencing stereotype consensus offers some preliminary guidelines that help us address the question of why stereotypes with certain types of contents become socially shared. All else being equal, traits indicating negative behavioral tendencies are most likely to become central to culturally shared stereotypes of outgroups, and those negative traits are most likely to pertain to counternormative behaviors. These beliefs are especially likely to be consensual under conditions in which communication about those groups is most likely to occur. These are useful guidelines for prediction, and they do – on the surface – seem to fit the nature of stereotypes about real groups. WHAT WE DON'T KNOW These general rules, however, offer rough and highly fallible means of prediction. Empirical research reveals that stereotypes are not made up of simply negative and counternormative behaviors. Many traits that are perceived to be stereotypic of minority outgroups are, in fact, valued very highly. Among Americans, for instance, certain outgroups (e.g., Germans, Japanese) have historically been viewed in many positive ways (Karlins et al., 1969; Katz & Braly, 1933). Even chronically maligned ethnic minority groups are the subject of stereotypes with positive elements: Jews tend to be perceived as intelligent and industrious (Karlins et al., 1969), and African Americans are perceived to be athletic and rhythmic (Devine & Elliot, 1995). Furthermore, although the range of negative and counternormative traits is huge, only a small number of these traits actually ever becomes central to the shared stereotype of any particular minority outgroup. Among Americans, honesty is valued at least as highly as intelligence and industry (Rothbart & Park, 1986), yet traits connoting dishonesty are relatively rare in cultural stereotypes of American ethnic minority outgroups compared with the frequency of traits connoting laziness or ignorance (e.g., Karlins et al., 1969; Katz & Braly, 1933). Additionally, traits that offer the most diagnostic descriptions of group members or of those group members' social circumstances are not necessarily the most likely to become central to consensual stereotypes – even under conditions in which communication about those groups is most likely to occur. The traits central to many group stereotypes are clearly not diagnostic in differentiating among groups. In one descriptive study of stereotype content among American perceivers, the trait (industrious) that was most commonly believed to be descriptive of the ingroup (Americans) was also commonly believed to be descriptive of most other outgroups as well (Katz & Braly, 1933). Clearly, this trait was not diagnostic of group membership. Furthermore, lots of traits offer equally accurate descriptions of groups or their circumstances, but not all become central to shared stereotypes. In relation to their overall demographic frequency in the United States, African Americans are overrepresented in popular professional sports; not surprisingly, athletic is central to White Americans' stereotype of African Americans (Devine & Elliot, 1995). African Americans are also overrepresented among the ranks of fatherless families, and white Americans are well aware of this overrepresentation (McCauley, 1995). Yet traits pertaining to this descriptively diagnostic characteristic have not once appeared in the various descriptive studies on the contents of stereotypes about negroes, blacks or African Americans (Devine & Elliot, 1995). In fact, one study found that the term loyal to family ties to be the single most commonly endorsed descriptor of the group (Dovidio & Gaertner, 1986) – a finding that violates every single one of our general heuristics for predicting stereotypes that matter. So the question remains: Why do some traits become central to the contents of shared stereotypes while other traits do not? Our reflection on the past does help

< previous page

page_169

next page >

< previous page

page_170

next page >

Page 170 clarify a direction for the future. Questions of content do demand a consideration of individual goals and other cognitive phenomena. Questions of consensus do demand a consideration of interpersonal influence and other communication phenomena. Questions concerning the content of shared stereotypes demand; therefore, the explicit integration of individual- and interpersonal-level processes. Acting in dynamic synergy, processes of cognition and communication are likely to more effectively help us understand the origins of cultural-level structures such as stereotypes. Recently, we have been exploring the implications of one particular metatheoretical perspective along these lines. It is a social-evolutionary framework, and it yields a number of interesting hypotheses concerning the origins of stereotypes that really matter. A SOCIAL-EVOLUTIONARY FRAME-WORK The contents of socially shared cognitive structures are influenced by processes roughly analogous to those that influence the emergence of organic structures within biological populations (Boyd & Richerson, 1985; Campbell, 1965; Cavalli-Sforza & Feldman, 1981; Hull, 1988; Sperber, 1990). As a result of biological evolutionary processes, the likelihood that some specific genetic information is to be transmitted from one individual to another predicts the likelihood that anatomical characteristics associated with that information become widely descriptive of a population. Similarly, as a result of social-evolutionary processes, the likelihood that some specific knowledge is to be communicated from one individual to another predicts the likelihood that this specific knowledge is to become common across a population. Applying this idea to group stereotypes, it follows that the specific characteristics of shared stereotypes are likely to be influenced by selection pressures through which certain variants of stereotypes are selected for or against within the processes of interpersonal communication. A Trait's-Eye View of Stereotypes To appreciate the implications of this social-evolutionary framework, it is useful to adopt the perspective of a potentially stereotypic characteristic or trait. To become stereotypic in a way that matters, a trait must establish itself as part of the network of associative memory links that define stereotypes at the individual level, and to do so within the cognitive structures of many individuals across a population. In theory, all traits have equal potential to do this. In fact, however, that potential is usually achieved by only a handful of traits at any given time. To gain the competitive edge necessary to attain such a place in cognitive space, a trait must be likely – relatively more likely than other traits – to be communicated from person to person. From the perspective of a potentially stereotypic trait that aspires to make a home in as many individual cognitive structures as possible, a high likelihood of being communicated is a highly desirable characteristic. So here we arrive at a fundamental hypothesis underlying this social-evolutionary approach to stereotypes: The more likely that some particular stereotypic belief is to be communicated to other individuals, the more likely that it (rather than some less likely to be communicated belief) is to become shared across a population of people. This basic hypothesis is a deceptively simple statement. Of course beliefs are more likely to be shared if people talk about them. If you tell me that your Alabama-born coworker Billy gobbled down a kilo of pulled pork with peppered vinegar sauce, then sure, there is now one more person in the world who believes that Billy likes his pig. But that is not what this hypothesis says. Rather, it suggests that, if you think Billy is both a barbecue fan and a bigot, but you are more likely to tell me that he made a racist remark about your Nigerian neighbors than you are to tell me that his eyes go moist and misty in the presence of grilled meats, then I am more likely to get the impression that Billy is a bigot and not so likely to get the impression that he is a barbecue aficionado. Moreover, this may not be merely your idiosyncrasy. Perhaps there is some general population-wide tendency to communicate more readily about others' racist statements than about their gastronomic weaknesses (we discuss later some answers to the important question of why certain beliefs are more likely to be communicated than others). If so, each individual within the population will both hear and tell stories about prejudicial tendencies performed by all sorts of people, but these individuals will not be trading tales about dietary habits so much. Although individual-level appraisals of reality might suggest that both traits are equally stereotypic, the two traits do not emerge as equally stereotypic at all. The inevitable implicit influence that occurs through communication is such that the emerging shared stereotype of Southerners coalesces around one of these traits, but not the other. The implication is this: To the extent that our decisions about what we talk to each other about are not random or idiosyncratic, then these decisions influence the contents of emerging cultural stereotypes. The

< previous page

page_170

next page >

< previous page

page_171

next page >

Page 171 decisions do not have to be based on anything that is either logically or phenomenologically relevant to stereotypes or prejudice. All they have to do is influence the contents of our communication and that, in turn, will influence the emerging contents of our culturally shared characterizations of groups. The Influence of Communication Likelihood on Stereotype Formation and Change Several pieces of evidence support the basic hypothesis that those traits that are most likely to be communicated to others are more likely to be perceived consensually as stereotypic. One intriguing piece of evidence relates to changes over time in the contents of cultural stereotypes. The social-evolutionary hypothesis predicts that the extent to which any given trait persists as part of a cultural stereotype is a function of the likelihood that people will communicate with others about the particular trait. Traits that are less likely to be communicated – for whatever reason – are less likely to remain strongly stereotypic. Traits that are more likely to be communicated are more likely to remain stereotypic over time. Schaller, Conway, and Tanchuk (2000) empirically tested this prediction on data pertaining to changes in stereotypes about African Americans held by European Americans in the United States. Katz and Braly (1933) reported 12 traits that Princeton students in the early 1930s most commonly indicated as stereotypic of Negroes. Each trait was identified with the exact percentage of subjects who indicated that trait to be especially central to their individual stereotype of the group. This percentage can be taken as a rough measure of the centrality of the trait to the culturally shared stereotype at that point in time. Two subsequent surveys of Princeton students (Gilbert, 1951; Karlins et al., 1969) and two more recent surveys conducted on other University campuses (Devine & Elliot, 1995; Dovidio & Gaertner, 1986) used similar methods and so reported the extent to which various traits defined the content of the culturally shared stereotype at later points in time. These descriptive results provide opportunities to assess the extent to which each trait remained stereotypic of African Americans over time. To assess persistence, the percentage of individuals endorsing a trait as stereotypic at a particular point in time was subtracted from the percentage endorsing it at subsequent points in time. Higher values indicate greater perseverance over time in the cultural stereotype of African Americans. The question of interest is whether variability in the extent to which traits remained stereotypic over time was predicted by a measure indicating the communication likelihood of each trait. The index of communication likelihood was based on an independent sample of 33 university undergraduates who rated a list of 76 trait adjectives (including each of the adjectives reported as most stereotypic of African Americans in the surveys identified earlier) in response to two questions: “Suppose you met a person who had this trait, and now you're talking to others about this person; how likely is it that you'll tell these others that this person has this trait?” and “In general, considering all the various traits of people that you talk about when you talk to others, how common is it to find yourself talking about this trait (or behaviors relevant to this trait)?” Responses to each question were averaged to yield two scores for each trait. The resulting scores were highly correlated and so were averaged to yield a single index for each trait indicating its communicability. Treating trait as the unit of analysis, correlations were computed between the communicability index and the extent to which each trait remained stereotypic across different points in time. Ten separate correlations were able to be computed, indicating persistence across one, two, three, and four generations. Nine of these 10 correlations were positive – indicating that the communicability of a trait predicted its persistence in a stereotype over time. These correlations remained positive (and substantial in magnitude) even when statistically controlling for differential tendencies toward statistical regression and for the general evaluative favorability of each trait. In addition to predicting stereotype change, a trait's likelihood of communication also predicts the trait's tendency to be part of a stereotype in the first place. In a controlled experiment (Schaller & Conway, 1999), participants in groups of two were presented with information describing individual members of two groups about which they had no previous knowledge (the Red Group and the Blue Group). The information was structured in such a way that the groups were objectively different on two personality characteristics; in general, people in the Red Group were both more aggressive and more intelligent. During the course of examining the group information, participants also communicated periodically with each other concerning their emerging impressions. We included a manipulation to directly influence the likelihood that information relevant to intelligence or information relevant to aggressiveness would be communicated within dyads. In one condition, both dyad members were told to try to communicate primarily about positive information; in another condition, both dyad members were told to try to communicate primarily about negative information. At the completion of the session, each participant completed measures

< previous page

page_171

next page >

< previous page

page_172

next page >

Page 172 assessing stereotypes of the groups. Consistent with the social-evolutionary hypothesis, results reveal that, although all individuals had identical information indicating that the two traits are equally stereotypic, the contents of the emerging stereotypes differed in the two conditions. Under conditions in which people were more likely to communicate about positive characteristics, intelligence emerged as a more central component of stereotypes. Under conditions in which people were more likely to communicate about negative characteristics, aggressiveness emerged as more central to stereotypes. Additional analyses (not reported by Schaller & Conway, 1999) revealed that these particular stereotypes with particular contents were not merely individual-level knowledge structures; the stereotypic beliefs formed by each individual were correlated with the beliefs formed by their dyad partner. In fact, this within-dyad similarity remained positive even after controlling for the predicted effects of the manipulation. This indicates that the emerging consensus could not be accounted for merely by similar inferential processes operating within individuals, but was the product partially of unique communicative interactions between individuals. It is also noteworthy that these effects emerged in the absence of any explicit goal to form stereotypes of some specific sort or to create consensus. Apparently the things people communicate about, even if not directly intended to influence the beliefs of self or others, do have an impact on the contents of emerging shared stereotypes. The Role of Cognitive Variables in Directing Social-Evolutionary Mechanisms So, traits that are more likely to be communicated between individuals are more likely to become and remain central to shared stereotypes. This basic social-evolutionary hypothesis begs the more interesting question: What factors influence the likelihood that a specific trait or belief will be communicated to others? Or, more to the point of prediction, what specific features of traits, persons, and situations might we attend to to predict the content of stereotypes that matter? Advantageous Properties of Stereotypic Beliefs Dawkins (1989) highlighted two general key features of beliefs that can influence the extent to which they are likely to become widespread: copying fidelity and fecundity. Copying fidelity refers to the extent to which the particular belief resists alteration and change in the process of communication. Fecundity refers – very generally – to the extent to which the particular belief will be acceptable to others. What determines a potentially stereotypic trait's copying fidelity? Given the limits of languages and the limits of the people who use them, it is nearly impossible for any communicator to communicate a stereotypic belief that is fully representative of the form that belief takes in her head. Moreover, as any student of rumor can attest, beliefs often are received in a different form than the one in which they were transmitted. Stereotypic beliefs that, in general, are least likely to suffer changes in meaning are more likely to become and remain part of stereotypes that matter. Among the various characteristics of potentially stereotypic traits that might influence their copying fidelity, one strikes us as particularly likely to exert an influence: Complexity. The more complex some characteristic is, the less likely it is to be communicated with fidelity. Complex beliefs are more likely to be sharpened by communicators to ease the act of communication. Complex beliefs are more likely to be sharpened further by cognitively fatigued or strongly schematic receivers. Given these pressures, it is likely that the characteristics of groups that become stereotypic are those that are simple to express and easy to understand. Of course, complex and simple are vague and relative terms. Beliefs may more complex because they include qualifying or limiting statements (“Southerners sure like barbecued meats, but the preferred style of meat sure varies from region to region”) or for a variety of other reasons as well. What makes a trait fecund? It seems likely that the fecundity of a trait is determined largely by the implicit rules of conversation that govern the exchange of information between any two individuals (Cushman & Whiting, 1972; Grice, 1975). Of particular importance is the social maxim concerning relevance: Savvy conversationalists observe the social rule that the information they transmit to others should be interesting, useful, and relevant to others' needs and goals. To some extent, individuals surely assume that if they find some group's traits and behaviors to be interesting and personally relevant, then so will others. Indeed the psychological literature on causal attribution reveals considerable overlap between the sorts of social information that inspires individual interest (as indicated by inferential activity) and the information that is judged to be interpersonally relevant to communication (e.g., Edwards & Potter, 1993; Hilton, 1990). This link between causal attribution and interpersonal communication suggests that the traits most likely to enter conversation – and so to become stereotypic – are those that correspond most unambiguously to behaviors that have important personal

< previous page

page_172

next page >

< previous page

page_173

next page >

Page 173 consequences. There is little point to transmitting social information that is nondiagnostic of actual behavior, or that is diagnostic only of trivial and inconsequential behaviors. Research on behavior-trait correspondence (e.g., Jones & Davis, 1965; Rothbart & Park, 1986) implies that traits that are especially counternormative or negative are perceived to be especially predictive of correspondently unpleasant behavior. Other research (e.g., Fiske, 1980; Pratto & John, 1991) indicates that people are especially attentive to the implications of negative characteristics of others – probably for the purpose of avoiding unpleasant social interactions. It follows that negative (or other counternormative) characteristics of groups are especially likely to be judged as interesting and relevant, and so more likely to be communicated to others. However, not all counternormative behaviors and negatively valued traits are equivalent in their practical implications. Recent functional perspectives on stigma and prejudice suggest that, for reasons based ultimately on biological evolutionary pressures, individuals may be especially sensitive to traits that most clearly imply the threat of physical injury or illness (Neuberg, Smith, & Asher, in press; Schaller, 1999; Stangor & Crandall, in press). For instance, although both stupidity and hostility are similarly negative, a group's cruelty more clearly indicates a potential danger. Consequently, people may generally be more attentive to hostility than stupidity, and may be more likely to communicate to others about the hostility rather than the stupidity of a group. Thus, the individual-level processes that conspire to tilt the content of stereotypes in a particular direction are likely to be amplified through the communication process. These brief and preliminary speculations suggest promising and fruitful directions for future research. If we can identify those features of traits that influence their fidelity and fecundity, we may discover that the explanations underlying the contents of cultural stereotypes lie not only within ourselves but also within the properties of what we talk about. Dawkins (1989, p. 200) once suggested that, ‘‘a cultural trait may have evolved in the way that it has, simply because it is advantageous to itself.” Standing alone, this is merely a provocative statement; when considered in the present context, it offers a potentially fruitful twist on the usual functional perspective we bring to the study of stereotypes. Interactions Between the Properties of Beliefs and the Properties of Individuals If traits differ along dimensions of fidelity and fecundity, it is because people in general share similar cognitive limitations and psychological goals. Although the presence of cognitive limits and certain psychological goals may be common across individuals, the magnitudes of these limits and goals differ between individuals. To more fully predict those potentially stereotypic traits that really do emerge as central to stereotypes that matter, it may be valuable to consider the variable features of beliefs in conjunction with the variable features of the people who traffic in those beliefs. Any personality variable that influences the successful communication of information from one person to another is potentially relevant to the process of stereotype formation and stereotype change. Some of these variables surely pertain to cognitive abilities and rhetorical styles. One such variable is the need for cognition – the desire for and enjoyment of effortful cognitive activity. Research on collective decision making reveals that individuals with a higher need for cognition generate more valid arguments to support their perspectives, and so are more persuasive in determining the eventual outcome of collective decisions (Shestowsky, Wegener, & Fabrigar, 1998). Similar processes might operate when individuals discuss different beliefs about the characteristics of groups; those individuals who support their perspectives with more compelling arguments may ultimately have a greater influence on the eventual contents of stereotypes shared across a populations. Of course, such an effect may not operate equally across all traits. Traits differ in the extent to which they are easily confirmed or disconfirmed by supportive evidence (Rothbart & Park, 1986). To the extent that rational argumentation about a group's traits depends on reference to evidence, it is most likely to be persuasive in regard to those traits that lend themselves to clear confirmation or disconfirmation. Another individual difference variable that may merit attention is the need for structure – a personality variable that relates to dogmatism and close-mindedness (Thompson, et al., chap. 2, this volume). Persons with higher chronic needs for structure are more inclined to infer stable traits from others' behaviors (Moskowitz, 1993; Webster, 1993), more likely to form erroneous stereotypes of groups (Schaller, Boyd, Yohannes, & O'Brien, 1995), less inclined to be persuaded by new information once they have formed a preliminary belief (Kruglanski, Webster, & Klem, 1993), and, when conversing, more inclined to use linguistic constructions that implicate stable traits in others (Rubini & Kruglanski, 1997; Webster, Kruglanski, & Pattison, 1997). Considering these tendencies together, it may be that stereotypic beliefs formed by persons higher in need for structure would be especially likely to become shared

< previous page

page_173

next page >

< previous page

page_174

next page >

Page 174 across a population of communicating individuals. If so, the implication is disturbing: Those individuals especially likely to form erroneous stereotypes on their own are also especially likely to infect others with their stereotypic beliefs. Meanwhile, those individuals who are less likely to form stereotypic beliefs on their own are nonetheless likely to contract those erroneous stereotypes as a result of their contact with others. Some empirical evidence (Conway & Schaller, 1998) supports the general line of thinking that people high in need for structure are sometimes especially influential bearers of stereotypes, although the evidence is preliminary and must be interpreted with considerable caution. In general, the consideration of individual differences in cognitive styles offers an interesting perspective on the role of individuals in the emergence of public opinion. Obviously some individuals are more influential than others in determining the shape of norms and beliefs that coalesce over time in any population. However, the factors that determine who is influential are not always obvious. Certainly opinion leaders are likely to be those who are most blessed with features that facilitate intentional persuasion, such as physical size and wealth (Latané, 1996). Some of the characteristics of opinion leaders may be considerably more subtle – pertaining to cognitive style rather than material strength – and may operate even in the absence of any deliberate attempt at influence. In addition to variables pertaining to cognitive style and epistemic goals, it is also worth considering the implications of individual differences in the need for belongingness. If the need to belong plays a role in our desire to achieve consensus, then those people who are particularly high in belongingness needs may be more likely to engage in behaviors that create consensus – such as accepting others' stereotypic beliefs about groups and attempting to pass those normative beliefs on to others as well. If so, then people who are high in the need to belong may be especially effective receivers and transmitters of stereotypic traits. Moreover, it may be functionally most advantageous to communicate content that is especially relevant to the defining features of the social group to which individuals wish to belong. If individuals within a particular population define themselves and their group according to their antiracist beliefs, the stereotypic belief that “they [outgroup members] are racist bigots” strongly supports the social belongingness needs of the ingroup members. Consequently, more so than many other potentially stereotypic traits, traits relevant to this group definition are likely to be especially virulent within this population (especially among individuals with stronger needs to belong). Within a different population of individuals in whom a sense of belongingness was predicated on a different social identity, the same belief would be less likely to become central to the stereotypes that matter. Interactions Between the Properties of Beliefs and Social Contexts Motives, goals, and other cognitive processes that influence the contents of communication not only differ across individuals, but also across social contexts. Thus, depending on the features of local cultural norms and expectations, one might expect different traits to actually emerge as stereotypic. Within certain normative contexts, it may be more likely that some beliefs about groups are more or less acceptable to discuss openly. Indeed, the potentially powerful normative influence on the acceptability of group-relevant communications underlies virtually all formal and informal interventions designed to undermine existing intergroup perceptions and behavior (e.g., political correctness). But norms do not have to be either logically or phenomenologically relevant to groups or stereotypes to exert an influence on the contents of stereotypes about groups. Given the operation of social-evolutionary processes, any contextual and normative variable that influences communication may have a indirect (and entirely unintentional) impact on the emerging contents of group stereotypes. Consider, for instance, the impact of impression management motives on the formation of brand-new group stereotypes. One's desire to make a good impression on others appears to bear little relevance to anybody's tendencies to form group stereotypes. But appearances are deceiving. A social-evolutionary perspective illuminates the potential impact of impression management motives on stereotype formation. Interpersonal communication is a medium through which individuals craft desired public images. Depending on social context or local norms, people may strategically choose to talk about certain characteristics of groups rather than others to cast a desirable impression. Those characteristics of groups that are most likely to be communicated are most likely to emerge as central to emerging culturally shared stereotypes of those groups. Two recent experiments yielded evidence of this indirect impact of impression management motives on emerging stereotypes (Schaller & Conway, 1999). Participants in groups of two were led to believe that desirable personal outcomes in life were associated with either tendencies to talk about positive characteristics of others or tendencies to talk about negative characteristics of others. They then received, over a period of time,

< previous page

page_174

next page >

< previous page

page_175

next page >

Page 175 information describing individual members of two groups (e.g., the Red Group and the Blue Group) who actually did differ on two personality characteristics – one of which was positive (e.g., aggressiveness) and the other negative (e.g., intelligence). Participants also communicated with each other periodically concerning their emerging impressions. It was expected that, in their desires to cast desirable personal impressions on their communication partner (who held the same belief concerning the success implications of talking about positive [or negative] traits), both dyad members would strategically alter their communications to talk more about positive (or negative) traits. Given the implicit influence of communication content, this strategic self-presentation was expected to exert an indirect influence on the emerging contents of their stereotypes. Results from both studies show these predicted effects. The contents of interpersonal communication were indeed influenced by the subtle impression management norm. For instance, compared with other dyads, individuals within dyads operating under a “talk about negative things to make a good impression” norm were more likely to communicate about the negative characteristics of group members. As a result, there was also an effect on the emerging contents of group stereotypes. Although all participants read and communicated about the same objective information, participants operating under an implicit ‘‘talk about negative things to make a good impression” norm were more likely to form stereotypes based on differences in negative traits (e.g., aggressiveness) and less likely to form stereotypes based on differences in positive traits (e.g., intelligence). The contents of these emergent social structures cannot be attributed solely to individual-level cognitive processes. Comparisons with several control conditions (e.g., a condition in which participants wrote notes with the intention of exchanging them, but never actually made the exchange) revealed that the manipulation required successful consummation of communication intentions to exert its full effects on stereotype content. Thus, although there certainly may be some effects of purely individual-level processes associated with the transmission of communication (e.g., Zajonc, 1960), the effects observed in these studies demand a explanation that attends to both individual-level cognition and interpersonallevel communication. These effects of impression management motives on stereotype formation offer a glimpse into a potentially fruitful realm of scientific investigation exploring the indirect and unintended effects of individual cognitive process variables on the contents of emerging cultural stereotypes. Any individual-level cognitive structure may – if it has any effects on the contents of interpersonal communication – shape the formation of stereotypes that matter. Operating within social-evolutionary processes, individual-level cognitive variables that superficially seem irrelevant actually play important roles in the origins of stereotypes that matter. FROM COGNITION TO CULTURE Stereotypes are but one example of the consensually shared belief systems that define any culture. Processes that influence the contents of cultural stereotypes influence the contents of cultural norms of all kinds. Just as we acknowledge that some of these norms really matter, it is easy to dismiss many of these norms as merely superficial fad and fashion. The norms that take the form of cultural self-concepts or group stereotypes have substantial consequences; those that take the form of regional accents or favored styles of barbecue may seem trivial and inconsequential by comparison. This sort of judgment is dangerous; it fails to consider the functions that every one of these norms serves for the individuals and populations described by the norms. The extent to which an outside eye judges a normative belief system to be important or trivial is probably uncorrelated with the actual significance of that belief system to the people defined by those beliefs. Within the American South, stereotypes matter and so too does barbecue. (John Egerton, an expert on things Southern, once wrote that “the pit wars rage from eastern North Carolina to the western tip of Texas, and if you don't think it's serious business, try asking for tomato-based sauce out east of Raleigh, or telling a Texan how you really prefer pork. I'd sooner stroll unarmed through downtown Beirut” [Egerton, 1990, p. 68].) Cultural norms of all kinds matter, and the origins of them all might be fruitfully examined through a social-evolutionary lens. Science too is defined by its norms, and scientific norms emerge and evolve in the manner of any other cultural structure. Once upon a time and not so long ago, social-psychological inquiry into group stereotypes was defined by a method that merely described the contents of shared stereotypes. When the cognitive pandemic replicated across the psychological landscape, the emphasis shifted. Description of contents gave way to description of processes. The shift in emphasis has been spectacularly successful; we can now no more rid ourselves of a cognitive perspective than the American South could rid itself of barbecue. And that's good. Of course, while the recipe for a tasty meat snack might well remain successfully unchanged for generations, the recipe for successful science calls for new ingredients all

< previous page

page_175

next page >

< previous page

page_176

next page >

Page 176 the time. It is in this context that a social-evolutionary approach to socially shared beliefs seems especially useful. This framework invites us to consider the many still undescribed effects that motives, goals, and other cognitive products have on the contents of communication. It invites us to consider too the many still unknown influences that communication has on the emergence of cultural beliefs. These invitations are well worth accepting. A vast amount of potential knowledge about the cognitive origins of culture awaits discovery.

< previous page

page_176

next page >

< previous page

page_177

next page >

Page 177 11 The Dynamic Relationship Between Stereotype Efficiency and Mental Representation Jeffrey W. Sherman Northwestern University Stereotypes are “cognitive structures that contain the perceiver's knowledge, beliefs, and expectations about a human group” (Hamilton & Trolier, 1986, p. 133). In the past 20 years, a great deal of attention has been devoted to understanding how stereotypes influence the perception of individual members of stereotyped groups. This research has shown stereotypes to have a profound impact on many basic cognitive processes that underlie social perception. They determine the kinds of information to which people attend, the ways in which that information is interpreted and encoded, and the manner in which the information is stored in memory and subsequently retrieved. Of course, stereotypes also influence the content of resulting judgments about target individuals as well as perceivers' behavior toward those individuals (for a review, see Hamilton & Sherman, 1994). More recently, attention has also turned to defining more precisely the nature of these stereotypic cognitive structures or mental representations (for reviews, see Hamilton & Sherman, 1994; Sherman, 1996). Broadly speaking, a mental representation is “an encoding of some information, which an individual can construct, retain in memory, access, and use in various ways” (Smith, 1998, p. 391). Research on the mental representation of stereotypes has been concerned with identifying the manner in which knowledge about social groups is constructed and retained in memory. When we say that a stereotype has been activated, what, specifically, do we mean? What exactly is being represented and activated in memory? To this point, research on the influence of stereotypes on social perception processes and on the mental representation of stereotypes has proceeded in a largely independent fashion, with little overlap having been noted or pursued between the two topics. However, these topics are much more closely related than has been previously acknowledged. We cannot directly observe people's mental representations. We can only infer these representations through the processes that act on them. As such, any model of representation is also a model of cognitive processing and is constrained by what we know about cognitive processes. In turn, what we learn about mental representation constrains the viability of different process models. Thus, one cannot be fully understood without a corresponding understanding of the other (Anderson, 1978; Barsalou, 1990; Smith, 1998). The primary goal of this chapter is to detail how research on stereotype representation and stereotypic effects on information processing (particularly as they relate to stereotype efficiency) inform one another. First, I describe recent research on stereotype representation and detail how concerns for efficient information processing play a vital role in the nature and development of stereotypic representations. Next, I explain how these representations in turn influence the processing of stereotype-relevant information. This discussion focuses on a recently developed model of how stereotypes confer efficiency in social perception. I argue that the efficient processing of stereotype-relevant information is largely dependent on the representational nature of stereotypes. Finally, I describe how these efficient processing mechanisms feed back into and reinforce the representational nature of stereotypes. Thus, the relationship between mental representation and processing efficiency is a dynamic one. Concerns about efficiency influence representation, which in turn influences processing efficiency, which in turn influences representation. THE MENTAL REPRESENTATION OF STEREOTYPES At one level, the mental representation of a stereotype can be defined in terms of the particular content that is included in the representation (e.g., the particular traits thought to characterize the target group). However, such a definition is merely descriptive and uninformative as to

< previous page

page_177

next page >

< previous page

page_178

next page >

Page 178 the mechanisms by which stereotypes produce their effects. At another level, stereotype representation can be defined by the specificity with which the particular stereotypic traits are represented in memory. Research on this question has focused on the distinction between abstraction-based and exemplar-based knowledge (see Hamilton & Sherman, 1994; Sherman, 1996). Abstraction-based knowledge summarizes the features of a concept that have been extracted from experiences (both first- and second-hand) with specific instances of the concept. Generally speaking, these abstractions can be thought of as category averages along different attributes (including, perhaps, summary estimates of category variability; Hamilton & Sherman, 1994; Park & Hastie, 1987). In the case of knowledge about a social group, an abstract stereotype would consist of a group summary along a particular dimension (e.g., friendliness) that had been extracted from relevant experiences with the group and its members and had become associated with the group. In contrast, exemplar-based knowledge consists of the specific instances of the concept in question. Exemplar-based stereotypes would consist of representations of particular group members and their behaviors (e.g., specific friendly and unfriendly behaviors). A central argument of this chapter is that specifying the different contributions of these two types of knowledge to stereotype representation has significant implications for understanding the functional significance of stereotypes in information processing, and that defining stereotypes at this level of specificity can clarify the roles that stereotypes play in attentional, encoding, retrieval, and judgment processes.1 Historically, abstractionist views have held sway, with stereotypes being conceived of as (over)simplified generalizations about groups of people (e.g., Brigham, 1971; Lippmann, 1922). In more recent years, these generalizations have been described alternatively as schemata (e.g., Taylor & Crocker, 1981), prototypes (e.g., Brewer, 1988), expectancies (e.g., Hamilton, Sherman, & Ruvolo, 1990), and Bayesian base rates (e.g., McCauley & Stitt, 1978). Alhough there are subtle differences among these models, they share a most important feature in common: They all suggest that people develop and store abstract impressions that summarize the behavioral tendencies of social groups. These abstractions may initially be extracted from encounters with specific group members. However, once formed, they may be retrieved and used independently of information about particular individuals and their behaviors. Although these conceptualizations are clearly abstractionist in nature, they are not highly specified models of mental representation. Specifying and testing underlying representational assumptions have never been primary concerns to researchers in these traditions. Indeed, few attempts have been made to verify the proposed abstract nature of these stereotypes. Moreover, the potential role of specific exemplars in group knowledge is usually not considered in these models. This is not to say that these models posit no role for exemplars. Rather, the question is simply never raised. Although these models have frequently been characterized as “pure-abstraction” models that are hostile to exemplar-based processes (e.g., Hamilton & Sherman, 1994; Sherman, 1996; Smith, 1990), they are so more out of neglect than out of clear intent. Serious consideration of the role of exemplar-based knowledge in stereotype representation has appeared only recently. Researchers have proposed pure exemplar-based models of stereotyping that directly challenge traditional abstractionist conceptions (e.g., Linville, Fischer, & Salovey, 1989; Smith, 1990; Smith & Zarate, 1992). According to these models, stereotypes do not exist as independently stored knowledge structures. Rather, group knowledge consists solely of information about particular group member behaviors. If knowledge of the group as a whole is required, it must be created on the spot by retrieving and summarizing information about specific individuals. A Developmental Mixed-Model of Group Representation The preceding discussion of abstraction and exemplar-based models of stereotypes suggests that group knowledge is either entirely abstraction- or exemplarbased. However, it has become clear that neither pure abstraction nor pure exemplar models accurately describe group knowledge. It has been relatively straightforward to demonstrate that particular category exemplars may sometimes form the basis of people's perceptions of social groups (e.g., Sherman, 1996; Smith & Zarate, 1990) or at least influence those perceptions (e.g., Bodenhausen, Schwarz, Bless, & Wänke, 1995; Lewicki, 1985; Schwarz & Bless, 1992).2 Likewise, it is clear that group knowledge is not always based on information about particular group members and is sometimes abstracted (Park & Hastie, 1987; Sherman, 1996; Sherman, Klein, Laskey, & Wyer, 1998). As a result, many theorists now subscribe to mixed models of mental representation. According to these models, knowledge of a group may be based on either abstract or exemplar information depending on the circumstances. For example, the extent to which group knowledge is based on

< previous page

page_178

next page >

< previous page

page_179

next page >

Page 179 abstractions and exemplars has been posited to be mediated by group membership (ingroups vs. outgroups; Park & Judd, 1990; Park, Judd, & Ryan, 1991), group cohesion (Hamilton & Sherman, 1996), target stereotypicality (Sherman, Klein, et al., 1998), and learning order of abstract and exemplar information (Smith & Zarate, 1990), among other factors. In my own research, I have proposed and tested a model of group representation that specifies group familiarity as a key determinant of the extent to which group knowledge is based on abstractions or exemplars (Sherman, 1996). According to this model, in the early stages of learning about a social group, judgments of the group are based on information about particular group members because too few exemplars have been encountered to support the formation of abstract knowledge. With sufficient experience with group members (or secondhand accounts of their attributes), perceivers form abstract representations of the attributes that are stereotypical of the group. Once formed, these abstractions may be retrieved independently from group exemplars to make judgments about relevant features of the group. Thus, group knowledge is exemplar based primarily when no abstract knowledge exists (although other factors may encourage exemplar based judgments once abstractions have developed; see also Klein & Loftus, 1993; Klein, Loftus, Trafton, & Fuhrman, 1992; Sherman & Klein, 1994, for familiarity-based models of the mental representation of self and other individuals). The predictions of this model were supported in two experiments. In the first experiment, participants formed impressions of an ill-defined group of college students who all belonged to the same club on campus. No other information was given about the group that would allow participants to rely on preexisting group stereotypes in forming their impressions. After learning either a relatively small or large amount of information about the group, participants performed a judgment priming task that assessed the extent to which judgments about the group depended on the retrieval of information about particular group members. Briefly, this task required participants to either make a judgment about the group or perform a control task, and then retrieve a particular behavior performed by a member of the group. If participants spontaneously activate behaviors about specific group members when they make judgments about the group, then it should take relatively little time to retrieve one of these behaviors following the judgment task as compared with the control task. In this case, the behaviors would have already been activated and would be relatively accessible. The results demonstrate that group judgments were based on activated exemplars only when participants had received little information about the group. At high levels of group experience, judgments no longer depended on exemplar retrieval. In this case, it appeared that participants were able to rely on abstract, trait-based stereotypes that had been formed about the group. The second experiment examined the mental representations of more familiar social groups. Participants learned either a relatively small or large amount of information about either a group of engineers or a group of priests. For these groups, participants may rely on preexisting stereotypes to form judgments of the groups along relevant dimensions, regardless of how little information they have received about particular group members. If these stereotypes exist as prestored abstractions, then judgments along relevant dimensions would not involve the retrieval of information about particular group members. However, if stereotypes are merely exemplar-based constructions, then relying on them would involve the retrieval of the novel, experimentally presented group members (e.g., Smith, 1990). That is, the most recently encountered exemplars would be activated in making the group judgment. The results demonstrate that judgments along stereotypical dimensions (engineer-intelligent, priest-kind) did not involve the retrieval of group exemplars, regardless of how little information had been presented about the groups. Even when participants had learned a small amount of information about the groups, judgments along these dimensions did not depend on exemplar retrieval, presumably because participants were instead relying on their preexisting abstract stereotypes. By contrast, judgments along nonstereotypical dimensions (engineer-kind, priest-intelligent) did rely on exemplar retrieval. This research suggests that stereotypes are often stored as abstract knowledge structures in memory, and that their availability reduces the role of exemplar-based processes (see also Sherman, Klein, et al., 1998). The Need for Efficiency Influences Stereotype Representation The developmental mixed model of stereotype representation is based on a functional analysis of memory. According to this analysis, the ability to summarize experiences across time, situations, and individuals is critical to the efficient operation of memory. There are two aspects to this efficiency. First, by capturing patterns of invariances in the environment, people are able to learn generalizable skills, predict the future, and explain novel events (McClelland, McNaughton, & O'Reilly, 1995; Nosofsky, Palmeri, & McKinley, 1994; Schank,

< previous page

page_179

next page >

< previous page

page_180

next page >

Page 180 1982; Sherry & Schacter, 1987). In this way, information that is learned from past experiences can be brought to bear on a wide variety of novel experiences. Thus, an individual's behavior may be predicted and understood in light of the trait generalizations that have been made about the person. Likewise, a group member's behavior may be understood (or misunderstood) in light of the abstract stereotypes that have been formed about that person's group. This is what Bruner (1957a) referred to as going beyond the information given, and it is an important factor in the development of abstract knowledge. The second reason that the development of abstractions is efficient has to do with the need for streamlined representations and cognitive processes. Theoretically, it would certainly be possible to create generalizations solely via repeated exemplar activation and summation. However, there are at least three reasons to posit that this would not be the case. First, the ability to create and apply knowledge that may be used in response to general and invariant features of the environment may be inhibited by the levels of temporal, spatial, and contextual detail preserved in exemplars (particularly behavioral episodes; Sherry & Schacter, 1987). As such, it would be useful to formulate and store abstract knowledge in addition to exemplars. Second, the predictive validity of these summarizations depends on the number of experiences included in the summarization. This is simply a question of sample size. Because each experience represents a single sample from the environment population, it is necessary to aggregate across many experiences to obtain a reliable population estimate (McClelland et al., 1995). This is related to the third point – that it is simply more efficient to maintain prestored abstractions than it is to re-create them every time they are needed by retrieving and summarizing specific events or exemplars. Most researchers assume that there are capacity constraints on the number of relevant events that may be found and summarized in a timely manner (Bruner, Goodnow, & Austin, 1956; Hamilton & Mackie, 1990; McClelland et al., 1995; Nosofsky et al., 1994; Rothbart, Fulero, Jensen, Howard, & Birrell, 1978). To constantly do so would be inefficient compared with keeping a running summary of relevant exemplars that could be accessed whenever necessary. Indeed, the retrieval and application of specific episodes is more easily disrupted than the application of abstract knowledge structures (e.g., Rothbart, et al., 1978; Sherman & Bessenoff, 1999, as described later; see Johnson et al., 1993; Tulving, 1983, for reviews). Thus, although it would be theoretically possible to retrieve and summarize large numbers of exemplars to generate valid population estimates, it would be inefficient to do so.3 This discussion makes clear the intimate relationship between form and function. The development, storage, and use of abstract knowledge are based, in part, on the need for efficient use of memory. Abstractions provide relatively large amounts of generalizable knowledge at relatively little cognitive expense, providing a healthy ratio of information gained to effort expended. Supporting this view is the impressive spontaneity with which perceivers draw abstractions from particular episodes. Past research has shown that perceivers derive abstract trait representations about the self and others relatively quickly and without being encouraged (e.g., Klein & Loftus, 1993; Klein et al., 1992; Sherman & Klein, 1994). Other research has shown that the leap from behavioral perception to trait inference occurs spontaneously, if not automatically (see Uleman, Newman, & Moskowitz, 1996, for a review). Related findings in a variety of experiments led Hastie and Park (1986) to conclude that it was difficult to invent situations in which abstraction processes did not spontaneously occur (see also Anderson, 1989). Of most direct relevance to the concerns of this chapter, my work on stereotype representation (Sherman, 1996) demonstrated that much the same thing happens when perceivers are learning about social groups. As information about the group is acquired, trait summaries are extracted. Moreover, this research showed that group judgments were based on exemplars only when an abstract stereotype was not available. This suggests that there is a functional basis to the development of abstract stereotypes. It is more efficient to store and retrieve abstract stereotypes than it is to re-create group stereotypes anew each time they are needed by retrieving and summarizing information about particular group exemplars. Comparisons to Other Models Of course, the idea that stereotypes promote efficient social perception has a long history in social psychology. Both Lippmann (1922) and Allport (1954) focused extensively on the efficiency of stereotyping in their early analyses. More recent theoretical statements by Brewer (1988) and Fiske and Neuberg (1990) were also largely based on a functional analysis. Those models are specifically concerned with the extent to which judgments of individual group members are based on stereotypes about the groups to which they belong versus specific behaviors performed by the individuals. Both models rely on the distinction between “top-down’’ and “bottom-up” processing in their analyses. Whereas

< previous page

page_180

next page >

< previous page

page_181

next page >

Page 181 top-down processing relies on preexisting knowledge (such as stereotypes) to deduce the properties of a new stimulus, bottom-up processing relies on raw perceptual input (such as observed behaviors) to induce the properties of a stimulus. Both models argue that top-down processing is more efficient than bottom-up processing in forming impressions of individuals. Clearly there are important similarities between these models and the developmental/representational model described earlier (Sherman, 1996). Most significant, all three models argue that stereotype-based social perception is more efficient than perception based on more specific kinds of information (exemplars, behaviors).4 However, there are important differences between the models as well. One obvious difference is that, whereas the developmental/representational model describes the formation of group impressions or stereotypes (although it may be generalized to impression of individuals; Klein et al., 1992; Sherman & Klein, 1994), the dual process models developed by Brewer (1988) and Fiske and Neuberg (1990) are concerned with the formation of impressions of individuals. A second difference is that, whereas the developmental/representational model focuses both on the roles of efficiency in the initial extraction of abstract stereotypes and in their subsequent use, the dual process models focus only on the efficiency of using stereotypes that are already fully developed. Little attention is paid in the dual process models to exactly how and why stereotypes develop the way they do. Finally, and most important, the developmental analysis in our model is particularly concerned with detailing the evolving representational basis of stereotypes and how it is related to stereotype efficiency. In contrast, although they are concerned with the relative contributions of top-down (stereotype-based) and bottom-up (behavior-based) processes in social judgments, the dual process models are less concerned with the underlying representation of the top-down knowledge that is being applied and the implications that it has for stereotype use. Fiske and Neuberg suggested that stereotypes may be either abstraction or exemplar based (e.g., Fiske, Neuberg, Beattie, & Milberg, 1987), but no discussion was devoted to identifying the conditions that determine the type of representation or what difference the two types of representation would make in stereotyping processes. Although Brewer (1988) clearly defined stereotypes as abstract group prototypes, little attention was given to how the abstract nature of stereotypes particularly influenced information processing. Thus, whereas the developmental/representational model is a model of both representation and judgment processes, the dual process models are primarily models of judgment processes. As we see later, the focus on underlying representation in the developmental/representational model provides insights into questions surrounding stereotype efficiency and encoding processes that are not addressed by the dual process models. First, however, a broader discussion of stereotype efficiency and encoding processes is warranted. STEREOTYPE EFFICIENCY AND THE ENCODING OF SOCIAL INFORMATION As described previously, it is widely accepted that stereotyping is more efficient than individuation (Bodenhausen et al., 1999; Brewer, 1988; Fiske & Neuberg, 1990; Hamilton & Sherman, 1994). This view has been supported by a considerable amount of research in recent years that has shown that stereotyping is particularly prevalent when attentional capacity is depleted. Whether due to physical depletion (e.g., Bodenhausen, 1990; Kim & Baron, 1988), task difficulty (e.g., Bodenhausen & Lichtenstein, 1987; Kruglanski & Freund, 1983; Pratto & Bargh, 1991), multiple task demands (Gilbert & Hixon, 1991; Macrae, Hewstone, & Griffiths, 1993), anxiety-induced arousal (e.g., Wilder & Shapiro, 1989), or positive moods (e.g., Stroessner & Mackie, 1993), situations that decrease the availability of processing resources have been shown to increase the extent to which perceivers rely on their stereotypes. Miser Models: Stereotypes as Crutches Historically, two different types of models have been proposed to account for the relative efficiency of stereotyping compared with individuation. First, many researchers have interpreted such findings from a cognitive miser perspective (e.g., Bodenhausen & Wyer, 1985; Brewer, 1988; Fiske & Neuberg, 1990; for an overview, see Fiske & Taylor, 1984). According to this orientation, perceivers are miserly with their resources and devote as little processing capacity to a given task as possible — what Allport (1954) called “the principle of least effort” (p. 173). The application of this analysis to stereotyping is further rooted in the heuristic processing tradition in the judgment and decision making literature, which suggests that people often rely on short-cuts or heuristics to achieve resource conservation (March & Simon, 1958; Tversky & Kahneman, 1974). As such, miser models of stereotyping suggest that perceivers rely on stereotypes as judgmental heuristics to relieve themselves of the cognitive effort of having to systematically process individuating information. Alhough these models are clear that perceivers may be more or less motivated to

< previous page

page_181

next page >

< previous page

page_182

next page >

Page 182 process systematically (Brewer, 1988; Fiske & Neuberg, 1990), stereotype use is always equated with processing in a miserly and unmotivated fashion. Thus, as perceivers become more miserly, stereotype use increases and individuation decreases; as perceivers become more motivated, stereotype use decreases and individuation increases. Of course, as capacity-saving devices, stereotypes are also presumed to be particularly useful when resources are depleted and careful processing is difficult to achieve. Given these models' historical basis in the judgmental heuristics tradition, it is not surprising that their primary focus has been on predicting the outcomes of judgment processes. In particular, they have been largely concerned with predicting the relative weights that will be given to category-based (i.e., stereotypes) and individuating information in determining the content of evaluative and descriptive judgments about group members. However, stereotypes and individuating information do not merely exert independent and separate influences on social perception. Rather, stereotypes and individuating information mutually influence and constrain one another. Of particular relevance to this chapter, stereotypes have been shown to have a significant impact on the manner in which stereotype-relevant individuating information is attended to, encoded, and retrieved. Furthermore, concerns for processing efficiency play an important role in determining exactly how stereotypes influence these processes. Though the miser models have not been so concerned with delineating these effects, other models have made them of central importance. Stereotypes as Filters The second type of model that has been proposed to explain stereotype efficiency is the filter model (e.g., Bodenhausen, 1988; Bodenhausen & Lichtenstein, 1987; Bodenhausen, Macrae, & Garst, 1997; Hamilton & Sherman, 1994; Macrae et al., 1993; Macrae, Milne, & Bodenhausen, 1994; Macrae, Stangor, & Milne, 1994; Miller & Turnbull, 1986; Stangor & Duan, 1991; Stangor & McMillan, 1992; Taylor & Crocker, 1981). In contrast to miser models, filter models are largely concerned with how stereotypes influence the processing of individuating information. Also, in contrast to miser models, filter models do not suggest that increased stereotype use is always accompanied by diminished reliance on individuating information. Rather, certain kinds of individuating information benefit from the use of stereotypes. Based on schematic principles of memory (e.g., Minsky, 1975; Neisser, 1976), these models suggest that information that is consistent with a stereotypic expectancy is encoded and represented in memory more completely than information that is unrelated to the expectancy. In contrast, information that is inconsistent with the expectancy is particularly unlikely to be successfully encoded into memory. Thus, stereotypes act as filters that let in information that confirms people's expectancies and keep out information that disconfirms them. There are two varieties of filter models. The weak or passive filter model suggests that the encoding advantage for consistent information is based on the fact that, because it fits with existing expectancies, it is simply easier to comprehend than inconsistent information. By providing this conceptual fluency to consistent information, stereotypes reduce the amount of capacity necessary to encode that information, freeing up processing resources for other tasks. The strong or active filter model further suggests that, because consistent information is easier to encode, attention is directed toward consistent and away from inconsistent information (Bodenhausen, 1988; Bodenhausen & Lichtenstein, 1987; Bodenhausen et al., 1997; Fiske & Neuberg, 1990; Hamilton & Sherman, 1994; Macrae, Milne, & Bodenhausen, 1994; Stangor & Duan, 1991; Taylor & Crocker, 1981). Thus, stereotypes further promote efficiency by directing resources away from information that is difficult to encode and toward information that is easy to encode. This attentional hypothesis is based on the cognitive miser analysis that people generally prefer to do as little cognitive work as necessary. It is also derived, in part, from principles of selective exposure (e.g., Festinger, 1957; Frey, 1986), which suggest that people prefer to not attend to information that challenges their beliefs (see also Thompson, Naccarato, Parker, & Moskowitz, chap. 2, this volume), particularly if they do not possess the resources to refute that information. Indeed, it is important to note that, as is the case with miser models, both filter models predict that the influence of stereotypes is especially likely to be observed when processing resources are low and efficient processing is at a premium. Thus, in these conditions, the encoding advantages for consistent over inconsistent information should be particularly strong. However, in contrast to miser models, these models suggest that perceivers who are particularly motivated to process target information may be especially likely to rely on their stereotypes to arrive at a preferred impression. STEREOTYPE REPRESENTATION, ENCODING FLEXIBILITY, AND STEREOTYPE EFFICIENCY Recently we have proposed and tested an alternative

< previous page

page_182

next page >

< previous page

page_183

next page >

Page 183 model of stereotype efficiency that makes different predictions from both the miser and filter models (Sherman & Frost, 2000; Sherman, Lee, et al., 1998). This model is particularly concerned with the manner in which stereotyperelevant individuating information is attended to and encoded, and how concerns for efficiency influence these processes. The predictions for this model were derived in part from our earlier work on the mental representation of stereotypes (Sherman, 1996), in which the presence of an abstract stereotype was shown to reduce the extent to which stereotype-consistent group judgments were based on particular stereotype-consistent behaviors. In conditions in which perceivers possessed abstract stereotypes, they were freed from having to rely on specific stereotypical behaviors as the basis for their judgments. Note that the implications of these findings are opposed to the predictions of filter models (particularly strong filter models). Whereas those models suggest that stereotypes increase the encoding and influence of stereotype-consistent behaviors, our results suggest that stereotypes decrease the importance of carefully encoding stereotypical behaviors. Because stereotypical information is already provided by abstractions, the necessity of attending to, carefully encoding, and retrieving confirmatory behaviors is reduced (see also Johnston & Hawley, 1994; von Hippel, Jonides, Hilton, & Narayan, 1993; von Hippel, Sekaquaptewa, & Vargas, 1995). Moreover, whereas filter models suggest that encoding should be particularly biased toward expected information when capacity is depleted, our results imply that careful encoding of expected information is particularly unlikely under these conditions. If, as we have suggested, it is more efficient to rely on abstract knowledge than behavioral exemplars, then it would seem unlikely that stereotypical behaviors would be particularly well encoded when processing resources are scarce. What Is an Efficient System? However, our model is not a miser model. We do not wish to suggest that perceivers rely on stereotypes out of laziness or that stereotype use necessarily implies the disuse of individuating information. On the contrary, rather than minimizing effort, we argue that stereotype use is about maximizing efficiency — the amount of information gained for effort expended. Indeed, the other major impetus for our model was a theoretical consideration of what an efficient cognitive system ought to accomplish. A number of researchers have argued that cognitive systems that are either too stable or too flexible would be at an evolutionary disadvantage (e.g., Johnston & Hawley, 1994; Sherry & Schacter, 1987; Tulving, Markowitsch, Kapur, Habib, & Houle, 1994). On the one hand, if knowledge is too easily changed, then it is not useful for making consistent predictions about the environment. Hence, there must be some stability in our expectations. On the other hand, if expectations are not open to alteration, they may not adequately represent the world. If expectancies have no predictive value, then they also are not very useful. At some point, people must be able to recognize that the data do not fit their expectancies, and they must revise them. Thus, for maximum predictive value, efficient systems must encode both invariances in the environment, which encourage the development of expectancies, and variances (unexpected events), which suggest that expectancy reorientation may be necessary (e.g., Johnston & Hawley, 1994; McClelland et al., 1995; Nosofsky et al., 1994; Schank, 1982; Sherry & Schacter, 1987; Tulving et al., 1994). Given the importance of each of these goals, it might be expected that there would be provisions in the cognitive system for encoding both expected and unexpected information when processing resources are scarce. This functional analysis is at odds with both miser and filter models. Both of those types of models propose cognitive systems that are inherently conservative. According to miser models, when resources are depleted, perceivers rely on their existing stereotypic expectancies to the relative exclusion of all novel behavioral information. Strong filter models suggest that, when capacity is low, people focus on confirmatory behaviors and direct their attention away from disconfirming information. Finally, weak filter models propose that, when resources are low, perceivers are unable to encode unexpected behaviors because they are simply too difficult to comprehend. Alhough people may want to encode this information, they are simply unable to do so. Thus, in all three models, the overwhelming trend is toward maintaining stability in existing expectations, particularly when capacity is restricted. In contrast, a model based on the notion of an efficiency-maximizing system suggests that efficient stereotype use ought to facilitate, in different ways, the encoding of both expected and unexpected individuating information when capacity is low. Encoding Flexibility and Stereotype Efficiency Recently, we have proposed and tested such an encoding flexibility model of stereotype efficiency (Sherman, Lee, et al., 1998). According to this model, stereotypes facilitate the encoding of consistent behaviors by providing explanatory frameworks that render those behaviors conceptually fluent. As such, those behaviors are relatively easy to comprehend, even when processing

< previous page

page_183

next page >

< previous page

page_184

next page >

Page 184 resources are scant. This aspect of our model coincides with the passive filtering hypothesis. However, our model suggests that the conceptual fluency of consistent information has a different impact on the encoding of consistent and inconsistent information than do filter models. In particular, we argue that, because consistent information can be understood with relatively little effort, substantial attention is not devoted to encoding the details of this information (e.g., Graesser, 1981; Johnston & Hawley, 1994; Sherman, 1996; von Hippel et al., 1993), particularly when capacity is depleted. Instead, those resources may be redirected to assist in the encoding of inconsistent information, which is difficult to understand. Yet this does not mean that the inconsistent information is fully understood — only that the effort will be made. Thus, when capacity is low, conceptual encoding (encoding for gist meaning) favors consistent information, whereas attentional allocation and perceptual encoding (encoding for details) favor inconsistent information. As such, stereotypes do not merely simplify impression formation for lazy perceivers. Rather, they permit the flexible distribution of resources in a way that maximizes the amount of information gained for the effort expended. This encoding flexibility is functional because it promotes both stability and plasticity in the mental system. The results from a number of experiments have supported the predictions of this model. Three experiments tested our hypotheses about the distribution of attentional resources under different encoding conditions (Sherman, Lee, et al., 1998). In all three experiments, participants read stereotype-consistent and -inconsistent information about a target person while either under a cognitive load or not. The first experiment measured the amount of time participants spent reading the consistent and inconsistent behaviors as a function of processing capacity. Results show that participants spent an equal amount of time reading consistent and inconsistent information when capacity was high, but spent a greater amount of time reading inconsistent than consistent information when capacity was low. Thus, participants were not directing their attention away from the inconsistent information when they were under load. The second experiment used a dual-task paradigm to examine the amount of attention paid to consistent and inconsistent information as a function of cognitive capacity. As participants read about the target, they were also asked to monitor auditory tones emitted by their computers. When capacity was high, participants responded to this secondary task equally quickly, regardless of whether it occurred during the encoding of consistent or inconsistent items. In contrast, when capacity was low, participants took more time to respond to the secondary task when it occurred as inconsistent items were being encoded than when consistent items were being encoded. This response interference demonstrates that greater attention was devoted to encoding the inconsistent than consistent items when capacity was depleted. Finally, the third experiment forced participants to attend selectively to either consistent or inconsistent information by presenting consistent and inconsistent items in pairs for a brief period of time. The results show that, when capacity was high, participants recognized consistent and inconsistent items from a pair equally well. However, when capacity was depleted, the inconsistent item in the pair was recognized with significantly greater accuracy than the consistent item. Together these three experiments provide strong evidence that, when processing capacity is limited, greater resources are devoted to the encoding of inconsistent than consistent information, which is in direct opposition to the predictions of filter models. Two other experiments tested our hypotheses about the perceptual and conceptual encoding of stereotypeconsistent and -inconsistent information (Sherman, Lee, et al., 1998). Once again, participants read stereotypeconsistent and -inconsistent information about a target person while either under a cognitive load or not. Subsequently, they engaged in a priming task that measured either perceptual or conceptual encoding. Both tasks required participants to identify words that were flashed very briefly (33 ms) on computer screens. The perceptual priming task examined the extent to which participants could identify words that had appeared in the original stimulus items, but were unrelated to the gist meaning of the items (e.g., the word salesgirl from the sentence, “Swore at the salesgirl”). Participants' ability to identify these words reflects the extent to which the perceptual details of the items had been extracted during encoding. In contrast, the conceptual priming task examined the extent to which participants could identify words that reflected the gist meaning of the original stimulus items, but had not actually appeared in those items (e.g., trait terms such as kind and mean). Participants' ability to identify these words reflects the extent to which the gist meaning of the items has been extracted during encoding (for a methodological overview, see Roediger, 1990). The first experiment showed that perceptual encoding was greater for inconsistent than consistent behaviors under conditions of both high- and low-processing capacity. The second experiment showed that, under high-capacity conditions, conceptual encoding was equally strong for consistent and inconsistent behaviors. In contrast, when encoding capacity was limited, the conceptual meanings of consistent behaviors were much

< previous page

page_184

next page >

< previous page

page_185

next page >

Page 185 more likely to be extracted than the conceptual meanings of inconsistent behaviors. Thus, despite the attentional and perceptual encoding advantages for inconsistent information when resources were low, conceptual encoding favored consistent information under such conditions. These five experiments provide strong initial support for the encoding flexibility model of stereotyping. When resources are limited, stereotypes facilitate the encoding of both stereotype-consistent and -inconsistent information. Inconsistent information receives greater attention and more thorough perceptual encoding. However, despite these advantages, conceptual encoding favors consistent information in these same conditions. Through these encoding flexibility processes, when resources are scarce, stereotypes are able to promote their own stability (through conceptual encoding) while maintaining vigilance (through attentional distribution and perceptual encoding) that reorientation may become necessary. We argue that this encoding flexibility is facilitated by an abstract representational structure of stereotypes. As described previously, well-developed stereotypes are stored as abstract knowledge structures that summarize grouplevel information across time, situations, and individuals (Sherman, 1996). As such, they may aid in understanding novel behaviors performed by novel group members. This is particularly true for behaviors and individuals that are congruent with the basic gist of the stereotype. Because the information provided by these acts/individuals is redundant with, and may be assumed by, the broader abstraction, perceivers need not attend carefully to it. Instead, those resources may be directed to assist in the encoding of other information that does not fit so comfortably with the stereotype (e.g., unexpected information), particularly when resources are low. If ever, it may be in the early stages of stereotype development, when group-based expectancies are too weak to support the formation of abstract stereotypes, that attentional allocation and perceptual encoding would be biased toward stereotype-confirming information (e.g., Klayman & Ha, 1987; Skov & Sherman, 1986). In such cases, the stereotype is less useful for inferring the meanings of novel stereotypical behaviors. As such, particular stereotypeconsistent acts demand greater attention. Moreover, the stereotype may not present a strong enough expectancy to produce clearly identifiable, inconsistent data that would draw special attention (e.g., Hamilton & Sherman, 1996; Sherman, 1996; Sherman & Klein, 1994; Srull, Lichtenstein, & Rothbart, 1985). Finally, it may be that information that facilitates the establishment of an abstract expectancy that allows for future generalization may receive more careful processing than information that challenges an expectancy that is weak and ungeneralizable to begin with. Thus, prior to the completion of the abstraction process, consistent information may attract more attention and perceptual encoding than inconsistent information. However, as stereotypic expectancies congeal and become abstract, consistent information becomes redundant and may be readily inferred from the stereotype. At the same time, inconsistent information becomes easier to identify and gains in importance. As a result, resources are more likely to shift from consistent to inconsistent information. EFFICIENT PROCESSING INFLUENCES STEREOTYPE REPRESENTATION First I described how concerns for efficiency influence the initial development of abstract stereotypes. Next I showed that, once these abstractions are formed, they influence the efficiency with which stereotypes impact the subsequent encoding of stereotype-relevant information. Now I take the discussion full circle and describe how efficient encoding flexibility processes feed back into and influence the mental representation of stereotypes and stereotype-relevant information. One of the more important questions arising from our research on encoding flexibility has to do with the mechanisms of stereotype plasticity. The attentional and perceptual encoding advantages for inconsistent information under cognitive load must ultimately contribute to stereotype revision, but how? Among other reasons, we believe these encoding advantages are important for stereotype plasticity because they help people retrieve and reconstruct the details of unexpected events at a later time when greater resources are available for comprehension. Because these items of information are difficult to interpret during encoding (especially if resources are low), it is of particular importance to be able to retrieve their details at a later time, when their implications may be more fully understood. Retaining the details of these behaviors would also be important so that they may be compared to newly observed behaviors that may help clarify matters and promote conceptual consolidation (e.g., McClelland et al., 1995). In fact, a number of researchers have argued that the essential purpose of episodic memory is to record the details of unexpected events for later inspection and comparison (McClelland et al., 1995; Nosofsky et al., 1994; Schank, 1982; Sherman & Bessenoff, 1999; Sherry & Schacter, 1987). In contrast, retaining the specific details of expected information is not such a pressing matter. In this case, the basic gist may simply be extracted and stored as semantic memory.

< previous page

page_185

next page >

< previous page

page_186

next page >

Page 186 Our demonstrations of attentional and perceptual encoding advantages for stereotype-inconsistent information are certainly consistent with this functional analysis of episodic and semantic memory. However, we may also make specific predictions about the extent to which the encoding of stereotype-consistent and -inconsistent behaviors would encourage the development and use of abstraction-based versus exemplar-based group knowledge. In particular, if people simply extract the basic gist of consistent behaviors and store it as semantic memory, then we would expect perceivers to quickly develop abstract group knowledge pertaining to those behaviors. This is essentially what Sherman (1996) showed. In contrast, if people are more concerned with retaining the perceptual and contextual details of inconsistent behaviors, then we would expect that the development and use of abstract knowledge pertaining to those behaviors would be substantially delayed, and the role of exemplar-based processes would be enhanced. The results of two recently reported experiments support these hypotheses (Sherman, Klein, et al., 1998). Participants were first randomly assigned to one of two social categories (overestimators or underestimators) based on a minimal group manipulation (Howard & Rothbart, 1980). Next, they were presented with positive or negative descriptors of members of either their ingroup or their outgroup. Finally, participants performed the previously described judgment priming task that assessed the extent to which group judgments depended on the retrieval of information about particular group members (Sherman, 1996). When the descriptors confirmed participants' expectancies that ingroups would be positive and outgroups would be negative, judgments about the groups did not involve the activation of specific exemplars. Instead, judgments were apparently based on abstract group summaries created during the encoding of the expected behaviors. In contrast, when the stimulus information suggested that the ingroup was negative or the outgroup was positive, judgments of the groups were constructed by retrieving from memory specific group exemplars (for related results, see Maass, Milesi, Zabbini, & Stahlberg, 1995; Maass, Salvi, Arcuri, & Semin, 1989). This implies that perceivers did not form semantic summaries during the encoding of unexpected behaviors. These data suggest that, if perceivers are unable (or unwilling) to extract the basic gist of unexpected behaviors, those episodes may be stored and retrieved for future use. Based on our other work on encoding flexibility, we might expect that these encoding/representational differences between consistent and inconsistent behaviors would be particularly evident when processing capacity is restricted during the initial encoding of group behaviors, and inconsistent information is particularly difficult to comprehend. These studies demonstrate how the differential processing of stereotype-consistent and -inconsistent information may influence the manner in which that information gets represented in memory. Because perceivers are relatively unconcerned with encoding the specific details of consistent behaviors, that information may be ‘‘semanticized” and stored as abstract knowledge relatively quickly. By contrast, perceivers' concerns with encoding the details of inconsistent behaviors delay the abstraction process and enhance the role of episodic memory in the representation of such information. Thus, not only does the initial abstract nature of stereotypes influence the manner in which stereotype-relevant behaviors are encoded, but those encoding processes also subsequently feed back into and influence the mental representations of stereotypes and stereotype-relevant information. In both cases, concerns for efficiency are an important contributing factor. ON THE NEED FOR ENCODING FLEXIBILITY I have argued that it is important to encode and represent the perceptual and contextual details of stereotypeinconsistent but not -consistent information. In part, this is because the meanings of inconsistent behaviors are relatively difficult to discern during encoding. As such, it is useful to be able to reconstruct the details of those behaviors at a later time when further attempts at comprehension may proceed. However, there is another important reason that perceivers would more carefully encode the details of inconsistent than consistent behaviors. If the details of consistent behaviors are forgotten or are temporarily inaccessible, perceivers may simply rely on their abstract stereotypes to infer the basic gist of what has occurred. On the contrary, if the details of inconsistent behaviors cannot be retrieved, the meaning conveyed by those behaviors may also be lost if the initial conceptual encoding of those behaviors has been less than ideal. This is because the basic gist of inconsistent information cannot be reconstructed from existing abstract group knowledge after the fact. Thus, the careful encoding of the details of unexpected information not only helps advance the ultimate conceptual consolidation of that information, but also acts as insurance against the irreplaceable loss of such information. We conducted an experiment to examine this issue in more detail (Sherman & Bessenoff, 1999). One important kind of contextual information is source information. To fully benefit from episodic

< previous page

page_186

next page >

< previous page

page_187

next page >

Page 187 memory, perceivers must be able to accurately attribute events to their proper source (e.g., Johnson et al., 1993). Was the source of a news story the New York Times or the World Weekly News ? Was that great research idea mine or my graduate student's? Was it John or Juan who threw the first punch in the bar brawl? In this last case, stereotypes about the aggressiveness of Hispanic men may influence the likelihood that people attribute the first punch to Juan rather than John. In this experiment, we examined such source attributions for stereotype-consistent and -inconsistent information. In particular, we were interested in the extent to which perceivers would misattribute stereotype-consistent and -inconsistent behaviors to a person who did not, in fact, perform them. After reading lists of stereotype-consistent and -inconsistent behaviors that did and did not describe a target person, participants were asked to engage in a source monitoring task that required them to accurately attribute the behaviors to their proper source. Some participants were placed under a cognitive load as they performed this source monitoring task. Results show that, when the true target behaviors and false behaviors shared a similar encoding context (i.e., were encoded in the same time and place), and were therefore easily confused, participants relied on their stereotypes as cues in the source monitoring task and misattributed more false stereotype-consistent than -inconsistent behaviors to the target. Because the proper source of the behaviors was difficult to discern under these conditions, participants relied on their stereotypes as judgmental cues (see also Banaji & Greenwald, 1995). However, an interaction with the processing capacity variable demonstrated that this effect was only found when participants' ability to rely on episodic memory was impaired by the imposition of the cognitive load. Thus, when participants needed to rely on episodic memory because there was source/context confusion, but were unable to do so because of the imposed cognitive load, they relied on the stereotype as a heuristic cue in making their source attributions. In contrast, the stereotype was not used as a source cue (i.e., an equal number of misattributions were made for stereotype-consistent and -inconsistent behaviors) when participants possessed full processing capacity, although source confusion was high. In this case, participants relied on a more systematic analysis of episodic memory to reconstruct the source information about the behaviors and make their attributions. These results highlight the necessity of carefully encoding and representing the episodic details of unexpected behaviors. When episodic memory is not available to perceivers, as in the cognitive load condition of this study, they may yet rely on abstract knowledge to guide social cognition. In such cases, the basic gist of stereotype-consistent behaviors may be inferred even if the details of those behaviors are inaccessible. In contrast, the ability to rely on abstract representations to fill in knowledge gaps does not extend to stereotype-inconsistent behaviors, for which there are no relevant abstractions. Therefore, it is of critical importance to retain the episodic details of these behaviors, particularly to the extent that their conceptual meanings have not been extracted during encoding. This suggests why the encoding flexibility processes outlined earlier (i.e., directing attentional and perceptual encoding toward unexpected behaviors when capacity is depleted) may be pursued in the first place: to ensure the adequate representation of unexpected events that may not be reconstructed if lost. SUMMARY AND FUTURE DIRECTIONS That mental representations and cognitive processes are mutually constraining in memory and judgment processes is axiomatic. Nevertheless, social-cognitive research on representation and processing has frequently proceeded in an independent fashion, with too few explicit attempts having been made to take stock of what the findings in one domain imply for theoretical development in the other domain (notable exceptions include Carlston, 1994; Klein & Loftus, 1993; Smith, 1990, 1996; Smith & DeCoster, 1998; Srull & Wyer, 1989; among others). The main goal of this chapter has been to demonstrate the utility of analyzing representation/processing dynamics in the context of stereotyping research. I began by describing research aimed at specifying the representational basis of stereotypical knowledge. The initial theoretical impetus for this research and the representational model that evolved from it were based, in large part, on a functional analysis of efficient information processing, and how it may be facilitated by the development of generalizable knowledge structures (e.g., Bruner, 1957b; Klein & Loftus, 1993; Klein et al., 1992; Lippmann, 1922; McClelland et al., 1995; Nosofsky et al., 1994; Schank, 1982; Sherman & Klein, 1994; Sherry & Schacter, 1987). Thus, our theorizing about representation was heavily influenced by existing theories of processing from the outset. Subsequently, I described our recently developed model of stereotype efficiency and encoding flexibility. The development of this model was heavily influenced by the conclusions drawn from our initial work on stereotype representation. The representational perspective provided by that initial research enabled us to make novel (and sometimes nonintuitive) predictions about

< previous page

page_187

next page >

< previous page

page_188

next page >

Page 188 stereotype-based encoding processes that had not been identified by prior models of stereotype efficiency. Thus, in this case, it was the knowledge of representational issues that informed the research on processing. Next, I presented research demonstrating how concerns about efficient processing and the differential encoding of stereotype-consistent and -inconsistent information feed back into and influence the mental representation of stereotype-relevant information. Thus, not only does the representational nature of stereotypes influence processing, but so too do processing strategies influence the representational nature of stereotypes. Finally, I described an experiment that further illustrated how the representational nature of stereotypes influences encoding processes and why it is that efficient encoding proceeds in the manner that it does. Altogether, the research described in this chapter makes a strong case for the value of integrating analyses of mental representation and information processing. On Efficiency One common factor running through all of the research described in this chapter is that stereotypes and stereotyping are influenced by perceivers' need for cognitive efficiency. Both the initial development and nature of stereotypes as well as their subsequent uses in social perception are molded by such concerns. In no way is this meant to suggest that other factors (e.g., identity-related motives, ego-defensive motives, etc.) are not critically important in stereotyping processes. Nor is this meant to imply that efficiency concerns are necessarily the most important factor in understanding stereotypes. Nevertheless, it is clear that efficiency is one critical factor that must be considered in models of stereotype representation and use (cf. Bodenhausen et al., 1999; Kunda & Thagard, 1996). Although the important role of cognitive efficiency in stereotyping has long been recognized (e.g., Allport, 1954; Brewer, 1988; Fiske & Neuberg, 1990; Hamilton & Sherman, 1994; Lippmann, 1922), the findings outlined in this chapter suggest that some refinements may be needed in how stereotype efficiency is conceptualized. The efficient stereotype typically has been conceived of as either a crutch or a filtering device. Both metaphors stress the effortreducing and simplifying properties of stereotypes as well as the conservative, stereotype-confirming nature of efficient processing. However, efficiency is defined here as the ratio of information gained (production) to effort expended (cost). As such, considerations of stereotype efficiency must consider not only savings in effort but also gains in information. Our research (Sherman & Frost, 2000; Sherman, Lee, et al., 1998) demonstrates that the primary motive behind stereotyping when capacity is depleted is not always simplicity for simplicity's sake. Perceivers are not so much interested in effort reduction per se as they are in efficiently distributing their resources. Furthermore, all aspects of efficient encoding are not motivated by, and do not necessarily result in, stereotype reconfirmation. Rather, both stereotype stability and plasticity are reinforced in different ways through the efficient use of stereotypes. It is important to emphasize that, in contrast to claims made by some researchers (e.g., Oakes & Turner, 1990; Spears & Haslam, 1997), the idea that stereotypes are used to enhance rather than reduce information processing is in no way inconsistent with the view that stereotype use is often influenced by concerns for efficiency (for further discussion of this matter, see Sherman, Macrae, & Bodenhausen, 2000). To the contrary, stereotype contributions to information gain are one critical component of their efficiency, as demonstrated by our research. It may be true that researchers have tended to place greater emphasis on the effort reduction than the information gain side of the equation. However, our research in no way challenges (and, in fact, strongly supports) the basic point that stereotype use is influenced by the need for efficiency. Implications for Dual-Process Models of Stereotyping. The model of efficiency we have outlined and the accompanying data have important implications for a number of central issues in stereotyping research. Dual process models have been developed to account for the conditions under which judgments about individual group members are dependent on top-down, stereotype-driven processes versus bottom-up integration of target behaviors (see Bodenhausen et al., 1999, for a review). In both of the prominent dual process models of stereotyping (Brewer, 1988; Fiske & Neuberg, 1990), increases in stereotype use are associated with decreases in individuation, particularly in the amount of attention paid to and the encoding quality of stereotype-inconsistent information. That is, because perceivers are using their stereotypes, they need not expend extra resources on difficult individuating processes. However, our results show that decreases in processing capacity increased both stereotyping processes (via conceptual encoding) and certain individuating processes (attentional allocation toward and perceptual encoding of inconsistent compared with consistent information) at the same time. These results demonstrate that efficiencydriven stereotype use need not exclude individuation processes. In fact, they suggest that stereo-

< previous page

page_188

next page >

< previous page

page_189

next page >

Page 189 type use and individuation should be conceived as two separate but related continua, rather than as mutually exclusive processing modes (see Chaiken, Liberman, & Eagly, 1989). As such, movement along the two continua may proceed along different dimensions of encoding at the same time. Thus, stereotyping may be increased via one mode of encoding (e.g., conceptual), while individuation is increased via a different mode of encoding (e.g., perceptual) simultaneously. The Interaction of Motivation and Processing Capacity in Stereotyping. In a related matter, our findings also shed light on the relationship between perceiver motivations and processing capacity in stereotyping. Dual process models of stereotyping suggest that motivations to individuate a target may override or diminish the influence of stereotypes (see Monteith, chap. 23, this volume). Yet these and other models of stereotype efficiency suggest that, when processing capacity is depleted, the impact of perceivers' processing goals are limited, The working model has been that motivations may be realized only if sufficient capacity is available. Thus, perceivers who may have initially been motivated by accuracy to individuate a target are presumed to abandon that motivation when resources are scant, and revert to more miserly pursuits such as purely heuristic processing or schematic filtering processes. In contrast, our results suggest that motivations continue to exert an important influence when cognitive resources are depleted (see also Moskowitz, Gollwitzer, Wasel, & Schaal, 1999). In our efficiency studies, participants were largely motivated by concerns for accuracy and did not forsake those concerns when capacity was diminished. Rather, they distributed their resources in such a way that enabled them to flexibly encode different aspects of both stereotype-consistent and -inconsistent information. Of course, there are many situations in which perceivers are motivated by concerns for ego-defense or social identity rather than by accuracy. In these contexts, we would expect that perceivers would, in fact, be especially likely to engage in miserly/filtering processes if resources are scant. Thus, stereotypes are flexible tools that are adapted to the current goals of the user. However, resource scarcity does not determine the processing goal, but rather how an already chosen goal is pursued. Conclusion The preceding analyses suggest that stereotypes are much more versatile tools than crutches or filters. Perhaps the metaphor of the Swiss Army knife, with multiple tools working simultaneously, better captures the flexibility with which efficient processing may be pursued. In terms of stereotyping, the tools correspond to different attentional, encoding (including perceptual and conceptual encoding), and inference processes that proceed in parallel. On some occasions, the tools may be working in concert toward achieving a single goal (e.g., stereotype confirmation). On other occasions, different tools may be used to pursue different goals at the same time (as in the case of encoding flexibility processes). One of the major challenges for future research on stereotyping should be to more closely examine how different processing conditions (e.g., variations in attentional capacity) and different processing motives (e.g., accuracy vs. defense vs. self-presentation) interact to determine how perceivers use stereotypes in forming impressions of others. This challenge must be met with a more complete understanding of cognitive efficiency that takes into account both effort reduction and information gain strategies, as well as both stability-maintaining and plasticity-seeking processes. Finally, progress in these endeavors can be greatly advanced by further attempts to integrate what is known about the mental representation of stereotypes and the processing of stereotype-relevant information. ACKNOWLEDGMENTS Preparation of this article was supported by NIMH Grant 55037 to Jeffrey W. Sherman. Thanks to Galen Bodenhausen, Jim Sherman, and Eliot Smith for their thoughtful comments. ENDNOTES 1. At still a third level, the mental representation of a stereotype may be defined in terms of a particular representational format (e.g., associative network models vs. localist connectionist models vs. distributive connectionist models). This level of analysis attempts to define at some hypothetical neural level the manner in which knowledge about a group is stored in memory. All of the prominent format models describe a memory system that is based on activation spreading between associated quasineuronal units or nodes that represent concepts or their features (e.g., connections or overlapping patterns of connections between social category representations and stereotypical features). Although the different models share much in common in terms of their basic metaphors, they make different assumptions about the nature of memory nodes, associations, and the ways that activation spreads through a representation. A full discussion of the implications of each of these models for stereotyping goes well beyond the scope of this chapter

< previous page

page_189

next page >

< previous page

page_190

next page >

Page 190 (for an excellent review, see Smith, 1996). However, it is important to note that each type of format can (and must) account for the storage and retrieval of both exemplars and abstractions. As such, the issues surrounding the differential impact of abstraction- and exemplar-based stereotypes discussed in this chapter are largely independent of questions of format. 2. Note that demonstrations that accessible exemplars may influence group judgments do not indicate that abstract group impressions do not exist. Accessible exemplars may influence group judgments in addition to, and not instead of, preexisting abstract knowledge. In a similar fashion, the fact that group judgments may be influenced by contextual variations does not indicate a lack of abstract knowledge. 3. It is important to note that this is not an argument about economy of storage. In this model, all encountered exemplars may be retained in memory. However, storing those exemplars and using them are different matters. The point here is that retrieving and summarizing large numbers of exemplars is less efficient and more easily disrupted than activating a stored abstraction. 4. There are certainly cases in which attribute-based impressions are as easily derived as stereotype-based impressions. There are also cases in which stereo-type-based inferences depend more on bottom-up and less on top-down processes than attribute-based inferences. This has led some researchers to conclude that the distinction between stereotyping and individuation is not a useful one, and that stereotype use is not inherently more efficient than individuating (e.g., Kunda & Thagard, 1996). Although we concur that it is not necessarily the case that stereotyping will always be more efficient than individuation, we also believe that it just so happens to be true most of the time. This is the case for a number of reasons. For example, whereas stereotypes typically provide access to knowledge along a wide range of dimensions, many attributes do not generalize beyond their own specific meaning (Andersen, Klatzky, & Murray, 1990). Moreover, some kinds of individuating information (e.g., behavioral information) must be extracted in a bottom-up fashion, which is rarely the case for many stereotypes (particularly those associated with obvious physical characteristics). For a complete discussion of this matter, see Bodenhausen, Macrae, and Sherman (1999).

< previous page

page_190

next page >

< previous page

page_191

next page >

Page 191 12 A Cornerstone for the Science of Interpersonal Behavior? Person Perception and Person Memory, Past, Present, and Future Leonard S. Newman University of Illinois at Chicago “To discover where the action is in psychology,” begins a recent undergraduate social psychology text, “look to social psychology” (Zimbardo, 1993). However, the typical beginning student may be surprised to learn that for many years much of the action has involved having people record their impressions of Donald – a man notable only for the ambiguity of his behavior (e.g., Higgins, Rholes, & Jones, 1977); recording how many of a hypothetical person's honest and dishonest behaviors study participants can remember (Hastie & Kumar, 1979); asking college freshmen and sophomores what they really think about the woman being interviewed on the silent videotape (Gilbert, Pelham, & Krull, 1988); figuring out whether people spontaneously inferred that the secretary who solved the mystery halfway through the book is clever or that the plumber who slipped an extra $50 into his wife's purse is generous (Winter & Uleman, 1984); and determining what leads to the inference that the businessman who tripped over his girlfriend's feet while dancing the fox-trot did so because he is clumsy (McArthur, 1972). As the late Roger Brown (1986) observed, for “the new student” it is hard to “grasp what it is that led’’ the people who do this kind of research “to care about the questions they ask; the original lure is lost” (pp. 378–379). Brown need not have restricted that observation to the new student; some prominent social psychologists (e.g., Aronson, 1992; Ickes & Gonzalez, 1994; see also Neisser, 1980) have also registered their discomfort with a field dominated by such investigations (i.e., a field dominated by laboratory studies of person perception and person memory). Why have social psychologists devoted so much time and energy to observing in microscopic detail the processes involved in making sense of people's behavior? And why should they continue doing so? The remainder of this chapter is devoted to addressing these questions. PERSON PERCEPTION AND PERSON MEMORY As defined by Gilbert (1998), person perception (or personology, as he prefers to call it) is the study of how “people come to know about each other's temporary states (such as emotions, intentions, and desires) and enduring dispositions (such as beliefs, traits, and abilities).” Viewed in this way, the term person perception also encompasses much of the research that falls under the categories of impression formation, social categorization, social inference, attribution, and even social construct accessibility. In other words, person perception research is concerned with how, when we are exposed to other people's behavior, we figure out what they did, why they did it, and what kind of people – including how likable – they are. Classic studies in this area include those by Higgins et al. (1977), Jones and Harris (1967), and Anderson (1965a, 1965b). The priming research of Higgins et al. shed light on the processes involved in categorizing other people's behavior. Participants in their studies were asked to form impressions of people based on descriptions of their behavior that could be construed and evaluated in different ways (e.g., engaging in high-risk sporting activities). In a previous (and bogus) perception study, they had been exposed to trait terms that could be used to disambiguate the behaviors (e.g., adventurous, a favorable trait, or reckless, an unfavorable one). Higgins et al. found that interpretations of the behaviors tended to be consistent with the primed traits. These findings indicate that the sheer accessibility of trait concepts – regardless of whether the accessibility stems from an irrelevant source – significantly affects how social information is categorized. Jones and Harris (1967) focused instead on how people explain behaviors that they have already categorized. They asked participants to guess the true attitudes of people who had delivered

< previous page

page_191

next page >

< previous page

page_192

next page >

Page 192 unambiguously favorable or unfavorable speeches about Fidel Castro. In some conditions, the speakers had been randomly assigned to the positions they took, and this was quite apparent to the participants. Nonetheless, participants underutilized this situational information and generally assumed that overt behavior (regardless of how constrained) reflected underlying attitudes and dispositions. Finally, Anderson (1965a, 1965b), building on a procedure introduced by Asch (1946), systematically varied the number, valence, and order of personality traits that were attributed to hypothetical people and asked participants in his study to form impressions of those people on the basis of the different trait aggregations. Together these three studies (all of which are being replicated and extended to this day) represent many of the traditional concerns of person perception researchers, including what determines how behavior is interpreted and categorized; how information about people, their behavior, and the situations they are in gets integrated when one tries to infer the meaning of their behavior; and how trait inferences and other pieces of individuating information are combined into a coherent (especially evaluative) impression (see Fiske & Taylor, 1991; Gilbert, 1998; Jones, 1990; Krull, chap. 13, this volume, for reviews of this research).1 In contrast, person memory is the study of how information about other people is represented in memory. The central issues are what kind of information gets stored, how it is organized, and what subset of that information is most accessible for later recall. Typical dependent measures in person memory studies include the recall and recognition of people's traits and behaviors, the order in which pieces of information are recalled, and the latency of response to memory probes. Hastie and Kumar's (1979) seminal research is perhaps the most frequently cited in this area. Participants in Hastie and Kumar's studies were told that a person could be characterized by a particular personality trait (e.g., intelligence or dishonesty) and then they were exposed to various behaviors performed by that person. Some of those behaviors were consistent with expectations, some were inconsistent (e.g., dishonest behaviors performed by an honest person), and some were irrelevant to the trait. Then participants were asked to recall the behaviors. This research is remembered most for its counterintuitive findings: Participants’ memory for inconsistent behaviors was better than their memory for consistent behaviors. But of more significance to the study of person memory is that those findings led Hastie and Kumar to propose a general model of how information about people is represented in long-term memory. This model was further refined on the basis of research by Srull (1981) and others. In general, the model assumes that information about a given person – behaviors, traits, and general evaluations – is stored in a particular location in a hypothetical memory space. Each piece of information occupies a node, and nodes can be connected by associative pathways. These pathways are formed when two pieces of information are thought about in relation to each other. The probability of recalling a piece of information is a function of the number of pathways leading to it. This kind of model can easily account for the prominence in memory of information that is incongruent with one's general impression of a person. Such behaviors prompt efforts to resolve the inconsistency, leading one to consider the behavior in light of other information one has about the person. That, in turn, leads to the formation of many associative links involving the incongruent information. If the beginning student of social psychology is surprised to discover how much effort has been devoted to studying person perception and person memory, another surprise is that the two research areas generally exist as separate literatures (cf. Wyer & Srull, 1989). Wyer and Carlston (1994) noted that a “limitation of research on the cognitive representation of social information is that, with a few exceptions, it has proceeded virtually independently of research on social inference” (p. 43). Newman and Uleman (1993) made the same point at some length, showing how certain distinctions that are quite prominent in social inference research (e.g., the distinction between categorizing or identifying behaviors in trait terms and inferring that the people emitting those behaviors have those traits; see Trope, 1986) have no obvious counterparts in many person memory models. More recent evidence of this sealing off of the two literatures is the chapter on “Personology” (Gilbert, 1998) in the latest Handbook of Social Psychology (Gilbert, Fiske, & Lindzey, 1998). It is an extensive and thoughtful review of person perception research that has next to no references to the person memory literature. Person memory is covered in a separate chapter of the handbook (Smith, 1998). An assumption of this chapter, however, is that the two areas of research not only emerge from a common perspective but are inextricably bound. As social beings, we encounter other people; we categorize their behavior (Cantor & Mischel, 1979) and generate immediate, automatic impressions of them (Uleman, Newman, & Moskowitz, 1996); those impressions can be shaped by chronically or temporarily accessible knowledge structures (Higgins, 1996); we can generate alternative interpretations of a person's behavior and modify our

< previous page

page_192

next page >

< previous page

page_193

next page >

Page 193 impressions of that person based on other, less accessible knowledge structures (Martin & Tesser, 1992); or, we can go beyond our initial impressions to take into account contextual information (Gilbert, 1989); our impressions and inferences are then stored in some organized way in memory (Carlston & Smith, 1996; Smith, 1998; Wyer & Carlston, 1994); some subset of that information can or will be retrieved when we later encounter the target person again; and how and how much we process new information about the person will be affected by our previous impressions (Srull & Wyer, 1989). Our representations of other people are shaped by what we infer about them and their behavior, and what we infer about people is affected by our preexisting representations of them. Both person perception and person memory researchers focus on the general cognitive processes involved when we make sense of each other's behavior, and to do justice to those processes, they need each other. Therefore, the remainder of this chapter (as much as possible) downplays the distinction between person perception and person memory research. THE PAST Beginnings To return to the question posed earlier, how did social psychology come to focus on these topics? One way to answer this question is to take a historical approach. To briefly recount a story told in greater detail by others (e.g., Gilbert, 1998): The earliest researchers who identified their specialty as person perception in the postwar period focused primarily on how accurate people are when judging each other's personalities. Another important topic was the assessment of individual differences in accuracy (see Bruner & Tagiuri, 1954). It is not hard to see why social psychologists would find these to be interesting and important issues. Personality psychologists and clinicians, then as now, often relied on trait ratings to describe people and predict their behavior. Thus, a compelling question for the field was how good the judges are who provide these trait ratings. The hypothesis that skilled judges of personality would also be effective clinicians (Schneider, Hastorf, & Ellsworth, 1979) provided a related reason to determine what kinds of people make accurate trait ratings and in what circumstances they do so. Unfortunately, sophisticated methodological critiques of such research (e.g., Cronbach, 1955, 1958) had a paralyzing effect on researchers, and the study of person perception became mired in debates about the most appropriate way to conceptualize and measure accuracy. Reinvigoration of the area was fueled by the increasing attention paid to Asch's (1946) work. With his studies of impression formation, Asch essentially executed an ‘‘end-run” around the accuracy debate by focusing exclusively on the process of person perception. In the Gestalt tradition, Asch was interested in how people construct simple unified impressions of others based on knowledge about their distinct characteristics. His method was to present participants with lists of traits said to describe different people. He then asked participants to report their impressions of the people by writing brief descriptions of them and also by selecting from another list of traits the ones they thought best characterized them. Asch found that a simple additive model did not seem to account for his results: In other words, participants’ impressions did not seem to be a simple sum of their evaluations of the individual items on the list. For example, Asch found a strong primacy effect. Traits appearing earlier in a list were more likely to drive final impressions than those appearing later. In addition, some traits were more central than others; that is, certain traits (e.g., warm) consistently influenced and even polarized impressions more than others (e.g., polite). The questions Asch raised about how people integrate social information were addressed by many other researchers in the following years (see especially work by Anderson, 1974). In addition, his concern with how people deal with and resolve inconsistencies in their knowledge about themselves and others remains an important issue to this day, especially in the person memory literature. But as discussed by Gilbert (1998), Asch's findings and the specific variables he focused on were ultimately less important than his methods, especially because of their implications for the accuracy problem. In the context of a study like Asch's, in which people formed impressions of hypothetical people on the basis of lists of personality traits, the accuracy issue was meaningless. Instead, the interesting questions included what pieces of information were more influential than others and how manipulating the way in which information was received affected perceivers. This general approach was formally endorsed at a now legendary 1957 conference at Harvard University (Tagiuri & Petrullo, 1958). A group of the field's most influential thinkers declared that, indeed, social psychologists should focus on how , not on how well, judgments are made about people. Asch, of course, had already modeled how they might go about doing that. The work of Fritz Heider (an attendee at the conference), which culminated with his classic The Psychology of Interpersonal Relations (1958), was equally influential. Heider's theorizing about the phenomenology of the average

< previous page

page_193

next page >

< previous page

page_194

next page >

Page 194 person and “naive psychology” inspired other researchers to think more systematically about the processes people use to comprehend, explain, and predict behavior. It should be noted that although the preceding account has become widely accepted, other developments in psychology in the 1940s and 1950s seem to have also played a role in leading researchers to focus on the process of knowing other people. Tagiuri (1969) emphasized Lewin's work on leadership (Lewin, Lippitt, & White, 1939), which involved the question of how people perceive leaders; he also pointed to the work by Bruner and others on the motivational factors in perception (e.g., Postman, Bruner, & McGinnies, 1948), factors that were seen to be particularly important when the things being perceived were people. Finally, he noted that because the Thematic Apperception Test (TAT) was in such heavy use by clinicians, the belief that we can learn about people based on their interpretation of human figures portrayed on TAT cards also led to an interest in that process of interpretation. These historical events provided enabling conditions for the development of research on the processes involved in interpreting and storing information about behavior, but they cannot explain why such research has continued to be carried out for so long in so many laboratories by so many people. Jones (1985) mentioned one cynical explanation: People were maybe attracted to Asch's work not only because it pointed the way out of an intellectual quagmire, but also because it was easy to replicate and extend. At the every least, Asch's procedures, along with the many paperand-pencil tasks that are descendants of those procedures (and the paper, pencil, and slide projector and paper, pencil, and computer tasks that are even more common as of this writing) are much easier to execute than the ones used in many other kinds of social-psychological investigations (see Aronson, Ellsworth, Carlsmith, & Gonzales, 1990). Yet it is doubtful that simple convenience could sustain the study of person perception and person memory for so long. Even if it could, this is obviously not how researchers in these areas explain to other people and themselves why they do what they do. An Organizing Framework for Social Psychology? Jones (1985) presented a cynical perspective on the history of person perception and person memory, but it is only fair to also give him credit for a more noble one. He also argued that “Person perception research was almost a syllogistic conclusion to the subjectivist premise: If people respond to the perceived environment and if other people are some of the most important entities in that environment, then it is important to study how people are perceived” (p. 87). In other words, Jones was saying that, given the field's most basic assumptions, the study of person perception is a necessary and primary aspect of the whole enterprise of social psychology. In the very first Handbook chapter on person perception, Bruner and Tagiuri (1954) made a similar point. They reminded their readers that “The first step in reacting to another is forming an impression of him. Later reactions depend on this first step. If there is to be a science of interpersonal behavior, it will rest upon a cornerstone of social perception” ( p. 650). This, then, is another way of understanding why so many social psychologists have devoted themselves to fine-grained analyses of how people make sense of each other's behavior: not because Asch showed them how, not because Heider told them to, and not because the research was easy to do, but because they were trying to create a cornerstone for the science of interpersonal behavior. Person memory researchers, from the time their specialty emerged as a distinct area of study (i.e., from the publication of the seminal 1980 volume edited by Hastie et al.), also seem to have claimed the same agenda. For example, Ebbesen (1980) pointed out: Most forms of social influence begin with or include a rather common event: one person observing the stream of another person's behavior. Conformity, imitation, group pressure, persuasion by communication, impression formation, and so on, generally involve one or more people (the observers) being exposed to the ongoing behavior of other people (the actors). Investigations of these social influence situations have focused primarily on ‘context’ variables that moderate the impact of an actor's behavior on an observer. Only peripheral attention has been given to the processes that determine: (1) how the observer extracts information from the ongoing behavior of the actor; (2) how that information is represented in the observer's memory; and (3) how the information that is in memory affects the behavior of the observer. (p. 179) Ostrom, Lingle, Pryor, and Geva (1980) also claimed that “important social behaviors, such as conformity, aggression, and attitude change, may ultimately be found to be dependent on impression organization” (pp. 85-86). And Hastie and Carlston (1980) pointed out that, ‘‘For many of us, the goal of developing models for person memory grew out of a desire to extend accounts of the formation, representation, and retrieval of first impressions of other people”(p. 1). Indeed, in the preface to the 1980 book, the editors (Hastie et al.) wrote that, “we had concluded that an adequate theory of social memory was the necessary foundation for solutions to

< previous page

page_194

next page >

< previous page

page_195

next page >

Page 195 many of the questions concerning social perception and judgment” (p. vii) that they had been interested in previously. Overall, what these scholars are saying is that if we conceptualize social-psychological phenomena as interactions in which people with social knowledge respond to the social situations they are in on the basis of how they make sense of their own or others’ behavior, social psychologists across the board could perhaps ultimately ground their investigations on the findings and principles yielded by person perception and person memory research. One could even argue that person perception and person memory researchers have always taken comfort in the idea that, although their investigations tend to trigger less immediate excitement and fascination than do the field's classic high-impact or “action” studies, their research was in fact foundational to the whole field of social psychology. Therefore, person perception and person memory have never been just other topics of study for social psychologists (cf. Higgins & Kruglanski, 1996). The long-term goal of researchers in these areas has been to develop models that could serve as common frameworks for the study of those issues that fall under the heading of “Interpersonal Phenomena” in the Handbook of Social Psychology (Gilbert et al., 1998) – classic topics such as altruism, aggression, interpersonal attraction, stereotyping, and social influence. What is exciting about person perception and person memory research is their promise to jointly occupy the same privileged and primary position in the study of interpersonal behavior that the general social cognition approach arguably holds in the study of social psychology defined more broadly. THE PRESENT Accomplishments The project undertaken by person perception and person memory researchers was thus a very ambitious one. And while it has certainly not been completed, much has been learned about how people interpret and represent social information. For example, among the contributors to this volume are researchers who have been intimately associated with the work in person perception that has taught us about, among other things, the stages of processing that occur when we form first impressions (Gilbert, Pelham, & Krull, 1988; Krull, chap. 13, this volume; Trope & Liberman, 1993); the effects of accessible social information on person perception (Higgins & Brendl, 1995), and when assimilation to that information occurs and when contrast effects are found (Moskowitz & Roman, 1992; Newman & Uleman, 1990; Stapel & Koomen, chap. 14, this volume; Wegener, Dunn, & Tokusato, chap. 17, this volume); the causal reasoning involved when we do the careful analysis of behavior that we often undertake to make sense of others’ behavior (Morris, Ames, & Knowles, chap. 9, this volume; Roese, chap 19, this volume); how goals direct and constrain social perception (Vorauer & Ross, 1993; Vorauer, chap. 16, this volume); and important individual differences in all of these processes (Moskowitz, 1993; Thompson, Naccarato, Parker, & Moskowitz, chap. 2, this volume). As other contributors to this volume cover much of this ground, the focus here will instead be on representative developments in the study of person memory (see Wyer & Carlston, 1994, and Carlston & Smith, 1996, for more comprehensive reviews of this literature). The Resolution of Inconsistency. Much research has built on the early work of Hastie and Kumar (1979) and Srull (1981). Bargh and Thein (1985), for example, used both contextual and dispositional moderating variables to provide converging evidence for the hypothesis that expectancy-incongruent information is recalled more readily than congruent information because it receives extensive processing when initially encountered. As in the earlier studies, participants were presented with trait expectancies and lists of behaviors that varied in terms of their consistency with the expectations. Bargh and Thein found that when the behaviors were presented at a very rapid pace (1.5 seconds each), the recall advantage for inconsistent behaviors disappeared because participants in this condition were denied the opportunity to think about how they compared to the target person's other behaviors. In contrast, participants in a condition where presentation of the information was self-paced were able to do the kind of elaborative processing that increases the number of associative links between incongruent and other behaviors. They spent more time reading the incongruent behaviors, and, as a consequence, those behaviors were more easily recalled. Interestingly, for a subset of participants in the rapidly paced attentional overload condition, the pattern of recall was equivalent to that for participants in the self-paced condition. These were people for whom the trait in question – honesty – was chronically accessible. Bargh and Thein had identified in advance participants who regularly thought about and described people in terms of that trait. These are people who might be expected to be well practiced in processing information relevant to honesty (Higgins, 1996). Indeed, these participants were relatively unfazed by the overload manipulation: Like participants in the self-paced condition, they recalled more incongruent than congruent information. In summary, when either tempo-

< previous page

page_195

next page >

< previous page

page_196

next page >

Page 196 rary situational factors or more enduring dispositional factors freed up cognitive capacity for participants, and only then, their recall performance matched that found by Hastie and Kumar (1979). Finally, Bargh and Thein (1985) highlighted the finding (previously reported by other researchers as well) that people not only devote attentional resources to comparing incoming information to their expectancies, but are also actively involved in developing and revising their expectancies. When participants were presented with sets of behaviors in which the majority were inconsistent with the experimenter-provided trait expectancy, they seemed to quickly revise those expectancies. The raw behavioral evidence was more important than the abstract trait in determining what participants treated as congruent and incongruent information. More recently, Sherman and Hamilton (1994) provided novel support for the hypothesis that the kind of extra processing received by expectancy-incongruent behavioral information involves the reactivation of previously encountered information about the person in question. Participants in their studies were led to believe that a target person was either friendly or intelligent. Then they were presented with a list of other behaviors performed by that person. The list contained equal numbers of congruent, incongruent, and irrelevant behaviors. Reading of the behaviors was periodically interrupted so that participants could answer a recognition question; specifically, they were presented with a behavior description and asked whether it was one of those they had already seen. Recognition latencies were fastest when the question followed an incongruent behavior. This finding is consistent with the hypothesis that incongruent behavior cues people to bring into mind previously encountered information about the target person. This is less likely to be the case when congruent information is encountered, and that would account for why recognition latencies were longer when the recognition questions followed behavior descriptions of that kind. The results of studies focusing on trait moderators of behavior recall are also consistent with these findings. For example, people high in Need for Cognition (Srull, Lichtenstein, & Rothbart, 1985) and Uncertainty Orientation (Driscoll, Hamilton, & Sorrentino, 1991) would be expected to process information more carefully than other people; indeed, they also recall higher proportions of expectancy incongruent information than do other people. Overall, Hastie and Kumar's (1979) and Srull's (1981) signature finding has been well replicated, and their account of that finding has been well supported. As noted previously, both the findings and the account were central to the development of the earliest person memory models. How Social Information Is Represented . Basic person memory models have undergone quite a bit of development in the last 20 years. A detailed discussion of those models is beyond the scope of this chapter. However, as just one illustration of their complexity, one of the most influential of them, Srull and Wyer's (1989; Wyer & Srull, 1989) makes the assumption that any one observed behavior can be coded in more than one way at the location in memory reserved for the person engaging in that behavior. A given behavior can be part of a trait-behavior cluster, which consists of a trait concept and the behaviors exemplifying that concept, and also part of an evaluative person representation, a subset of behaviors clustered together as a separate cognitive entity that allows one to form a general conception of how likable a person is. (For more on how knowledge about other people might be stored and represented see Chen, chap. 8, this volume; Sherman, chap. 11, this volume). Information-Processing Goals. Both person perception (Jones & Thibaut, 1958) and person memory researchers (Hamilton, Katz, & Leirer, 1980a, 1980b) recognized early on that the goals or objectives with which people approach their social interactions will affect how they interpret and store information about other people. For example, Hamilton et al. (1980a) focused on the special memorial consequences of processing social information with an impression formation goal (perhaps the most common goal of participants in person perception and person memory experiments). They presented participants in their studies with a series of sentence predicates describing behaviors (e.g., ‘‘Cleaned up the house before company came”). In addition, they asked half of the participants to form an overall impression of the person who performed the actions and the other half to simply memorize the information. Overall memory for the behaviors was actually higher for people in the impression-formation condition. Hamilton et al. explained these results by arguing that an impression-formation set, more than a memory set, encourages people to think about various behaviors in relationship to each other and organize them in a coherent way. This organization, in turn, facilitates recall. In support of this explanation, Hamilton et al. (1980b) showed that the order in which impression-goal participants recalled behaviors indicated that they were clustering the behaviors in memory in terms of trait categories (intelligence, sociability, etc.). Information-processing goals also have implications for the relationship between judgments about people and recall of the pieces of evidence (behaviors,

< previous page

page_196

next page >

< previous page

page_197

next page >

Page 197 traits, etc.) on which those judgments are assumed to be based. As Hastie and Park (1986) pointed out, it is usually assumed that “There ought to be a relationship between memory and judgment. Our intuition tells us that … evaluations of people should be related to the amounts of good and bad information we have about them” (p. 258). However, Hastie and Park's review of the evidence indicated that this commonsense assumption quite often turns out to be wrong. The key, they suggested, was whether the judgments in question are on-line or memory-based. When people construct judgments and evaluations while behavioral information is being received, they can later retrieve those judgments and evaluations directly. In other words, when judgments are made on-line, subsequent reports of those judgments are a reflection of how people reacted to and assessed the information when it was received, and not a function of whatever pieces of information happen to be accessible at the time of the report. For example, in the typical person memory experiment, although people are better able to recall behaviors that were evaluatively incongruent from the majority of the behaviors (e.g., the 3 dishonest behaviors in a set of 12 mostly honest ones), their evaluations of the target person are typically driven by the valence of the majority behaviors (positive, in this case). Incongruent behavior is more carefully processed, but much of that processing is devoted to attempts to explain it away or discount it (Crocker, Hannah, & Weber, 1983; Hastie, 1984). In contrast, when judgments are not made on-line, people have to rely on the evidence they retrieve from memory to reach reasonable conclusions. Thus, judgments are much more likely to directly correspond to what is remembered. Lichtenstein and Srull (1987) extended this research by showing that, relative to impression sets, the goal of simply understanding or comprehending information is also more likely to lead to memory-based judgments. Spontaneous Trait Inference. The work reviewed in the previous section would seem to imply that people generally do no more than superficially process social information unless they have the explicit goal of trying to figure people out. In fact, Wyer and Srull's (1989) model explicitly predicts that people will not interpret behavior in terms of abstract trait concepts unless some processing objective requires it. A research program initiated by Winter and Uleman (1984), however, challenged that idea. Winter and Uleman presented participants with a series of written descriptions of trait-implying behaviors (e.g., “The secretary solves the mystery halfway through the book”; “The tailor carries the old woman's groceries across the street”) each performed by a different person. Participants were asked to simply memorize this material; after a brief distractor task, they were presented with a list of cues to help them recall the behaviors. Word associates to both the sentence actors (e.g., typewriter, clothes) and their actions ( reading, assisting) helped boost recall relative to a no-cue baseline condition. The most effective cues of all, however, were personality traits ( clever, helpful ). Winter and Uleman concluded that participants had inferred those traits when the behaviors were initially encountered. Therefore, the trait concepts served as effective cues because they were implicitly present when participants encoded the behaviors (see Tulving & Thomson's [1973] encoding specificity principle). At the same time, few participants reported any awareness of having thought about the behaviors in terms of trait implications. In summary, participants seemed to have spontaneously inferred traits from behavior – that is, they inferred traits without intending to, without being aware of doing so, and with the investment of minimal cognitive resources. Winter, Uleman, and Cunniff (1985) developed an experimental procedure that made it even less likely that participants would consciously adopt the goal of inferring traits. In their study, behavior descriptions like those used by Winter and Uleman (1984) were presented as distractors in the context of a number-recall experiment. Participants had to read each sentence aloud while simultaneously rehearsing different sets of five numbers that they had been asked to memorize. Winter et al. also varied the difficulty of the number-recall task; some participants had to memorize single-digit numbers, whereas others had to memorize more complex strings (e.g., 14, 123, 44, 62, 105). Winter and Uleman's (1984) cued-recall results were replicated. In addition, the manipulation of cognitive load had no effect on spontaneous inference, leading Winter et al. to tentatively suggest that the trait inference process is an automatic one that is impervious to disruption and impossible to control. However, subsequent research has shown that spontaneous trait inference can be eliminated by even higher levels of cognitive load than were used by Winter et al. (Uleman, Newman, & Winter, 1992) or with the imposition of certain processing goals (Uleman & Moskowitz, 1994). But as Bargh (1989, 1994) argued, the automatic-controlled distinction is actually a continuum. Thus, although trait inference can be considered to be only a relatively automatic process, most cognitive processes of interest to social psychologists are complex enough that they consist of both automatic and controlled components. Alternative explanations for the results of studies using the basic cued-recall procedure developed by

< previous page

page_197

next page >

< previous page

page_198

next page >

Page 198 Winter and Uleman (1984) have been proposed by some commentators (e.g., Wyer & Srull, 1989, p. 146). Nevertheless, empirical support for the spontaneity of trait inference has now been found using a variety of procedures utilizing many different dependent variables (for a review, see Uleman et al., 1996). Overall, the evidence indicates that, although impression goals promote rich and elaborate trait interpretations and the complex organization of information in terms of trait categories, such goals are not necessary for trait inferences to occur. Self-Flagellation Clearly, much progress has been made in explicating person perception and person memory processes. Nonetheless, a reader of recent literature reviews will often detect a tone very different from pride. Instead, a certain amount of defensiveness and a somewhat self-critical quality – perhaps even what Jones (1985, p. 99) described as selfflagellation – are quite common. The writers of these reviews all had their own perspective on why these problems are maybe exaggerated, irrelevant, or actually signs of strength. But even among its most accomplished practitioners, one can detect a concern that work on person perception and person memory is becoming increasingly method-bound, self-referential, about nothing but itself, divorced from the real world, and not so clearly related to social behavior, which is presumably what social psychologists are most interested in. Such qualities would seem to be liabilities for any program of research seeking to position itself at the center of the study of interpersonal behavior. Wyer and Carlston (1994, pp. 42-43; see also Wyer, Budesheim, Lambert, & Gruenfeld, 1992), for example, worried that The field has tended to focus on a restricted set of issues, processes, and paradigms, leaving many important aspects of social representation unexplored…Relatively little research has been conducted on the representations that are formed from information acquired in actual social interactions…The manner in which these social representations contribute to actual human behavior and interaction is even less clear…The conditions investigated often seem only remotely related to those in which people receive information in daily life. Earlier, Schneider (1991) discussed the complaints of critics who argued that “social cognition today is often driven more by model testing than by the traditional social psychological efforts to explain intuitively interesting and immediately important social phenomena …(it) can seem sterile and bloodless as much of the richness of the everyday social world gets processed and abstracted out …” (p. 530). And Smith (1994, p. 100) worried about his colleagues’ impression that ‘‘Endless investigations look at details of memory structure or question how many processes can dance on the head of a pin”. Such self-critiques are not new, however, and certainly do not signal a crisis in the field. Hastie, Park, and Weber, in their comprehensive review of the social memory literature in 1983, worried that theoretical advances in the study of how behavior is perceived, stored, and remembered had “occurred at the expense of developments in our knowledge of the structure of the real social world and its phenomena outside the laboratory” (p. 154). Fifteen years before that, Tagiuri (1969), in his Handbook of Social Psychology chapter on person perception, noted that “The technology is by now very complex and tends, at times, to obscure the main issues…How does the perception of the other relate to action toward him? This is…a complex problem. But was not the desire to understand this interplay one of the sources of the impetus for much work in person perception?” (pp. 432, 435). These issues were already on his mind in 1958, as noted in the introduction to the published volume that emerged from the influential Harvard conference mentioned earlier (Tagiuri, 1958), and one can find similar concerns about the relevance of person perception for actual interpersonal behavior in the very first Handbook chapter on the topic (Bruner & Tagiuri, 1954). In summary, person perception and person memory researchers have always engaged in some degree of selfflagellation. A USEFUL LEVEL OF ANALYSIS As discussed, social-psychological phenomena can be conceived of as interactions in which people respond to the situations they are in on the basis of how they make sense of their own or others’ behavior. Therefore, one could argue that the primary task of social psychology should be to understand how people make sense of their own or others’ behavior. It would follow that, despite their many bouts of self-criticism, person perception and person memory researchers could also take pride in having focused on a level of analysis that could be an organizing framework for much of the field. Arguments like that can be misleading, however. For example, Buss (1995) wrote that, “Because all behavior depends on complex psychological mechanisms, and all psychological mechanisms, at some basic level of description, are the result of evolution by selection, then all psychological theories are implicitly

< previous page

page_198

next page >

< previous page

page_199

next page >

Page 199 evolutionary psychological theories”(p. 2). As Harris and Christenfeld (1995) pointed out, one could replace the references to evolution in that sentence with references to neurotransmitters or subatomic particles and it would make just as much sense. Their point, of course, was that social psychologists can adopt different levels of analysis (e.g., evolutionary, biological, cultural, and cognitive), and all are potentially useful for understanding social behavior (see also Schaller, 1997). Any social-psychological research enterprise will have its limitations, but the most important issue is whether people working in that area develop a level of analysis that other psychologists find to be a useful and important one for describing, explaining, and predicting social behavior. Therefore, the most important question for students of person perception and person memory is whether their contributions and insights have made an impact on social psychologists studying the field's classic topics. Have they focused on a level of analysis that other social psychologists have found to be useful for shedding light on those aspects of social behavior of interest to them? The answer is a resounding “yes.” The following brief review of evidence supporting that conclusion is by no means intended to be a comprehensive review of all the research informed by models of person perception and person memory, nor does it identify all of the content areas that have benefited from those models. Instead, it is a selective set of examples illustrating the kind of contribution the study of person perception and person memory has made toward understanding a variety of interesting and important aspects of interpersonal behavior. Aggression Aggressive behavior is one of social psychology's classic topics. Quite a few contextual variables playing a role in the inhibition or facilitation of aggression have been studied, such as crowding, heat, frustrating circumstances, and one's relative anonymity (Geen, 1998; Krebs & Miller, 1985). But a number of researchers (e.g., Kulik & Brown, 1979) have focused on our interpretations of others’ behaviors as proximal causes of aggression. Dodge and his colleagues (see Dodge & Crick, 1990) have found, among other things, that aggressive individuals are often predisposed to decide that other people's unpleasant behavior is intentional and specifically directed toward them. In developing this idea and designing research to test it, they have been guided by basic models of person perception and attribution, such as Jones and Davis’ (1965) explication of the processes involved in inferring intentions from behavior. Recently, Zelli and his colleagues (Zelli, Huesmann, & Cervone, 1995; Zelli, Cervone, & Huesmann, 1996) found that a more reliable predictor of individual differences in aggressiveness may be the extent to which people make spontaneous hostility inferences. Their research suggests that one can do an even better job of identifying aggressive people by looking at the kinds of inferences they make quickly, unintentionally, and without awareness than by looking at the kinds of judgments they make when they have a long time to deliberate about them. This research, of course, was made possible by the more basic work on spontaneous personality trait inference (Uleman et al., 1996; see also Tillema, 1998, for similar research on individual differences in loneliness). Altruism and Prosocial Behavior Similarly, the decision to engage in prosocial behavior and help other people can be analyzed in terms of social context factors, but altruism also depends in part on how we interpret the behavior of people in need and on the kind of people we judge them to be. Therefore, researchers interested in helping and resource allocation have also found the level of analysis occupied by person perception research to be a useful one. For example, Skitka and Tetlock (1992; see also Weiner, Perry, & Magnusson, 1988) found that when resources were scarce, participants in their studies (especially those with a politically conservative ideology) were less likely to assist claimants (by allocating medication, housing, and bodily organs to them) when they believed that the claimants had behaved in such a way as to cause their own plights. Skitka and Tetlock were guided by a more general attributional model developed by Weiner (1986, 1993). Using the same model, Zucker and Weiner (1993) found that particular attributions for poverty (e.g., laziness) are reliably associated with a decreased willingness to offer help. A consideration of person perception processes was also necessary to interpret the results of social psychology's most well-known studies of helping behavior – Latané and Darley's (1970) research on bystander intervention. Latané and Darley found that a major determinant of people's willingness to take action in emergency situations is their interpretation of those situations as ones that call for intervention. One of the most important cues used to disambiguate situations is the behavior of other people who are present. Unfortunately, Latané and Darley pointed out, “If each member of a group is trying to appear calm and not overreact to the possibility of danger, each other member, in looking

< previous page

page_199

next page >

< previous page

page_200

next page >

Page 200 to him for guidance, may be misled into thinking that he is not concerned” (1970, p. 41). In other words, people looking to others for information about how to avoid acting inappropriately often fail to consider that others are doing the same thing. They also might not appreciate how their own muted reaction is actually eliciting the same behavior in the people around them. When a person looks to others to define an ambiguous event, he or she often fails to adjust for the influence of both social norms and his or her own behavior on those other people. This can lead to what appears to be apathy and callousness. The extent to which people take into account situational pressures and constraints when interpreting each other's behavior has, of course, been a major focus of research and theorizing in person perception (Jones, 1990; Kelley, 1973). Person perception is thus intimately involved in the phenomenon of pluralistic ignorance, where group members assume that their own thoughts and feelings differ from others in the group (D. T. Miller & Prentice, 1994, 1996). Stereotyping A common theme in the stereotyping literature is that stereotypes often indirectly influence evaluations of other people and interactions with them by affecting how their behavior is interpreted (e.g., Darley & Gross, 1983; Duncan, 1976; Sagar & Schofield, 1980; see Hamilton, Stroessner, & Driscoll, 1994, for a review). Some researchers have gone beyond documenting this general phenomenon and have attempted to conceptualize the effects of stereotypes in terms of different stages of information processing. For example, Kunda and Sherman-Williams (1993) argued that, regardless of whether people are affected by stereotypes when forming impressions, they still engage in the same basic person perception and person memory processes. Stereotypes, they proposed, can just bias the input to those processes – what Trope (1986) and others have called the initial identification (or categorization) of behavior. Kunda and Sherman-Williams localized the biasing effects of stereotypes at a stage of social information processing described by researchers developing more general models of person perception. This approach also allowed them to use the more general model to speculate about the broader consequences of the particular biasing effects they had identified. Delmas (1992) presented evidence that stereotypes can actually have such effects spontaneously: Therefore, one might preferentially construe a shove by an outgroup member as an instance of hostility rather than playfulness without even being aware that any inference has been made. This occurs even when out-group membership is based on a trivial distinction (e.g., with a minimal-group manipulation; Otten & Moskowitz, 2000). Less work has been done to explicate in a precise way how social category information is incorporated into the more general person representations described by person memory researchers (cf. Gingrich, Houston, & Lupfer, 1998; Wyer & Martin, 1986). Stangor and Ruble (1989), however, adopted some of the experimental procedures developed by those researchers and used them to examine representations of groups and the processes involved in stereotype formation and maintenance. Participants in Stangor and Ruble's (1989) study were presented with behavioral information about two fictitious fraternities. The mix of information encouraged them to develop stereotypes about the groups’ relative extraversion. Then they were asked to recall the behaviors. Consistent with the results of most person memory studies, no recall advantage for expectancy consistent (i.e., stereotype consistent) information was initially found (see Hastie & Kumar, 1979; Srull, 1981). However, when more information was presented after the stereotypes had become well developed, participants subsequently recalled significantly more stereotype-consistent than stereotype-inconsistent information. This led Stangor and Ruble to propose that different processes are involved in stereotype formation (e.g., attributional processing) and stereotype maintenance (e.g., selective attention). For stereotypes to affect the processing of information about a person, that person must first be categorized in terms of some social category. Influential models developed by Brewer (1988) and S. T. Fiske and Neuberg (1990) suggest that an understanding of the circumstances in which people are viewed and treated simply as prototypical members of a group (i.e., the circumstances in which stereotypes will guide our interactions with others) is inseparable from an understanding of the circumstances in which attribute-by-attribute (or behavior-by-behavior) analyses of information about other people take place (i.e., the circumstances in which the kind of processing described in the person perception and person memory literatures is more important). According to S. T. Fiske and Neuberg (1990), category-based and piecemeal processing define two ends of a continuum of processes (although the former has priority over the latter). By contrast, Brewer (1988) described a branching sequence of stages in which category-based processing is the default and personalized processing only occurs when the observer has some degree of affective investment in the target (see S. T. Fiske, 1988, for a detailed comparison of the models). Brewer's model

< previous page

page_200

next page >

< previous page

page_201

next page >

Page 201 also suggests that the kinds of detailed person representations described by the person memory literature are only formed as a result of personalized processing. Otherwise a person is simply represented in memory as a categoryexemplar (e.g., as just one of many Asians, elderly people, or nurses). In both models, however, the assumption is that, to understand the role of stereotyping in interpersonal behavior, one must take into account the alternative to stereotyping and the circumstances in which one or the other kind of processing is more likely to occur. And learning about the alternative boils down to learning what the person perception and person memory literatures have to offer. Conformity Conformity and obedience have long been among social psychology's most compelling phenomena (Asch, 1956; Milgram, 1974). D. T. Miller and Prentice (1996) analyzed such phenomena in terms of how people construct group norms. For example, to conform, one must first identify a norm specifying the expected behavior. In some cases (e.g., “How do other people really feel about drinking or premarital sex?”) the norm is ambiguous. To figure out what it is, one needs to interpret other people's behavior and determine why they are doing what they are doing and how they feel about it. The norms that people confidently identify as a result of such an analysis are often distorted. Much can go wrong in the construction of norms, including all of the things that can go wrong in and bias interpersonal perception. In fact, Miller and Prentice's discussion of the processes that lead people to conform behaviorally to a norm – be it a real norm or an illusory one – leads them to literatures like the ones on the correspondence bias (Gilbert & Malone, 1995) and spontaneous trait inference (Uleman et al., 1996). Minority Influence Of course, people do not always bow to pressures to conform; sometimes lone dissenters prevail on the majority to accept their views. To understand the minority influence phenomenon, one could focus on how the style with which dissenters buck the majority affects their impact on the group (e.g., Moscovici, Lage, & Naffrechoux, 1969). One could also go one step further, however, and look at how group members perceive and form impressions of dissenters. That is the approach that Moskowitz (1996; see also Moskowitz & Chaiken, 2000) took, and he found that attributions for dissenter behavior play a significant and underexamined role in the minority influence phenomenon. Interpersonal Attraction Evaluations of people (or liking judgments) have long been common as dependent variables in person perception research,2 and recent models of person memory (e.g., Srull & Wyer, 1989) assume that one's overall evaluation of a person plays a prominent role in one's cognitive representation of that person. Berscheid (1985), however, noted that person perception and attribution models have not played a direct role in framing the kinds of questions asked by students of interpersonal attraction, “although the processes by which attributions are made of another's dispositional properties are understood, of course, to be critical to an understanding of attraction’’ (p. 426). The importance of the processes by which people interpret and recall others’ behavior has thus been implicitly assumed: For example, the well-documented similarity-attraction relationship (Byrne, 1971) depends on the ability to determine that another person has particular attitudes, beliefs, and personality traits. THE FUTURE Red Herrings In summary, person perception and person memory researchers have provided other social psychologists with important conceptual tools and a useful level of analysis. However, one could also argue that what has been learned has not yet been exploited to the fullest, and students of person perception and person memory still have riches to share with other social psychologists. Unfortunately, those riches may stay locked up in the vault unless some challenges are embraced. Before outlining those challenges, it should be acknowledged that much of the landmark research that in the past ended up inspiring the field's conceptual development did so accidentally or serendipitously because the research yielded unexpected and surprising findings. For example, as noted before, the development of person memory models was to a great extent oriented around Hastie and Kumar's (1979) finding that, in some circumstances, people preferentially recall expectancy-inconsistent information – a finding that, according to the writeup of the study, was not expected. Similarly, a great many grant dollars have been spent trying to understand why participants in Jones and Harris’ (1967) study seemed to think that one can get a pretty good read on a speaker's political orientation by observing what he or

< previous page

page_201

next page >

< previous page

page_202

next page >

Page 202 she is forced to say and do. This was a finding that did not directly follow from Jones and Davis’ (1965) correspondent inference theory. Nonetheless, some modest suggestions are offered. These suggestions are not the ones that might seem to follow most naturally from some f the standard self-critiques of the study of person perception and person memory. Therefore, it is necessary to first summarize what do not seem to be promising strategies for further invigoration of this area of research while acknowledging that these comments to a great extent echo those of others (e.g., Aronson et al., 1990; Mook, 1983; Ostrom, 1984; Wyer, Budesheim, Lambert, & Gruenfeld, 1992). Worries about the fact that the vast majority of person perception and person memory studies are laboratory-based seem to suggest that it is incumbent on investigators to design more field studies. Unfortunately, this recommendation can be a vacuous one. If the issue is that laboratory studies have ignored, or held constant, or even overemphasized certain variables, then the task of the researcher is to specify what those variables are so that they can be operationalized and controlled for in some way. If that can be done in the field, then a field study would be appropriate. If that can be done in the laboratory, then a laboratory study would be appropriate. It is not clear what there is to be gained by simply translating laboratory studies into field investigations. An unexpected finding or a failure to replicate an effect originally uncovered in the laboratory would simply require one to construct an ingenious post hoc explanation. And if one did replicate an effect in the field, that would be satisfying, but even then it could well be that the field setting that was picked was as exotic or unrepresentative as the researcher's lab room. A related suggestion might be that the next step for person perception and person memory researchers is to determine whether their models hold up in the real world. The problem, of course, is the assumption that there is some obvious consensual definition of the real world . In this context, it apparently means “not the laboratory,” but that, of course, is about as coherent a category as “not Princeton University” or “not the airport.” In addition, the isolation of the laboratory environment from the real world should not be exaggerated. One can assume that interactions between research participants and between experimenters and their research participants, while not necessarily representative of all interactions, follow most of the normal rules that govern interpersonal behavior in other contexts. Experimenters who ignore those rules probably do so at their peril. Finally, concerns about the apparent artificiality of the tasks presented to participants in laboratory-based person perception and person memory research might be taken for a call for the use of more naturalistic materials. One could hardly object to naturalistic materials in principle. By itself, though, this recommendation is no more compelling than a call for more field studies, and for many of the same reasons. There may be something to be gained from seeing whether people will spontaneously infer generosity when presented with a film clip showing that “The plumber slips an extra $50 into his wife's purse” (Winter & Uleman, 1984) as opposed to a sentence describing that behavior. What is to be gained, however, needs to be explicitly spelled out. Needless to say, no one would claim that there are no differences between the processes involved in decoding text and those involved in extracting meaning from live social targets. For some research questions, compelling arguments have been made for the importance of those differences. For example, Gilbert and Hixon (1991) pointed out that one cannot reach conclusions about the ubiquity of stereotyping in everyday social interaction by manipulating social categories verbally (e.g., by identifying a person in writing as a “Black fireman’’) because the verbal labels force on people distinctions that they might not have made by themselves. In general, without some compelling framework for thinking about the theoretically relevant differences between naturalistic materials and those more typically utilized for well-controlled studies of person perception and person memory, it is not clear what the payoff will be for the researcher or the field. New Contexts What person perception and person memory researchers study are particular phenomena that take place in particular contexts involving particular people. Happily for them, there are still interesting and unexplored interpersonal phenomena crying out for an analysis in terms of person perception and person memory processes; populations of people clamoring to teach us more about those processes; and contexts just waiting for person perception and person memory researchers to arrive so that they can reveal to those researchers new dimensions of their theories and models. A few years ago, Wyer and his colleagues departed from their standard experimental procedures and imported person memory principles into a conversational context (Wyer, Budesheim, & Lambert, 1990; Wyer, Budesheim, Lambert, & Swan, 1994; see also Wyer et al., 1992). In these studies, the researchers did what person memory researchers almost always do; that is, they presented participants with trait descriptions of

< previous page

page_202

next page >

< previous page

page_203

next page >

Page 203 others and let the participants learn about a number of those people's behaviors. In this case, however, rather than just presenting the information in a disembodied way with a slide projector or computer, they embedded it in a conversation between two people. A superficial reading of the results of these studies would probably horrify the student of person memory processes. That is because, at first glance, almost none of the classic person memory findings were replicated, and the conclusion one might reach is that all one needs to do is tweak the classic research methodology a little bit for the whole study to blow up in one's face. In a darker frame of mind, one might even conclude that these studies suggest that past person memory studies are about nothing but themselves, and that the results of those studies tell us nothing more than what happens in person memory studies. A less superficial reading of these papers would lead to different conclusions. A full description of the complicated results of these studies is beyond the scope of this chapter. Recall, however, some of the most basic person memory principles: that is, people have and/or develop expectations; the relationship between those expectations and new information determines how they process that information; unexpected information, especially early in the impression formation process, leads to enhanced processing; and that in turn affects the memory trace. Wyer et al.'s studies actually fit neatly into that story. In this case, however, the expectations involved were not just trait impressions (as is usually the case in person memory studies), but expectations based on conversational norms. For example, we expect people to be charitable when describing others, and we expect people to be relatively modest when describing themselves. When people violate those expectations, such behavior attracts notice and causes an observer to process that information more deeply. That information, of course, is then more accessible for recall later. These, in fact, are the kinds of findings reported by Wyer et al. For example, participants were better able to recall the favorable statements people made about themselves, but recalled more unfavorable statements when a person was making them about someone else. In summary, these data actually were arguably compatible with the basic Srull and Wyer (1989) person memory model, but they forced that model to stretch to accommodate new and important variables. Note that it would not do justice to these studies to say that they simply represented an effort to see whether a model would replicate in a more naturalistic or real-world setting. As discussed earlier, research framed in that way is unlikely to advance theory. Instead, it was a project in which a model was tested in a context (one in which conversational norms were salient) that differed in a specific and potentially theoretically important way from the usual one in which it was tested (because those norms have important implications for attentional allocation). The result was that the model developed in an arguably interesting way that could generate whole new sets of questions. Hamilton, Driscoll, and Worth (1989) provided a related example of the usefulness of stretching a research paradigm's boundaries. They noted that typical person memory study is set up so that participants receive information related to one possible trait that a person might possess. There may well be social contexts structured in such a way that a person's behavior is diagnostic of one and only one personality characteristic, but it is arguably more typical to encounter people in situations where their behavior is informative about many possible traits – friendly or unfriendly, intelligent or unintelligent, adventurous or unadventurous, and so on. The Hamilton et al. data suggest that, in situations like that, people engage in processes during the encoding of behavior that person memory researchers would not be able to detect by using only the typical single-trait procedure. For example, people might devote effort to verifying that a behavior had been encoded in terms of the right trait concept, and this leads to an increase in recall of behaviors congruent with that trait (in contrast to the typical recall advantage for incongruent behaviors). Again, a small (but meaningful) change – one that essentially represented a different social context – led to an extension of existing person memory models. There are, of course, other social contexts with features that, from the perspective of our models, might vary in interesting ways from the ones usually studied. For example, it is not clear that person memory models have been extended to situations where people have a vested interest in or a preference for the way in which they judge and represent other people. A number of models of biased inference processes hypothesize that people construct justifications for their preferred judgments and inferences by searching memory for supportive information (e.g., Kunda, 1990; see also Baumeister & Newman, 1994). Often those judgments are about people, and, in some investigations, the biased inferences that have been studied are inferences about the dispositional properties of those people. To illustrate, Klein and Kunda (1992) found that when people were motivated to see a target person as being high in ability – because that person was going to serve as their partner in a history quiz game – they did in fact conclude that the target was very knowledgeable, whereas those motivated

< previous page

page_203

next page >

< previous page

page_204

next page >

Page 204 to conclude the opposite – because that person was going to be an opponent – they judged the target to have relatively little knowledge (see also Dunning, Leuenberger, & Sherman, 1995, Study 3). In such investigations, predictions about how people selectively search their memories for information about people are rarely, if ever, informed by models of how we cognitively represent such information. It remains for future research to build bridges between the person memory and biased inference literatures. New Populations Instructors of undergraduate methods courses often give students an opportunity to demonstrate their knowledge of research design by writing up critiques of other people's research. In all probability, most of those instructors (like the author of this chapter) find that a popular theme in those papers is the claim that not much can be concluded from the results of a study because the research participants came from a specific population (e.g., Texans, Ivy Leaguers, students at a business school, shoppers at a particular supermarket, etc.). Hopefully those instructors point out to their students that such critiques are useful only if one can specify (or at least speculate about) why the characteristics of a particular group cast doubt on the generality of a finding. Simply to suggest testing models or replicating effects with different groups of people is not to suggest much. Even Sears (1986), who bemoaned social psychology's reliance on college students as research participants (and described how they might make up a unique subpopulation) warned that selective conversion and replication of social psychology studies makes sense primarily “when there is reason to believe that findings might be biased by our peculiar data base” (p. 527). In other words, selective replication makes the most sense when a group of people not only differs from the population usually studied, but when people in that group differ in terms of processes that are central to a model or phenomenon. Such groups can help provide both tests of a model and systematic ways of thinking about revisions of the model. Studies involving participants from different age groups provide a case in point. A number of writers have pointed out that social psychologists do not do themselves any favors when they ignore developmental data (e.g., Higgins & Wells, 1986; Ruble & Goodnow, 1998). Not only do 5-year-olds, 7-year-olds, and 10-year-olds differ from college students, but they differ in interesting ways that could be helpful for testing or extending theories. For example, the overjustification effect – i.e., the tendency of salient incentives for an activity to undermine intrinsic motivation for that activity (Deci & Ryan, 1985; Lepper & Greene, 1978) – is commonly assumed to be mediated by the use of the multiple sufficient causes schema and the discounting principle (Kelley, 1973; Morris, et al., chap. 9, this volume). This implies, of course, that if one can replicate the overjustification effect with 3-year-olds and 4-year-olds (Lepper, Greene, & Nisbett, 1973), one needs also to demonstrate that children that young have developed a multiple sufficient causes schema (Newman & Ruble, 1992). If not, then one's model of the overjustification effect is in need of revision. Similarly, if one asserts that the bias toward explaining others’ behavior with person rather than situation attributions (the “fundamental attribution error”; Ross, 1977) derives from some more basic perceptual salience effect (i.e., the tendency to focus attention on the figural actor and not the “ground’’ or situation; see Gilbert & Malone, 1995, for a discussion), one must also explain why young children do not reliably display the same bias (Rholes, Newman, & Ruble, 1990). In recent years, cross-cultural investigations of the attribution of causes for behavior have been particularly informative. A number of researchers have taken advantage of opportunities to study populations of people that differ in theoretically interesting ways from the ones typically studied by North American and Western European social psychologists. For example, J. G. Miller (1984) found that her participants from the United States showed the expected tendency to explain others' behavior predominantly in terms of their stable dispositional qualities. This tendency, however, was notable only in its absence among a sample of Hindu participants she recruited from India. Morris and Peng (1994) replicated and extended these findings by comparing North American college students to a group of Chinese students (see also Krull, chap. 13, this volume). Archival studies of the kinds of causal accounts offered for behavior in newspaper reports have also yielded results that are consistent with these cross-cultural differences (Lee, Hallahan, & Herzog, 1996; Morris & Peng, 1994). In general, person perception processes begin to seem unfamiliar in many ways when social psychologists study people from cultures that, according to cross-cultural psychologists (Markus & Kitayama, 1991; Triandis, 1990), are more collectivist and less individualist than those found in North America and Western Europe. People in collectivist cultures (such as traditional Asian ones) are less likely than people in individualist ones to conceive of individual people as free, independent agents who are defined separately from their social contexts. That profound psychological difference

< previous page

page_204

next page >

< previous page

page_205

next page >

Page 205 would seem to have implications for person perception and person memory, and it does. Some of social psychology's most cherished phenomena – e.g., the fundamental attribution error – suddenly become more complicated and less fundamental when we examine the social inferences and causal attributions of people without the individualist assumptions shared by the vast majority of the people who have contributed to our current data base (A. P. Fiske, Kitayama, Markus, & Nisbett, 1998; Krull, chap. 13, this volume). In addition, the findings emerging from this research are forcing researchers to highlight certain distinctions that did not seem in need of sharpening before, such as those involving inferential goals. People might not universally approach social situations with the goal of extracting individuating information about the other people they encounter there (see Jones & Thibaut, 1958; Krull, 1993). The implications of these differences in beliefs about behavior for the representation of social information have been less extensively explored, with one exception being a now sizable literature on cross-cultural differences in descriptions of the self (e.g., Cousins, 1989; Rhee, Uleman, Lee, & Roman, 1995; for a review, see A. P. Fiske et al., 1998). But it seems safe to predict that current models of person memory need to be revised to accommodate the individualism-collectivism distinction. It is not obvious, for example, that trait-behavior clusters play a prominent role in the way people from collectivist cultures cognitively represent their acquaintances (cf. Srull & Wyer, 1989). Not only does cross-cultural research have the potential to change the way we think about phenomena like the fundamental attribution error, correspondence bias, and spontaneous trait inference (Duff & Newman, 1997; Newman, 1993), but it also raises some interesting meta-issues about the field. It suggests another level of analysis for a question posed earlier: Why did the study of person perception and person memory become so prominent in social psychology? Historically the field has been dominated by researchers from the United States (Jones, 1985), and perhaps these researchers focused so much on how people think about, perceive, and store information about each other because their culture encouraged them to see the individual as the proper unit of analysis for social psychology. Had social psychology first flowered in Asia, the history of the field would arguably look quite different. In summary, person perception and person memory research involving currently underrepresented populations can do more than just provide researchers with the satisfaction of learning that their findings are not simply an artifact of a narrow data base (Sears, 1986). Instead it can provide valuable opportunities to test, challenge, and enhance the generality of existing models. Other opportunities await. For example, Penn, Corrigan, Bentall, Racenstein, and Newman (1997) noted that, although “schizophrenia is inherently an interpersonal disorder in which problems result from faulty construction of the social environment and one's place in it” (p. 114), few researchers have actually adopted a social-cognitive level of analysis in their studies of patients with schizophrenia. Penn et al. issued a call for such research, which has the potential to provide unique tests of assumptions about the roles of variables such as cognitive capacity, social schemata, and interaction goals in person perception and person memory (Newman, in press). New Phenomena Social psychologists probably do not need to lengthen their list of effects and phenomena with catchy names. A more worthwhile project for person perception and person memory researchers would be a commitment to confront and analyze the “intuitively interesting and immediately important social phenomena” to which Schneider (1991, p. 530) alluded. In other words, there are quite a few interpersonal phenomena identified on the basis of both everyday observation and clinical experience that have not received the attention that they may deserve. Shedding light on those phenomena could, first of all, highlight the explanatory power and usefulness of current models. Even more important is the possibility that confronting messy and complex phenomena will lead to theory development and make researchers spell out more explicitly the connections among different social psychological models, methods, and effects. A successful example of a project of this kind is the experimental work of Andersen and her colleagues on the concept of transference (Andersen & Cole, 1990; Andersen, Glassman, Chen & Cole, 1995; Chen, chap. 8, this volume). It has long been observed that a person frequently responds to others as if those other people were standins for important figures from the person's past, such as that person's parents (Scharff, 1992). In other words, how people construct their representations of others often seems to be guided not by normative considerations, but by the idiosyncratic personality configurations they have encountered in other individual people. Andersen and her colleagues have demonstrated that when particular personality traits are attributed to a target person, people have a tendency to misattribute other traits to that person in a way that could not be predicted from an abstract implicit personality theory (Schneider, 1973). Instead, people incorrectly recall that

< previous page

page_205

next page >

< previous page

page_206

next page >

Page 206 targets with one attribute also have other characteristics simply because other people they have known have also possessed traits in that combination. The starting points of this work were basic models of social categorization and construct accessibility, and the tools and ideas borrowed from those literatures allowed the researchers to shed light on a long-recognized and fascinating phenomenon. What these studies also did, however, was expand our understanding of person perception by showing how representations of significant others can have direct effects on how we construe other people and their behavior. Another phenomenon recently reconceptualized using concepts developed in part by students of person perception is defensive projection. The idea that people sometimes seem to be quick to detect in other people those very qualities that they are motivated not to see in themselves is as familiar to lay people as it is to psychologists. However, it seems fair to state that the modal belief among experimental psychologists has long been that defensive projection is a cute psychoanalytic idea for which there is little evidence (Holmes, 1968, 1978). Skepticism about the existence of defensive projection might also have been fueled by the lack of precision of Freudian definitions of the phenomenon (see Sandler & Freud, 1985). Newman, Duff, and Baumeister (1997) recasted projection in terms of the literature on social construct accessibility. In other words, they framed the phenomenon of defensive projection as follows: Certain trait concepts (undesired ones) are often chronically accessible, and therefore they easily come to mind when making certain judgments (i.e., judgments about other people) and so are frequently inferred (i.e., projected). To explain why unwanted or feared traits might become chronically accessible, and thus readily available for use when disambiguating other people's behavior, Newman et al. (1997) built a bridge between the person perception literature and research on thought suppression. A now extensive literature shows that vigorous efforts to avoid a thought paradoxically boost the accessibility of that thought – a phenomenon known as the rebound effect (Wegner, 1992). Newman et al. (1997) proposed that similar rebound effects might be observed when the thought being suppressed is the possibility that one might possess an undesired personality trait. In a series of studies, they provided support for the hypothesis that people's efforts to avoid thoughts about specific self-relevant and threatening personality characteristics lead people to attribute those characteristics to others. Defensive projection thus occurs as a side effect of suppression, not as a result of a direct effort to make oneself feel better by expelling a threatening thought (cf. Vaillant, 1977). One contribution of this research, besides shedding light on defensive projection, is its articulation of connections between person perception and the literature on thought suppression. In addition, it suggests a rethinking of the concept of motivated person perception. Motivated person perception typically implies a direct desire to perceive a particular characteristic in another person for some self-serving purpose (e.g., Dunning et al., 1995; Klein & Kunda, 1992). The Newman et al. research on projection shows how motives can instead have indirect effects on person perception, by enhancing the accessibility of concepts that are then only incidentally used to construe other people's behavior. CONCLUSION This chapter began with the observation that the kinds of studies that beginning students of social psychology encounter in abundance – studies in which “an individual subject is exposed in isolation and in the laboratory to sets of trait or behavior words or sentences presented via slides or printed lists” (Landman & Manis, 1984, p. 113) – seem to be a far cry from the high-impact studies of conformity (Asch, 1956), obedience (Milgram, 1974), and bystander intervention (Latané & Darley, 1970) that usually define the field for nonspecialists. Social psychologists have been known to voice their displeasure with the prominence of such studies, and some have even accused person perception and person memory research of being a self-referential, esoteric, and even unexciting enterprise. In fact, even the most prominent person perception and person memory researchers have occasionally expressed discomfort with certain features of their research. The history of the study of person perception and person memory was then briefly reviewed, along with some reasons that these areas of research became so prominent in social psychology. This review suggests that person perception and person memory researchers have provided a level of analysis that has been useful for social psychologists studying a wide range of the field's classic topics. In fact, it could even be argued that social psychologists seeking to understand the kinds of phenomena with which the field has traditionally concerned itself handicap themselves if they ignore basic person perception and person memory research. In summary, students of person perception and person memory have contributed enormously to the understanding of interpersonal behavior and still have more insights to offer. However, the range of application of their models will

< previous page

page_206

next page >

< previous page

page_207

next page >

Page 207 continue to expand only if they also continue to confront the challenges that will lead to conceptual and methodological growth. ENDNOTES 1. This definition could also be stretched to include the perception of groups, but a discussion of the similarities and differences between the processes involved in processing information about individuals and groups (e.g., Hamilton & Sherman, 1996; Wyer, Bodenhausen, & Srull, 1984) is beyond the scope of this chapter, as are the distinctions between self-perception and other perception (Bem, 1972; Prentice, 1990). 2. In fact, the evaluative component of impression formation dominated research before 1980 (Hamilton et al., 1980b). Hamilton et al. attributed this, in part, to the fact that researchers had at their disposal Anderson's (1968) widely utilized likableness ratings for a large set of personality traits.

< previous page

page_207

next page >

< previous page

page_208

next page >

page_208

next page >

Page 208 This page is intentionally left blank.

< previous page

< previous page

page_209

next page >

Page 209 Part III. Subjective Perception and Motivated Judgment

< previous page

page_209

next page >

< previous page

page_210

next page >

page_210

next page >

Page 210 This page is intentionally left blank.

< previous page

< previous page

page_211

next page >

Page 211 13 On Partitioning the Fundamental Attribution Error: Dispositionalism and the Correspondence Bias Douglas S. Krull Northern Kentucky University We often speculate about why things happen. Suppose we learn that a prestigious law firm hired a new attorney and within 3 years he is a partner in the firm. We might wonder why he was promoted so quickly. Is he incredibly talented? Is he the favorite nephew of the senior partner? Suppose we hear of an airplane crash. We might wonder about the cause. Did the engines fail? Was it the pilot's fault? Suppose we meet a colleague in another department and he seems grumpy. We might wonder why he acted that way. Is he a grumpy person? Is he just having a bad day? Such questions are ubiquitous, and, moreover, they are important. As Gilbert and Malone (1995) pointed out, the consequences of misunderstanding the behavior of others can be quite serious: Guilty people are set free, innocent people are jailed, poor political candidates are elected, people who thought they knew their dates are raped by them, wars are started, relationships are ended. It is difficult to overestimate the problems that can occur because we misunderstand the traits of other people and the causes of their behaviors. The question of how people understand and explain the behavior of other people is the domain of attribution theory. An attribution is an explanation or a judgment about the cause of some event. More specifically, the subset of attribution theory discussed in this chapter primarily deals with how people make judgments about a particular type of event — the behavior of others.1 This chapter compares and contrasts two aspects of what is often called the fundamental attribution error. Dispositionalism is the tendency to prefer dispositional attributions over situational attributions. Correspondence bias is the tendency to infer that dispositions correspond to behavior. It is suggested here that, although dispositionalism and correspondence bias are related phenomena, they are not identical, and a consideration of the differences and similarities may help clarify extant research and inform subsequent research efforts. This chapter also points out that, although the tendency to prefer dispositional attributions exists under some conditions, the tendency to prefer situational attributions exists under other conditions. Moreover, it is suggested that the tendency to infer that dispositions correspond to behavior is only one side of the coin because people also tend to infer that situations correspond to behavior. Thus, both correspondence bias and situational correspondence can be thought of as subsets of the more general tendency to draw correspondent inferences. PAST RESEARCH AND THEORY Contemporary attribution theory draws its ideas from many sources. Kelley's Covariation Principle can be found in the ideas of many writers, including Mill (1872/1973). Indeed, Hewstone (1990) indicated that White (1989) traced the covariation principle back to William of Ockham, who wrote during the 14th century. Hilton and Slugoski's (1986) Abnormal Conditions Focus Model makes use of Hart and Honore's (1961) insights about commonsense knowledge. Automaticity in impression formation, a cornerstone of many modern attribution theories, draws from both Icheiser's (1943) discussion of unconscious processes of interpretation and Michotte's (1946/1963) view that causality can be experienced immediately and directly. A consideration of the many individuals and ideas that have shaped contemporary attribution theory is beyond the scope of this chapter.2 Therefore, this consideration of the past of attribution theory begins with Heider. Heider is regarded as the father of attribution theory (Jones, 1985) and with good reason. Heider's (1944, 1958) insights laid the foundation for much of the research and theory in attribution. Consider three of Heider's ideas that have contributed to attribution theory. First, Heider suggested that when people view behavior, they want to understand the stable causes of the behavior so they can predict future behavior. He wrote:

< previous page

page_211

next page >

< previous page

page_212

next page >

Page 212 It is an important principle of common-sense psychology…that man grasps reality, and can predict and control it, by referring transient and variable behavior and events to relatively unchanging underlying conditions, the so-called dispositional properties of his world (1958, p. 79). Thus, just as we may seek to understand why an apple falls to the ground so that we can predict the behavior of a piano if the rope by which it is suspended were to break, we also try to understand why people behave as they do so that we can predict their future behavior. This control function can be found in Kelley's (1967) Covariation Model and in Jones and Davis' (1965) Correspondent Inference Theory. Second, Heider stressed the distinction between personal and situational causes. This distinction is central to most research in attribution, and it seems to capture a great deal of commonsense reasoning as well, because people often make comments that seem to fall into one of these two categories. For example, if Luis observes that Sheila looks nervous, he might decide that she has an anxious personality or that she is in an anxiety-provoking situation. If Tamela sees a homeless person, she might decide that the homeless person is lazy or a victim of circumstances. If a student receives a high score on an exam, a teacher might decide that the student is intelligent or that the exam was easy. Heider's view in this matter did need some fine tuning. For example, he viewed these two types of judgments as hydraulically related (i.e., as one increases, the other decreases), prompting some researchers to assess attributions with a single internal-external or person-situation scale. This hydraulic assumption has not been strongly supported by subsequent research (e.g., Miller, Smith, & Uleman, 1981; Solomon, 1978; Taylor & Koivumaki, 1976; White, 1991), which suggests that these two types of causes should be assessed separately. Other theorists have criticized the distinction, suggesting that it has more to do with how attribution theorists think about causes than with how lay people think about causes (e.g., Antaki, 1988; Edwards & Potter, 1993). Nevertheless, most work in attribution has operated under the assumption that the universe of possible causes can be placed into one of these two fundamental categories. Third, although Heider suggested that people can attribute causality to either personal or situational factors, he noted that a personal attribution was the more likely of the two (see also Icheiser, 1943). Heider pointed out that an actor and his or her act tend to be perceived as a unit, and so personal factors tend to be seen as the origin of behavior. Heider (1944) wrote: “Often the momentary situation which, at least in part, determines the behavior of a person is disregarded and the behavior is taken as a manifestation of personal characteristics’’ (p. 361). Of course, this tendency is what Ross (1977) later called the fundamental attribution error. Two Landmark Theories Many attribution theorists used Heider's insights as a starting point for their own theories and research. Two such theories that are particularly relevant to the current theses are Kelley's (1967) Covariation Model and Jones and Davis'(1965) Correspondent Inference Theory. Each of these is discussed briefly. Kelley's (1967) Covariation Model. Suppose an acquaintance of yours, Alisa, says to you, “You've got to try Aldo's, that Italian restaurant on Main. It's terrific!” What kind of an attribution should you make about why she said this? Something about the restaurant might have caused her to offer such a fine recommendation. However, she might have said this because of something about her. Kelley suggested there are three types of information that you should consider. First, you should consider what other people have to say about Aldo's restaurant. This is called consensus information. Second, you should consider whether Alisa typically raves about restaurants. This is called distinctiveness information. If Alisa makes positive comments about every restaurant she samples, her comment on this occasion is not very distinctive. Finally, you should consider how Alisa has reacted to this particular restaurant at other times. This is called consistency information. Different patterns of these three types of information suggest different conclusions. For example, what would the pattern of low consensus, low distinctiveness, and high consistency suggest about the cause of Alisa's recommendation? Low consensus indicates that other people did not recommend Aldo's. Low distinctiveness indicates that Alisa makes positive comments about many restaurants. High consistency indicates that she consistently recommends Aldo's. Apparently something about Alisa's personality is the cause of her recommendation. In contrast, high consensus (others liked Aldo's), high distinctiveness (Alisa rarely raves about restaurants), and high consistency (Alisa often raves about Aldo's) suggests that something about the restaurant is the cause (e.g., McArthur, 1972). Of course, sometimes One may lack the ability, motivation, or information to perform such a sophisticated analysis. In this case, Kelley (1972a) suggested that we can use causal schemata to draw an attribution. For example, suppose one learns that Josh lost a racquetball

< previous page

page_212

next page >

< previous page

page_213

next page >

Page 213 tournament. If one thinks that any of a number of causes (e.g., low ability, low effort, injury) might cause one to lose a racquetball tournament, a multiple-sufficient-causes-schema is applied. With this schema, if any cause is present, the effect will occur. So if one knows that Josh is injured, one should not draw an extreme attribution about his ability or effort because the injury alone could have caused him to lose (see Kun & Weiner, 1973; McClure, 1998; Reeder, 1993; Reeder & Fulks, 1980). This type of reasoning is an example of what is referred to as the discounting principle . The discounting principle states that, “the role of a given cause in producing a given effect is discounted if other plausible causes are present” (Kelley, 1972b, p. 8). In contrast, suppose one learns that Josh won a racquetball tournament. If one thinks that a confluence of facilitating factors is necessary for people to win racquetball tournaments, then a multiple-necessary-causes-schema (where multiple causes are required to produce the effect) is applied. Here discounting is unnecessary because, even if Josh is not injured, a certain level of ability and effort is still thought to be required. Jones and Davis' (1965) Correspondent Inference Theory . Suppose you observe the following three behaviors. First, Stanley is in a hospital waiting room and he looks anxious. Second, Denise declines an all expenses paid trip to Japan. Third, at a party, you overhear Joe arguing that it is every person's responsibility to recycle, reduce dependence on fossil fuels, and protect wildlife. What should you conclude about the personalities of these individuals? Jones and Davis' Correspondent Inference Theory suggests that two pieces of information are important to consider. First, you should consider the prior probability or expectedness of the behavior. When people behave in ways that are completely expected (e.g., anxious-looking Stanley in the hospital waiting room), you should not conclude that peoples' behavior reflects their personalities (e.g., Jones, Davis, & Gergen, 1961). Second, you should consider the distinctiveness or clarity of the behavior with regard to its implications about the person's personality. For example, if Zhitang buys one car rather than another and the cars differ in many respects, one cannot be certain why Zhitang made the choice he did. However, if the cars are identical except that one is red and the other is blue, one can conclude with confidence that his preference for a certain color is the reason for his choice (e.g., Newtson, 1974). In the case of Denise declining the trip to Japan, the low probability of her behavior suggests that a personality inference is warranted. However, it is not clear what type of personality inference should be drawn. Is Denise not interested in Asian cultures? Does she have important commitments that she wishes to honor? Is she afraid of flying? In summary then, when behavior is expected or does not have clear implications about personality, one cannot draw a correspondent dispositional inference.3 What about Joe, who argued at the party that people have a responsibility to be more environmentally conscious? The probability of his behavior is not particularly high, and so it would be appropriate to draw an inference about his personality. Moreover, although one could be mistaken, a clear and reasonable inference is that his comments reflect an underlying disposition — specifically, that he is an environmentalist. This is what Jones and Davis refer to as a correspondent inference . In Jones' (1990) words, A correspondent inference is a straightforward extrapolation from the behavior observed: The behavior is seen as corresponding to or reflecting an underlying disposition of the actor…for example, to call a person hostile after observing a hostile act would be to draw a correspondent inference (pp. 46-47). Two Different Theories - Two Different Questions - Two Different Judgments. The theories of Kelley (1967) and Jones and Davis (1965) address rather different questions. Kelley's model is a causal attribution theory. The goal of this theory is to explain how people should decide the cause of behavior. Jones and Davis' theory also addresses causal attribution (one should not draw a personality inference when behavior is expected because the behavior could be due to the situation). However, the primary focus of the theory (and particularly the focus of the research that followed) is to explain how people could draw correspondent inferences from behavior, or, as Jones and Davis stated it, how people go from acts to dispositions. In summary then, Kelley's model is a theory of causal attribution, but Jones and Davis' theory is a theory of correspondent (specifically trait) inference. This is not merely a semantic distinction. Both research and theory suggest that causal attributions and correspondent inferences are not the same type of judgment and are probably not drawn through the same mechanism (e.g., Bassili, 1989; Erickson & Krull, 1999; Hamilton, 1988; Hilton, Smith, & Kim, 1995; Johnson, Jemmott, & Pettigrew, 1984; Reeder & Spores, 1983; Semin & Marsman, 1994; Smith & Miller, 1983). Consider three findings that bear on this issue. First, Johnson et al. (1984) found that knowledge that a behavior was caused by situational forces did not immunize participants to the correspondence bias. These researchers concluded, “Correspondent inferences, these data suggest, may exhibit a startling independence from assessments of causality” (Johnson et al., 1984, p. 581).

< previous page

page_213

next page >

< previous page

page_214

next page >

Page 214 Second, Bassili (1989) found that participants who were instructed to form impressions of targets showed significant trait activation, but participants who were instructed to allocate causality showed little trait activation even when covariation information suggested a person attribution. Bassili (1989) concluded, “The present results suggest that causal allocation places very little emphasis on trait concepts” (p. 293). Third, although a growing body of research suggests that trait inferences may display aspects of automaticity (see Uleman, Newman, & Moskowitz, 1996, for a review), research by Smith and Miller (1983) suggests that dispositional and situational causal attributions may be less automatic (see the discussion in Anderson, Krull, & Weiner, 1996, for possible exceptions). Perhaps as a result, causal attributions may also be less common than trait inferences. As Hamilton (1988) suggested, causal attributions may only typically be drawn when behavior is unexpected. Thus, correspondent inferences may often be relatively quick and effortless summaries of behavior that contribute to an impression, but causal attributions may typically be slower and more deliberate judgments that are most often prompted by surprising or unexpected information. Interestingly, as Hamilton (1988) pointed out, this suggests that when perceivers judge a behavior, they probably do not allocate causality (e.g., Kirsten's behavior is caused by her personality) and then zero in on a more specific inference (e.g., Kirsten is a friendly person). Rather, they may quickly draw an inference from behavior and then use this inference to draw a true causal attribution when one is needed. This distinction between causal attributions and correspondent inferences is important to the current discussion because the tendency to prefer dispositional over situational attributions (dispositionalism) and the tendency to infer that personality corresponds to behavior (correspondence bias) differ in essentially the same way. Dispositionalism is an attributional tendency, but correspondence bias is an inferential tendency. Of course Kelley's Covariation Model and Jones and Davis' Correspondent Inference Theory do have similarities. Both make use of Heider's trade-off between persons and situations. Both point out that if a behavior is typical (high consensus in Kelley's model, high prior probability in Jones and Davis' theory), it is not very informative about a person's personality. Both suggest that people desire to predict others' behavior. Moreover, neither theory necessarily describes what people actually do when they form social judgments. Rather, they are idealized models of what an astute perceiver could do. Kelley (1972a) wrote that his Covariation Model was ‘‘an idealized model and not descriptive of most everyday, informal attributions” (p. 152). Jones and McGillis (1976) wrote that Correspondent Inference Theory “does not summarize phenomenal experience; it presents a logical calculus in terms of which accurate inferences could be drawn by an alert perceiver” (p. 404). Jones and McGillis went on to point out that a major contribution of the theory has been to highlight deviations from a logical calculus. They wrote, “the role of the theory has been as much to identify attributional bias as to predict precisely the course of the social inference process” (p. 404). It is to two of these biases that we now turn. Dispositionalism and the Correspondence Bias In a classic review, Ross (1977) coined the term fundamental attribution error and defined it as the tendency “to underestimate the impact of situational factors and to overestimate the role of dispositional factors in controlling behavior” (p. 183). It may be worthwhile to partition the fundamental attribution error into two distinct tendencies: dispositionalism and the correspondence bias. Dispositionalism is defined here as the tendency to prefer dispositional over situational attributions (e.g., McArthur, 1972; see also Morris & Peng, 1994). Note that dispositionalism differs from the fundamental attribution error in at least two ways. First, dispositionalism need not be an error because dispositional factors can be more important than situational factors in determining behavior, but the fundamental attribution error is an error by definition because it involves a failure to estimate dispositions and situations correctly. Second, dispositionalism addresses the relative importance of dispositional and situational factors, but the fundamental attribution error does not. The assignment of greater weight to dispositional factors always qualifies as dispositionalism, but it need not be an error. In contrast, the attribution of behavior primarily to the situation never qualifies as dispositionalism, but it could be an instance of the fundamental attribution error, providing situational factors are insufficiently emphasized and dispositional factors overemphasized. Thus, the fundamental attribution error can only be identified if the true influence of dispositions and situations is known, but dispositionalism is present whenever dispositions are emphasized more than are situations. Correspondence bias is defined as “the tendency to draw inferences about a person's unique and enduring dispositions from behaviors that can be entirely explained by the situations in which they occur” (Gilbert & Malone, 1995, p. 21). Evidence for this tendency abounds (e.g., see Gilbert & Malone, 1995; Jones, 1979, 1990, for reviews). Indeed, Jones (1990) argued that the correspondence bias is “the most robust and ubiquitous

< previous page

page_214

next page >

< previous page

page_215

next page >

Page 215 finding in the domain of interpersonal perception” (p. 164). This tendency differs from the fundamental attribution error in at least three ways. First, the correspondence bias need not be an error because personality can correspond to behavior even when situational forces are present. Second, whereas the fundamental attribution error involves a consideration of both dispositional factors and situational factors, the correspondence bias only pertains to the relationship between dispositional factors and behavior. Third, whereas the fundamental attribution error is a true causal judgment, the correspondence bias is not. It is important to highlight the differences between dispositionalism and the correspondence bias before proceeding. Recall that dispositionalism is a causal judgment that takes both dispositional and situational factors into consideration, with greater causality attributed to personality than to the situation. In contrast, correspondence bias is an inferential judgment about what personality is like. Although it is true that what is thought about the situation influences the production of correspondence bias, what is thought about the situation is irrelevant to the definition of correspondence bias. One may infer that the situation corresponds to behavior or that it does not, but if one infers that personality corresponds to behavior, one has displayed correspondence bias. For example, one may observe happy behavior in the context of a happy situation, infer that the person's situation is a happy one, and nevertheless infer that the person has a happy personality. In fact, whether behavior is thought to be caused primarily by personality or the situation is also irrelevant to the definition of correspondence bias. Strange as it may seem, one can attribute behavior to the situation and still infer that personality corresponds to behavior. For example, one may conclude that the situation caused a person to look knowledgeable and yet still infer that the person is knowledgeable (Johnson, et al.'s [1984] participants did exactly this). In summary, dispositionalism is displayed when behavior is attributed more to personality than to the situation, rather than when behavior is attributed equally or attributed more to the situation. In contrast, regardless of what is thought about the situation, the correspondence bias is displayed when personality is thought to correspond to behavior in a situation sufficient to produce the behavior. Of course, although dispositionalism and the correspondence bias may be distinct, they are not mutually exclusive. It is quite possible to display both dispositionalism and the correspondence bias simultaneously. As described earlier, correspondent inferences may be more automatic and more common than causal attributions, and often they may be the basis for causal attributions when a true causal judgment is made. If so, correspondence bias may also be more common than dispositionalism, but may be the basis for dispositionalism when a true causal judgment is made. PRESENT RESEARCH AND THEORY Having considered some of the background of attribution, particularly the theories of Kelley (1967) and Jones and Davis (1965), and having described two apparent deviations from the logical strategies outlined by these theories, we now turn to some of the explanations that have been offered for these tendencies. Why do people prefer dispositional attributions over situational attributions, and why do they infer that personality corresponds to behavior? Many different explanations have been offered for these tendencies, and no doubt both are multiply determined (e.g., Gilbert & Malone, 1995; Jones, 1990; Ross & Nisbett, 1991). In their recent review, Gilbert and Malone (1995) suggested that four mechanisms are responsible for the correspondence bias: lack of awareness of situational constraints, unrealistic expectations for behavior, incomplete corrections of dispositional inferences, and inflated categorizations for behavior. Let us notice whether each of these factors is also a cause of dispositionalism. In addition, let us notice that culture is a cause of dispositionalism and whether culture plays a role in the production of the correspondence bias. Explanations for the Correspondence Bias and Dispositionalism Lack of Awareness of Situational Constraints. It is often stated that dispositional bias occurs because behavior is salient (i.e., vivid, noticeable) compared with the relatively pallid situational background. Heider (1958) wrote that behavior “has such salient properties it tends to engulf the total field rather than be confined to its proper position as a local stimulus whose interpretation requires the additional data of a surrounding field-the situation in social perception” (p. 54). Many an attribution theorist has concluded from this that salient behavior is an explanation for dispositional bias in judgment. However, the extant literature suggests a somewhat different conclusion. The salience of actors does play a substantial role in judgments about people, as many experiments testify. For example, Taylor and Fiske (1975) had people observe a conversation between two people (actor A and actor B). When participants faced actor A, they rated actor A as the more dominant individual — the one who

< previous page

page_215

next page >

< previous page

page_216

next page >

Page 216 set the tone and direction of the conversation. But when they faced actor B, they indicated that actor B played the greater role. Ross and Nisbett (1991) summarized the work on salience this way: To be succinct, what you attend to is what you attribute to. Indeed, there is no generalization coming from the Heider-inspired attribution literature of the 1970s that is better supported than this (p. 140). However, there is a difference between assigning greater causality to a salient person and attributing causality for salient behavior to an actor's disposition. Jones (1979) pointed out that, although people are more likely to make causal attributions to salient features of their environment than they are to nonsalient features (e.g., Taylor & Fiske, 1978), it is not entirely evident that the salience of behavior should lead to person attribution rather than to situation attribution (p. 113). Quite so. Behavior may be salient, but why should salient behavior be attributed to dispositions rather than situations? Gilbert and Malone (1995, p. 30) explained, In classical attribution terms, behavior is an effect to be explained, and dispositions and situations are two possible causes of that effect. Why, then, should the salience of the effect facilitate attribution to one of those causes? It is not, after all, the person's dispositions that are salient, but the person's behavior. Indeed, in no other case would reasonable people argue that the salience of an effect “explains” its attribution: If a physical symptom such as vomiting were particularly salient, this would not explain why physicians attribute that symptom to a virus rather than to a bacterium. However, the salience of the situation is another matter. Skinner (1971) suggested that dispositional attributions increase to the extent that situations are inconspicuous. ‘‘If we do not know why a person acts as he does, we attribute his behavior to him” (p. 55). This corresponds to how attribution theorists have typically thought about behavior. If behavior is an effect that could be attributed to either personality or the situation, then as the likelihood that the situation is a cause of behavior decreases, dispositional attributions should increase. As we have noticed, this is the logic of Kelley's discounting principle. So, for example, if we see a man sleeping at a funeral and we know that he had to work all night, we might well attribute his behavior to the situation. In contrast, if we do not know that he worked all night, we may be more likely to attribute his behavior to his personality. Several findings in the literature suggest that a lack of awareness of situational forces can cause the correspondence bias. For example, in the well-known experiment by Ross, Amabile, and Steinmetz (1977), participants were assigned one of three roles: quizmaster, contestant, or observer. For each triad of participants, the quizmaster was asked to generate 10 challenging questions using his or her general knowledge. Then the quizmaster posed the questions to the contestant while the observer looked on. We all know things that most other people do not know, and so one can imagine how challenging these questions might have been: What is Jesus' first statement in the Sermon on the Mount? What structure in the limbic system is involved in transferring new memories into permanent storage? Where is the Caspian Sea? Not surprisingly, contestants answered only about 4 questions in 10 correctly. Then the participants rated the general knowledge of the quizmaster and the contestant. On the surface, it seems that the quizmaster is more knowledgeable than the contestant. After all, the quizmaster knew the answers to all 10 questions. Of course the quizmaster had an advantage in being able to choose the questions, but this is not easy to appreciate. Observers rated the quizmasters' general knowledge higher than that of the contestants, and the contestants even rated the quizmasters' general knowledge higher than their own (they both displayed correspondence bias). Only the quizmasters did not indicate that they had greater general knowledge than the contestants. Why not? Perhaps in part because the advantage of being in the role of the quizmaster was very salient for them. One can imagine them thinking, “Of course the contestant couldn't answer most of my questions. I picked questions that most people can't answer. That doesn't mean I know more in general.” Thus, when people lack awareness of situations, or fail to fully appreciate their presence, they tend to infer that personality corresponds to behavior.4 Is a lack of awareness of situational forces also a cause of dispositionalism? Yes, although the findings are not as common or powerful as one might think. For example, research reveals that when an actor 'interacts with others who are salient because they are moving or brightly illuminated (McArthur & Post, 1977), or because they have red hair or a leg brace (McArthur & Solomon, 1978), the actor's behavior is attributed more to situational factors. Thus, this work suggests that dispositionalism is more likely when situations are less obvious. However, salience effects on dispositional measures are typically not obtained, perhaps because the salience work often used salient and nonsalient interaction partners. Even nonsalient interaction partners are

< previous page

page_216

next page >

< previous page

page_217

next page >

Page 217 still physically present, whereas it is quite possible for powerful situations to have no physical manifestation (e.g., a woman is sad because her husband recently died). Thus, the research on salience may not have provided the strongest manipulation of salience. Recent work by Seger and Krull (1999) found that when otherwise salient situations are invisible (e.g., an employer is criticizing an employee and observers do not know why, rather than knowing that the employee behaved irresponsibly and cost the company a lot of money), situational attributions decrease and dispositional attributions increase. Unrealistic Expectations for Behavior. A second reason that people might be dispositionally biased is that they do not realize how situational forces are likely to influence behavior. Consider an experiment by Bierbrauer (1979). In this experiment, participants viewed a reenactment of Milgram's (1963) well-known obedience study. In Milgram's study, participants were asked by an authority figure to give potentially painful and dangerous electric shocks to another participant (actually a confederate, and actually the confederate received no shocks). Participants were asked to give shocks of increasing intensity, and the question of interest was the level of shock that participants would administer before refusing to continue. One might expect that participants would quickly refuse to obey the authority figure, but approximately 65% of the participants displayed complete obedience, stopping only when the maximum level of shock (450 volts) had been reached. Bierbrauer's participants observed an individual who administered the maximum level of shock; they were then asked to estimate the percentage of people who would display similar behavior. Bierbrauer's participants indicated that few people would behave in such an apparently despicable fashion, presumably concluding that the individual who they observed was atypical in some way — perhaps sadistic or particularly weak-willed (correspondence bias). In Bierbrauer's study, participants had unrealistic expectations about how people are willing to behave. Participants may also fail to appreciate how people are able to behave. For example, in work using the attitude attribution paradigm (e.g., Jones & Harris, 1967), it seems that participants fail to realize that people are capable of writing convincing essays that oppose their own attitude (e.g., Miller, Ashton, & Mishal, 1990; Reeder, Fletcher, & Furman, 1989). So when essays are more extreme than people expect, they tend to infer that the essayist holds an attitude that corresponds to the content of the essay. It seems that a failure to appreciate how situations influence behavior causes correspondence bias. Research suggests that this also causes dispositionalism. Consider the investigation of Pietromonaco and Nisbett (1982), based on Darley and Batson's (1973) well-known “Good Samaritan” study of helping behavior. In Darley and Batson's study, students who either were, or were not in a hurry encountered a person in need of help (a man lying in a doorway). Pietromonaco and Nisbett described a similar scenario (a woman who appears to have a knee injury and asks a student to call her husband) and asked their participants to indicate how several variables, including being in a hurry or not, would influence helping. In Darley and Batson's study, whether participants were in a hurry had a substantial influence on helping; 63% of those who were not in a hurry offered assistance, but only 10% of those who were in a hurry. Darley and Batson also found that the personality differences they assessed had little impact on whether people offered help. In contrast, Pietromonaco and Nisbett's participants indicated that whether a person was in a hurry would make essentially no difference, and they thought that personality differences would substantially influence helping. Apparently they failed to appreciate the power of the situation (i.e., a need to be elsewhere quickly), and so assigned a greater role to personality than was warranted. Incomplete Corrections of Dispositional Inferences . People may also display correspondence bias because they initially draw dispositional inferences that correspond to behavior and then fail to adjust these inferences sufficiently. To more fully appreciate this mechanism, consider a model of the social inference process developed by Gilbert and his colleagues (e.g., Gilbert, 1989; Gilbert, Pelham, & Krull, 1988). This model suggests that when people draw dispositional inferences about the behavior of others, a multiple-stage process occurs. First, people categorize the actor's behavior (e.g., John appears happy), next they characterize, or draw a dispositional inference that corresponds to the behavior (e.g., John must be a happy sort of person), and finally they may correct this inference by considering situational information (e.g., Then again, John and his family are about to leave on a vacation to Italy. Perhaps he's not such a happy sort of person after all).5 Gilbert and his colleagues also suggested that these three stages differ in their degree of automaticity. The first two stages were thought to be substantially effortless, but the last stage was thought to be more cognitively demanding. Thus, this model suggests that when people are distracted or unmotivated, they are still able to categorize behavior and draw a dispositional (correspondent) inference, but they may fail to correct this inference and so display correspondence bias.

< previous page

page_217

next page >

< previous page

page_218

next page >

Page 218 Numerous investigations provide support for this model. In one study (Gilbert et al., 1988, Experiment 1), participants viewed a silent videotape of an anxious-appearing woman with the goal of diagnosing her personality.6 Participants were told that the woman was discussing either relatively anxiety-provoking topics or relatively calm topics. Half of the participants in each of these conditions also performed an additional task designed to make them cognitively busy. As expected, nonbusy participants considered the information about the topics and corrected their dispositional inferences. However, cognitively busy participants were less able to correct their inferences, and so displayed greater correspondence bias (see Tetlock, 1985; Webster, 1993, for similar work that manipulated motivation rather than distraction). Moreover, just as people may fail to adjust their initial dispositional inferences because they are unable to do so, they may also fail to adjust their inferences because they think their inferences are accurate. Research conducted by Reeder and his colleagues (e.g., Reeder & Brewer, 1979; Reeder & Fulks, 1980; Reeder & Spores, 1983; see also Miller et al., 1990; Morris & Larrick, 1995) suggests that, for some behaviors (e.g., behaviors related to immorality or high ability), people do not think adjustments are needed. In some cases, they are right. If a man beats his wife, one should conclude that he is abusive. In other cases, however, an adjustment is appropriate, and yet people may think it is not. For example, as mentioned earlier, people are not expected to write essays that strongly oppose their own attitudes (e.g., Reeder et al., 1989). As a result, when perceivers encounter a convincing essay, they may infer that the essayist's attitude is consistent with the essay or at least not strongly inconsistent. In summary, people may display correspondence bias because they fail to sufficiently adjust their initial correspondent dispositional inferences. Could a failure to correct an initial attribution lead to dispositionalism? Research suggests that a failure to correct could lead to dispositionalism if dispositional attributions are drawn first (see Lee, Hallahan, & Herzog, 1996). However, dispositional judgments are not always drawn first, and sometimes situational judgments may be drawn first (see Krull & Dill, 1996; Lee et al., 1996). Notice what might occur in the relatively neutral situation, where both dispositional and situational judgments are drawn simultaneously. If one fails to correct dispositional and situational inferences, both might correspond to behavior, and so correspondence bias would occur. But if one fails to correct dispositional and situational causal attributions, both might be rated as relatively extreme, and dispositionalism would not occur. Thus, it seems that a failure to correct is at best a partial explanation for dispositionalism, but it is a good explanation for the correspondence bias. Inflated Categorizations for Behavior. This fourth cause of the correspondence bias is quite different from the other three. The three causes that have been considered thus far are similar in that, in one way or another, they all pertain to the tendency for people to fail to use situational information. But the fourth cause of the correspondence bias does not have to do with failing to use situational information. Rather, it has to do with using situational information. Ironically, people's knowledge of the situation can cause them to perceive behavior differently and so display correspondence bias. The model of Trope and his colleagues explains how this can occur (e.g., Trope, 1986; Trope & Alfieri, 1997; Trope, Cohen, & Maoz, 1988). Like Gilbert's model, Trope's model suggests that a multiple-stage process occurs when people draw dispositional inferences. First perceivers identify behavior (e.g., John appears happy), then they test alternative hypotheses about the target's disposition, including the hypothesis that the target's personality corresponds to the behavior (e.g., John must be a happy sort of person). This model suggests that situational information exerts opposite effects at these two stages. In the behavior identification stage, situational information may cause the behavior to be perceived as corresponding to the situation. For example, John's ambiguously happy behavior may be perceived as happier in a situation that is clearly a happy one (e.g., John appears even happier if one knows that he and his family are going to Italy on a vacation). In the hypothesis testing stage, this same situational information may be used to discount (e.g., John and his family are about to leave on a vacation to Italy, so perhaps he does not have a particularly happy personality). How can this model explain why inflated perceptions of behavior could lead to correspondence bias? Consider the following example. Suppose Prisca and Carl observe John, who looks happy. Prisca does not know about John's vacation, and so she perceives John's happiness accurately. But Carl knows that John is soon to be on a plane for Venice, and so, expecting him to be happy because of the situation, perceives even more happiness in John's behavior than is actually present (i.e., John looks extremely happy). Of course, as described earlier, Prisca might infer that John is a happy person because she does not know about John's happy situation. But Carl does know. In fact, it is Carl's knowledge of the situation that has caused him to perceive John's behavior

< previous page

page_218

next page >

< previous page

page_219

next page >

Page 219 as exceptionally happy. Now suppose Carl estimates the influence of John's happy situation perfectly (e.g., “A trip to Italy would certainly make a person happy”) and also adjusts perfectly (“Therefore I should not infer that John's personality is as happy as his behavior would suggest”). Carl still might display correspondence bias because his perception of John's behavior was too extreme at the identification stage (e.g., ‘‘But anyone who looks that happy must be a happy sort of person, even if he is about to leave for Venice”). Thus, even if one discounts for the situation, one may fail to consider the effect of the situation on one's initial behavioral identification. Consistent with this, research indicates that when people expect a particular type of behavior, and when the behavior is sufficiently ambiguous to permit perceptual assimilation effects, people seem to perceive the behavior as more extreme and so display correspondence bias (e.g., Snyder & Frankel, 1976; Trope et al., 1988). In summary, people may show correspondence bias because, even if they estimate the power of the situation perfectly and adjust perfectly, their knowledge of the situation might bias their perception of the behavior, the information on which subsequent judgments or adjustments are based. Does inflated categorization also cause dispositionalism? This author knows of no research that has examined the effect of inflated categorizations on attributions, but there is reason to think that inflated categorizations may be substantially unrelated to dispositionalism. In classical attribution terms, behavior is an effect to be explained by a person's disposition or situation. The perception that behavior is extreme might increase the extremity or number of attributions, but there would seem to be no reason for it to cause a preference for dispositional over situational attributions. Moreover, although more extreme behavior (e.g., John looks extremely happy) should influence correspondent inferences (e.g., John must have an extremely happy personality), more extreme behavior may not lead to more extreme attributions. For example, both moderate and extreme happiness could be caused very much or very little by personality. Consistent with this reasoning, recent research has found that the extremity of behavior influences the extremity of both dispositional and situational inferences, but has little effect on dispositional and situational causal attributions (Erickson & Krull, 1999; Krull & Seger, 2000). In summary, it seems that not all causes of the correspondence bias are causes of dispositionalism. Could it be that a cause of dispositionalism is not a cause of the correspondence bias? Culture . Research suggests that some cultures seem to predispose people toward dispositional thinking. A wellknown and informative investigation of cultural differences was conducted by Miller (1984). She asked participants from the United States and India to draw attributions about events they had observed. Unlike most investigations in attribution, her participants included both adults and children of various ages. This range of participants enabled her to track the development of any cultural differences that might be found. Young children showed no cultural differences in attributions. However, as age increased, a fascinating pattern emerged. Participants in the U.S. displayed no increase in situational attributions with age, but dispositional attributions increased sharply. In contrast, Indian participants displayed no increase in dispositional attributions with age, but situational attributions increased sharply. Consistent with Miller's work, Morris and Peng (1994; see also Lee et al., 1996) found that, although U.S. participants favored dispositional attributions, Chinese participants tended to favor situational attributions. What is the cause of this cultural difference? Current theory suggests that it may stem from differences in how people view both the self (e.g., Cousins, 1989; Markus & Kitayama, 1991; Rhee, Uleman, Lee, & Roman, 1995) and others (e.g., Miller, 1984; Shweder & Bourne, 1982). People in individualist cultures (e.g., Australia, England, United States)tend to view the self as “a unique, bounded configuration of internal attributes” (Markus & Kitayama, 1994, p. 569). In contrast, people in collectivist cultures (e.g., China, Guatemala, India) tend to view the self as “inherently social — an integral part of the collective” (1994, p. 570). Markus and Kitayama (1991) pointed out that these differing views of the self can be seen in the folk wisdom in these different cultures. For example, in the United States, “the squeaky wheel gets the grease,” but in Japan, “the nail that stands out gets pounded down” (H. Yamada, personal communication, February 16, 1989). Thus, individualists seem to view themselves as autonomous individuals whose behavior is driven by their own attitudes and wishes, but collectivists seem to view themselves as a part of the groups to which they belong, and greater emphasis is placed on behaving in a manner that is consistent with the goals of those groups. The difference between individualists and collectivists extends to beliefs about others' behavior. Shweder and Bourne (1982) cited anthropological evidence from Africa, Bali, Central America, Central Asia, New Guinea, and Tahiti, and a study comparing North Americans to citizens of Bhubaneswar, Orissa, India, which suggested that, although individualists may be “culturally primed to search for abstract summaries of the autonomous individual” (p. 130), collectivists may be “culturally

< previous page

page_219

next page >

< previous page

page_220

next page >

Page 220 primed to see context and social relationships as a necessary condition for behavior” (p. 129). Thus, individualists seem to view behavior as primarily caused by actors' personalities, but collectivists seem to view behavior as substantially due to situational forces. This cross-cultural research on the self and on how people think about the behavior of others suggests a simple conclusion about Miller's (1984) work on the development of attribution differences in the U.S. and India. It seems that people in different cultures develop different ideas about why people behave as they do (see Newman, 1991). In individualist cultures, people seem to learn that personality is the primary cause of behavior (and consistent with this reasoning, children's use of trait terms increases with age; e.g., Livesley & Bromley, 1973). But in collectivist cultures, people may learn that situations are the primary cause of behavior. Moreover, personality may indeed play a greater role in individualist cultures and situations may play a greater role in collectivist cultures (e.g., Argyle, Shimoda, & Little, 1978; Ross & Nisbett, 1991). If so, what type of attributions might we expect in these cultures? Attribution theorists have suggested that people draw attributions to enhance their ability to predict behavior (e.g., Heider, 1958; Jones & Davis, 1965; Kelley, 1967), and so people should be expected to explain behavior with what they may believe the primary determinant of behavior to be: dispositions in individualist cultures and situations in collectivist cultures. This, of course, is just what the cross-cultural research has found.7 It seems that dispositionalism may be partially due to aspects of individualist cultures. Is the correspondence bias then confined to individualist cultures? No. A growing body of research suggests that the correspondence bias is universal (e.g., Choi & Nisbett, 1998; Krull, Loy, Lin, Wang, Chen, & Zhao, 1999; Toyama, 1990, 1998). For example, Krull et al. (1999) compared participants from the U.S. and Taiwan using the attitude-attribution paradigm (e.g., Jones & Harris, 1967). Participants in both cultures strongly displayed correspondence bias, and the two cultures did not differ in the degree of correspondence bias. A second investigation compared participants from the U.S. and Hong Kong using the quizmaster paradigm (e.g., Ross et al., 1977). Again, participants in both cultures strongly displayed correspondence bias, and again, the two cultures did not differ in the degree of correspondence bias. Interestingly, some investigations have found that individualists display greater correspondence bias than collectivists when the situational constraints are very salient (e.g., Choi & Nisbett, 1998, Experiment 2). If individualists and collectivists display similar levels of correspondence bias with typical constraints, why might collectivists be better able to discount when situational forces are very salient? One possibility is that very salient situational constraints promote greater causal thinking. As described earlier, the causal theories of collectivists may be relatively situational, whereas those of individualists may be relatively dispositional. Thus, causal thinking may greatly enhance discounting for situations among collectivists, but not among individualists (although it may greatly enhance discounting for dispositions among individualists). It should be emphasized, however, that although there may be more exceptions in collectivist cultures, correspondence bias seems to be the rule across cultures. Situationalism As the work described to this point indicates, people sometimes favor dispositional attributions over situational attributions. However, people sometimes favor situational attributions over dispositional attributions (this tendency is referred to here as situationalism ). For example, as described earlier, people in collectivist cultures may prefer situational attributions over dispositional attributions (e.g., Miller, 1984; Morris & Peng, 1994). Moreover, Hofstede's (1980) research suggests that a number of nations in Central America and South America (e.g., Guatemala, Ecuador, Panama) may be more collectivist than the majority of Asian nations. Little is known about the attributions drawn by people in Central America and South America, but if greater collectivism produces greater situationalism, then perhaps the cultures that are the most situational have not yet been investigated. Of course, even in the United States, people sometimes prefer situational attributions. For example, it has b