Advances in Experimental Social Psychology, Volume 36 (Advances in Experimental Social Psychology)

Citation preview

ADVANCES IN

Experimental Social Psychology

VOLUME 36

This Page Intentionally Left Blank

ADVANCES IN

Experimental Social Psychology

EDITED BY

MARK P. ZANNA DEPARTMENT OF PSYCHOLOGY UNIVERSITY OF WATERLOO WATERLOO, ONTARIO, CANADA

VOLUME 36

Elsevier Academic Press 525 B Street, Suite 1900, San Diego, California 92101-4495, USA 84 Theobald’s Road, London WC1X 8RR, UK

This book is printed on acid-free paper. Copyright ß 2004, Elsevier Inc. All Rights Reserved.

No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the Publisher. The appearance of the code at the bottom of the first page of a chapter in this book indicates the Publisher’s consent that copies of the chapter may be made for personal or internal use of specific clients. This consent is given on the condition, however, that the copier pay the stated per copy fee through the Copyright Clearance Center, Inc. (www.copyright.com), for copying beyond that permitted by Sections 107 or 108 of the U.S. Copyright Law. This consent does not extend to other kinds of copying, such as copying for general distribution, for advertising or promotional purposes, for creating new collective works, or for resale. Copy fees for pre-2004 chapters are as shown on the title pages. If no fee code appears on the title page, the copy fee is the same as for current chapters. 0065-2601/2004 $35.00 Permissions may be sought directly from Elsevier’s Science & Technology Right Department in Oxford, UK: phone: (+44) 1865 843830, fax: (+44) 1865 853333, E-mail: [email protected]. You may also complete your request on-line via the Elsevier homepage (http://elsevier.com), by selecting ‘‘Customer Support’’ and then ‘‘Obtaining Permissions.’’ For all information on all Academic Press publications visit our Web site at www.academicpress.com ISBN: 0-12-015236-3 PRINTED IN THE UNITED STATES OF AMERICA 04 05 06 07 08 09 8 7 6 5 4 3 2 1

CONTENTS Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

Aversive Racism John F. Dovidio and Samuel L. Gaertner I. II. III. IV. V. VI.

The Nature of Aversive Racism. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . The Operation of Aversive Racism . . .. . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Emergency Intervention. .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Dissociated Attitudes.. . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Combating Aversive Racism . . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Summary and Conclusions .. . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . References . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. .

4 7 10 19 25 37 43

Socially Situated Cognition: Cognition in its Social Context Eliot R. Smith and Gu¨n R. Semin I. II. III. IV. V.

Cognition is for Action . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Cognition is Embodied. . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Cognition is Situated: Emergent from Interaction of Agent and Task and Social Environment. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Cognition is Distributed Spatially and Temporally Across Tools, People, and Groups .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Summary . . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . References . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. .

57 67 74 89 104 105

Social Axioms: A Model for Social Beliefs in Multicultural Perspective Kwok Leung and Michael Harris Bond I. II. III.

Beliefs in Social Psychological Research .. . .. . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Social Axioms as a Basic Psychological Construct. . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Dimensions of Social Axioms . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . v

122 127 132

vi IV. V. VI. VII. VIII.

CONTENTS A Global Study of Social Axioms . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Evidence for the Meaning and Usefulness of the Social Axiom Dimensions. . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Clustering Nations in Terms of Their Citizens’ Beliefs. . . . . . .. . . . . . .. . . . . . .. . . .. . Roles of Social Axioms for Individual Behavior. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Conclusions and Directions for Future Research . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Appendix.. . .. . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . References. . .. . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. .

139 157 170 172 180 190 191

Violent Video Games: Specific Effects of Violent Content on Aggressive Thoughts and Behavior Craig A. Anderson, Nicholas L. Carnagey, Mindy Flanagan, Arlin J. Benjamin, Jr., Janie Eubanks, and Jeffery C. Valentine I. II. III. IV. V. VI. VII. VIII. IX. X. XI.

Introduction . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . The General Aggression Model.. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . GAM and Violent Video Games . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Overview of the Present Studies . . .. . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Experiment 1 . . .. . .. . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Experiment 2 . . .. . .. . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Experiment 3 . . .. . .. . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Correlation Study 1 . . .. . .. . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Updated Meta-Analysis . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . General Discussion. . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Appendix.. . .. . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . References. . .. . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. .

200 202 204 207 207 215 224 232 237 240 244 246

Survival and Change in Judgments: A Model of Activation and Comparison Dolores Albarracı´n, Harry M. Wallace, and Laura R. Glasman I. II. III. IV.

Introduction . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Attitude Change and Survival: Preliminary Definitions . .. . . .. . . . . . .. . . . . . .. . . .. . Activation/Comparison Model .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . Conclusion . .. . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . References. . .. . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. .

252 255 258 291 307

CONTENTS

vii

The Implicit Volition Model: On the Preconscious Regulation of Temporarily Adopted Goals Gordon B. Moskowitz, Peizhong Li, and Elizabeth R. Kirk I. II. III. IV. V.

Introduction.. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Goals as Mental Representations: The Implicit Triggering of Goals. .. . . . . . .. . Implicit Regulation via Intentions .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . Compensatory Cognition and Compensatory Behavior . .. . . . . . .. . . . . . .. . . . . . .. . Conclusion . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . References . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. . . .. . . . . . .. . . . . . .. . . . . . .. .

317 326 371 383 402 404

Index ........................................................................................... Contents of Other Volumes ...............................................................

415 429

This Page Intentionally Left Blank

CONTRIBUTORS Numbers in parentheses indicate the pages on which authors’ contributions begin.

Dolores AlbarracI´n (251), Department of Psychology, University of Florida, Gainesville, FL 32611 Craig A. Anderson (199), Department of Psychology, Iowa State University, Ames, IA 50011 Arlin J. Benjamin, Jr. (199), Department of Psychological Sciences, University of Missouri-Columbia, Columbia, MO 65211 Michael Harris Bond (119), Department of Psychology, Chinese University of Hong Kong, Hong Kong S.A.R., CHINA Nicholas L. Carnagey (199), Department of Psychology, Iowa State University, Ames, IA 50011 John F. Dovidio (1), Department of Psychology, Colgate University, Hamilton, NY 133456 Janie Eubanks (199), Department of Psychological Sciences, University of Missouri-Columbia, Columbia, MO 65211 Mindy Flanagan (199), Department of Psychological Sciences, University of Missouri-Columbia, Columbia, MO 65211 Samuel L. Gaertner (1), Department of Psychology, University of Delaware, Newark, DE 19713 Laura R. Glasman (251), Department of Psychology, University of Florida, Gainesville, FL 32611 Elizabeth R. Kirk (317), Department of Psychology, Lehigh University, Bethlehem, PA 18015 Kwok Leung (119), Department of Management, City University of Hong Kong, Hong Kong S.A.R., CHINA Peizhong Li (317), Department of Psychology, Lehigh University, Bethlehem, PA 18015

ix

x

CONTRIBUTORS

Gordon B. Moskowitz (317), Department of Psychology, Lehigh University, Bethlehem, PA 18015 GU¨n R. Semin (53), Social Psychology Department, Free University Amsterdam, Amsterdam, THE NETHERLANDS Eliot R. Smith (53), Department of Psychological Sciences, Purdue University, West Lafayette, IN 47907 Jeffery C. Valentine (199), Department of Psychological Sciences, University of Missouri-Columbia, Columbia, MO 65211 Harry M. Wallace (251), Department of Psychology, Trinity University, San Antonio, TX 78212

AVERSIVE RACISM

John F. Dovidio Samuel L. Gaertner

Race relations in the United States, especially in terms of whites’ orientations toward blacks, have been characterized by inconsistencies and ambivalence. This duality has existed virtually from the beginning. The United States was founded on the principle of ‘‘liberty and justice for all’’ and on ‘‘the proposition that all men are created equal.’’ Nevertheless, it was not until 175 years after these basic human rights were proclaimed by the Declaration of Independence and guaranteed by the Constitution that the initial civil rights legislation was passed and the United States formally recognized that black and white people were equal under the law. Such structural changes in the law, however, do not mean that racial attitudes of white Americans will quickly reflect this new legal standard. Myrdal (1944) identified the paradox between historical egalitarian values and racist traditions in the United States, describing the ‘‘American dilemma.’’ According to Myrdal, the dilemma involves The ever-raging conflict between, on the one hand, the valuations preserved on the general plane which we call the ‘‘American creed,’’ where the American thinks, talks, and acts under the influence of high national and Christian precepts and, on the other hand, the valuations on the specific planes of individual and group living, where personal and local interests; economic, social, and sexual jealousies; consideration of community prestige and conformity; group prejudice against particular persons or types of people; and all sorts of miscellaneous wants, impulses, and habits dominate his outlook. ( p. xliii)

1 ADVANCES IN EXPERIMENTAL SOCIAL PSYCHOLOGY, VOL. 36

Copyright 2004, Elsevier Inc. All rights reserved. 0065-2601/04 $35.00

2

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

The American dilemma of 1944 was thus between egalitarian ideals and personal and social forces that promote prejudice and discrimination. In this chapter we examine one contemporary legacy of this historical dilemma—‘‘aversive racism.’’ Elements of the American dilemma still remain. On the one hand, in the United States the principle of equality continues as a fundamental social value (Schuman, Steeh, Bobo, & Krysan, 1997). Moreover, whites’ expressions of prejudice toward traditionally underrepresented groups, and toward blacks in particular, have declined substantially over time. As Bobo (2001) concluded in his review of trends in racial attitudes, ‘‘The single clearest trend in studies of racial attitudes has involved a steady and sweeping movement toward general endorsement of the principles of racial equality and integration’’ (p. 269). From 1960 to the present, public opinion polls have revealed that whites increasingly support integration in schools, public transportation, jobs, and housing; whites’ support for interracial marriage has also grown correspondingly. On the other hand, evidence of racial disparity and discrimination still remain. Although disparities can have causes other than discrimination, economic indices show consistent diVerences in status based on race. For instance, the median family income for blacks is less than two-thirds that of whites, a diVerential that has widened over the past two decades (Blank, 2001). Also, on several basic measures of health and well-being, the racial gap has been maintained or, in some cases (e.g., infant mortality), has widened substantially over the past 50 years (Jenkins, 2001). Furthermore, recent studies suggest that over the life span black and white patients receive unequal treatment from medical practitioners resulting in less favorable health-related outcomes for blacks (see Smedley, Stith, & Nelson, 2003; Whaley, 1998). Steady trends toward residential integration that were observed from the 1950 to 1970 have slowed in the South and stagnated in the North (Massey, 2001). Massey (2001) observed, ‘‘Either in absolute terms or in comparison to other groups, blacks remain a very residentially segregated and spatially isolated people’’ (p. 403). Evidence suggests that discrimination is a key factor in many of these disparities. In terms of career dynamics, whites have an advantage over blacks in initial wage level and opportunities for training (Rosenfeld, 1998). In addition, blacks are more likely to be discriminated against relative to whites when economic conditions require layoVs (Elvira & Zatzick, 2002, see also Murphy-Berman, Berman, & Campbell, 1998). In addition, diVerences in health care, which have been attributed to racism by health providers (Smedley et al., 2003), occur over and above access to health insurance. Clearly, discrimination and perceptions of discrimination continue to be dominant forces in the lives of minorities in the United States. For example,

AVERSIVE RACISM

3

within the government workforce, 55% of blacks and 28% of Hispanics reported that discrimination hinders their career advancement (U.S. Merit Systems Protection Board, 1997). In the general public, nearly half of black Americans (47%) reported that they were treated unfairly during the previous month in their own community in at least one of five common situations: while shopping, at work, in restaurants or other entertainment places, in dealing with the police, and using public transportation (Gallup, 2002). Moreover, white and black Americans have very diVerent perceptions of the racial discrimination that blacks face today. More than three fourths (79%) of whites reported that blacks ‘‘have as good a chance as whites’’ to ‘‘get any kind of job,’’ but fewer than half (46%) of blacks shared that view. Whereas the vast majority (69%) of whites perceived that blacks were treated ‘‘the same as whites,’’ the majority of blacks (59%) reported that blacks were treated worse than whites. Over the past 35 years, we have explored the nature of whites’ racial attitudes to understand this duality between the generally expressed nonprejudicial views of whites in contemporary U.S. society and the persistence of significant racial disparity and discrimination. Our work built on the conceptual framework of Kovel (1970), who distinguished between dominative and aversive racism. Dominative racism is the ‘‘old-fashioned,’’ blatant form. According to Kovel, the dominative racist is the ‘‘type who acts out bigoted beliefs—he represents the open flame of racial hatred’’ ( p. 54). Aversive racists, in comparison, sympathize with victims of past injustice, support the principle of racial equality, and regard themselves as nonprejudiced, but, at the same time, possess negative feelings and beliefs about blacks, which may be unconscious. Aversive racism is hypothesized to be qualitatively diVerent than blatant, ‘‘old-fashioned,’’ racism, is more indirect and subtle, and is presumed to characterize the racial attitudes of most welleducated and liberal whites in the United States. Nevertheless, the consequences of aversive racism (e.g., the restriction of economic opportunity) are as significant and pernicious as those of the traditional, overt form (Dovidio & Gaertner, 1998; Gaertner & Dovidio, 1986). This chapter reviews this program of research, highlighting basic assumptions, key findings, and major developments. Although our research has focused on race relations in the United States, the processes of aversive racism are not limited by national or geographic boundaries and could reflect attitudes toward a number of diVerent groups when overt forms of discrimination are recognized as inappropriate (see Esses, Dovidio, Jackson, & Armstrong, 2001; Kleinpenning & Hagendoorn, 1993; Pettigrew & Meertens, 1995). We begin by reviewing the nature of aversive racism, including the contributing psychological factors and the potential conflict between whites’ conscious endorsement of egalitarian principles and unconscious negative

4

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

feelings and beliefs about blacks. We next consider basic evidence about how aversive racism operates and its moderating factors. Then we consider our research examining separately the conscious and unconscious components of aversive racism. In the last two sections we explore ways of combating the eVects of aversive racism and consider conclusions and implications.

I. The Nature of Aversive Racism A critical aspect of the aversive racism framework is the conflict between whites’ denial of personal prejudice and underlying unconscious negative feelings toward and beliefs about blacks. Because of current cultural values, most whites have strong convictions concerning fairness, justice, and racial equality. However, because of a range of normal cognitive, motivational, and sociocultural processes that promote intergroup biases, most whites also develop some negative feelings toward or beliefs about blacks, of which they are unaware or which they try to dissociate from their nonprejudiced self-images. These negative feelings that aversive racists have toward blacks do not reflect open hostility or hatred. Instead, aversive racists’ reactions may involve discomfort, uneasiness, disgust, and sometimes fear. That is, they find blacks ‘‘aversive,’’ while at the same time find any suggestion that they might be prejudiced ‘‘aversive’’ as well. Thus, aversive racism may involve more positive reactions to whites than to blacks, reflecting a pro– in-group rather than an anti–out-group orientation, thereby avoiding the stigma of overt bigotry and protecting a nonprejudiced self-image. The existence of these nearly unavoidable racial biases and the simultaneous desire to be nonprejudiced represents a basic duality of attitudes and beliefs for aversive racists that can produce racial ambivalence (see also Katz & Hass, 1988; Katz, Wackenhut, & Hass, 1986). We recognize that all racists are not aversive or subtle, that old-fashioned racism still exists, that there are individual diVerences in aversive racism, and that some whites may not be racist at all. Nevertheless, we propose that aversive racism generally characterizes the racial attitudes of a large proportion of whites who express apparently nonprejudiced views. Some of the sources of this negative aVect and beliefs are reviewed in the next section. A. NEGATIVE ATTITUDES AND RACIAL AMBIVALENCE In contrast to traditional approaches that emphasize the psychopathology of prejudice (e.g., Adorno, Frenkel-Brunswik, Levinson, & Sanford, 1950; see Duckitt, 1992), the negative feelings and beliefs that underlie

AVERSIVE RACISM

5

aversive racism are hypothesized to be rooted in normal, often adaptive, psychological processes. These processes fundamentally involve the consequences of social categorization. People inherently categorize others into groups, typically in ways that delineate one’s own group from other groups (Hamilton & Trolier, 1986). The mere categorization of people into in-groups and out-groups, even on the basis of arbitrary assignment, is suYcient to initiate (often spontaneously; Otten & Moskowitz, 2000) an overall evaluative bias in which people categorized as members of one’s own group are evaluated more favorably than are those perceived as members of another group (Brewer, 1979; Tajfel, 1970). Social categorization also fundamentally influences how people process information about others. Perceptually, when people or objects are categorized into groups, actual diVerences between members of the same category tend to be minimized (Tajfel, 1969) and often ignored in making decisions or forming impressions, whereas between-group diVerences tend to become exaggerated (Abrams, 1985; Turner, 1985). Cognitively, people retain more information in a more detailed fashion for ingroup members than for outgroup members (Park & Rothbart, 1982), have better memory for information about ways in which ingroup members are similar to and outgroup members are dissimilar to the self (Wilder, 1981), and remember less positive information about outgroup members (Howard & Rothbart, 1980). In addition, they evaluate the products of ingroup members more favorably than those of outgroup members (Ferguson & Kelley, 1964) and they work harder for groups identified as ingroups (Worchel, Rothgerber, Day, Hart, & Butemeyer, 1998). The process of social categorization also influences aVective reactions. As Insko et al. (2001) have demonstrated, categorization in terms of group membership rather than individual identity evokes greater feelings of fear and lower levels of trust in interactions with others. In the United States, because of its historical and central importance, social categorization by race is virtually automatic. Without eVort or control, whites spontaneously diVerentiate people by race, with the activation of racial categories being triggered by the actual or symbolic presence of a black person. Because of sociocultural influences, these racial categories are associated with traditional negative stereotypes of blacks (Devine, 1989; Dovidio, Evans, & Tyler, 1986), as well as negative attitudes (Dovidio, Kawakami, & Beach, 2001). Motivational factors, such as needs for power, status, and control (Operario & Fiske, 1998), also contribute to whites’ biased feelings and beliefs. Although these needs may be personal, they are also often group related. In Social Identity Theory, Tajfel and Turner (1979) proposed that a

6

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

person’s need for positive identity may be satisfied by membership in prestigious social groups. This need also motivates social comparisons that favorably diVerentiate ingroup from outgroup members. Discrimination is one way of creating ‘‘positive distinctiveness’’ for one’s group, which, in turn, can boost one’s self-esteem and promote feelings of control and superiority (Fein & Spencer, 1997; Tajfel & Turner, 1979). Competition and perceived competition between groups further exacerbate intergroup biases (Campbell, 1965; Esses et al., 2001; Sherif, 1966). These biases are functional. Tangibly, discrimination oVers economic advantages to members of the majority group and maintains that group’s political, social, and corporate power (Blumer, 1958; Bobo, 1999). Because blacks have traditionally been perceived to threaten whites’ basic values and well-being, powerful cognitive and motivational forces that are a function of social categorization and perceived threat form the basis for the negative racial feelings of aversive racists. Because of current cultural values, however, most whites also have convictions concerning fairness, justice, and racial equality (Bobo, 2001). At the same time, however, psychological and social forces contribute to whites’ egalitarian orientation. Principles of fairness, justice, and equity are universal, and they profoundly shape human functioning and social life (Kelman, 2001). Equality and justice are not only fundamental principles in the United States, but they are steadily endorsed more strongly and broadly over time and by succeeding generations (Schuman et al., 1997). The vast majority of white Americans today believe that prejudice and discrimination are wrong, and they indicate strong support for social and political equality (Bobo, 2001). Thus, it is the existence of both almost unavoidable racial biases and conscious adherence to nondiscriminatory principles that forms the basis of the ambivalence that aversive racists experience. We note that other forms of contemporary racial biases, such as modern racism (McConahay, 1986) and symbolic racism (Sears, Henry, & Kosterman, 2000) also hypothesize a conflict between the denial of personal prejudice and underlying unconscious negative feelings and beliefs. What distinguishes aversive racism from modern and symbolic racism is the nature of the conscious beliefs that permit discrimination to be expressed. Symbolic Racism Theory emphasizes that beliefs about individualism and meritocracy that become racialized motivate opposition to policies designed to benefit racial and ethnic minorities. For instance, a measure of individual diVerences in symbolic racism predicted support for Proposition 209, which was designed to dismantle aYrmative action in California in 1996 (Sawires & Peacock, 2000). Modern Racism Theory similarly proposes that beliefs associated with conservative ideologies can justify discriminatory behaviors, but this theory places more emphasis on the moderating eVects

AVERSIVE RACISM

7

of contexts that provide a justification for negative responses to minorities (e.g., a previous negative decision about a comparable white job candidate; McConahay, 1983). Whereas modern and symbolic racism characterize the attitudes of political conservatives, the aversive racism framework focuses on biases among people who are politically liberal and who openly endorse nonprejudiced beliefs, but whose unconscious negative feelings get expressed in subtle, indirect, and rationalizable ways (Gaertner & Dovidio, 1986). For example, research by Nail, Harton, and Decker (2003, Studies 1 and 2) revealed that in a context in which race was very salient, politically liberal participants responded more positively toward a black than toward a white person, a pattern that could be expected among aversive racists, whereas politically conservative respondents responded more favorably toward the white person. In further support of the aversive racism framework, Nail et al. (Study 3) found that only liberals displayed greater physiological arousal to the touch of a black versus a white person, which Nail et al. argued reflected the intrapsychic conflict associated with aversive racism.

II. The Operation of Aversive Racism The aversive racism framework also helps to identify when discrimination against blacks and other minority groups will or will not occur. Whereas oldfashioned racists exhibit a direct and overt pattern of discrimination, aversive racists’ actions may appear more variable and inconsistent. Sometimes they discriminate (manifesting their negative feelings), and sometimes they do not (reflecting their egalitarian beliefs). Our research has provided a framework for understanding this pattern of discrimination. Because aversive racists consciously recognize and endorse egalitarian values and because they truly aspire to be nonprejudiced, they will not discriminate in situations with strong social norms when discrimination would be obvious to others and to themselves. Specifically, we propose that when people are presented with a situation in which the normatively appropriate response is clear, in which right and wrong is clearly defined, aversive racists will not discriminate against blacks. In these contexts, aversive racists will be especially motivated to avoid feelings, beliefs, and behaviors that could be associated with racist intent. To avoid the attribution of racist intent, aversive racists either will treat blacks and whites equally, or they will respond even more favorably to blacks than to whites. Wrongdoing, which would directly threaten their nonprejudiced self-image, would be too costly. However, because aversive racists still possess feelings of uneasiness,

8

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

these feelings will eventually be expressed, but they will be expressed in subtle, indirect, and rationalizable ways. For instance, discrimination will occur in situations in which normative structure is weak, when the guidelines for appropriate behavior are vague or when the basis for social judgment is ambiguous. In addition, discrimination will occur when an aversive racist can justify or rationalize a negative response on the basis of some factor other than race. Under these circumstances, aversive racists may engage in behaviors that ultimately harm blacks but in ways that allow whites to maintain their self-image as nonprejudiced and that insulate them from recognizing that their behavior is not color blind. Generally, then, aversive racists may be identified by a constellation of characteristic responses to racial issues and interracial situations. First, aversive racists, in contrast to old-fashioned racists, endorse fair and just treatment of all groups. Second, despite their conscious good intentions, aversive racists unconsciously harbor feelings of uneasiness toward blacks and thus try to avoid interracial interaction. Third, when interracial interaction is unavoidable, aversive racists experience anxiety and discomfort, and consequently they try to disengage from the interaction as quickly as possible. Fourth, because part of the discomfort that aversive racists experience is due to a concern about acting inappropriately and appearing prejudiced, aversive racists strictly adhere to established rules and codes of behavior in interracial situations that they cannot avoid. Finally, their feelings will be expressed, but in subtle, unintentional, rationalizable ways that disadvantage minorities or unfairly benefit the majority group. Nevertheless, in terms of conscious intent, aversive racists intend not to discriminate against people of color—and they behave accordingly when it is possible for them to monitor the appropriateness of their behavior. The aversive racism framework thus considers the situation as a critical factor influencing the expression of racial bias by whites toward blacks. Although social influences can directly aVect the level of bias that is expressed (Pettigrew, 1959), we emphasize the moderating role of situational factors on whether the unconscious negative aspects of aversive racists’ attitudes are manifested in terms of racial discrimination. That is, whether the situation is one in which a negative act toward a black person would be attributed to racial intent, by others or by the aversive racist himself or herself, determines whether bias will be expressed. We have found consistent support for the aversive racism framework across a broad range of situations (see Dovidio & Gaertner, 1998; Gaertner & Dovidio, 1986). Much of the research reported in this chapter focuses on the responses of white college students—well-educated and typically liberal people—who are presumed to represent a prime population for aversive racism. Nevertheless, we note that many of the findings and principles we

AVERSIVE RACISM

9

discuss extend to biases exhibited by liberal noncollege populations (e.g., Gaertner, 1973). In the next sections, we describe examples of a series of diVerent studies to illustrate the operation of aversive racism. The evidence presented in this section is derived from a variety of diVerent paradigms, such as ones involving interventions to help people in need and employment or admission selection decisions.

A. SERENDIPITY AND AVERSIVE RACISM We began our research on racism naively with a simple assumption: based on diVerences in their expressed racial attitudes (see Adorno et al., 1950), conservative whites would behave in a more racially discriminatory way than would liberal whites. However, we discovered, somewhat serendipitously, that racial discrimination was complex, and it occurs in subtle as well as overt ways. In an initial study of contemporary racism and interracial helping (Gaertner, 1973), white participants residing in Brooklyn, New York, were selected for a field experiment on helping on the basis of their liberal or conservative orientations, as indicated by their political party aYliations (i.e., liberal or conservative parties in New York State) that were a matter of public record. Both the liberal and the conservative households received ostensibly wrong-number telephone calls that quickly developed into requests for assistance. The callers, who were clearly identifiable from their dialects as being black or white, explained that their car was disabled and that they were attempting to reach a service garage from a public telephone along the parkway. The callers further claimed that they had no more change to make another call and asked the participant to help by calling the garage. If the participant agreed to help and called the number, ostensibly of the garage, a ‘‘helping’’ response was scored. If the participant refused to help or hung up after the caller explained that he or she had no more change, a ‘‘not helping’’ response was recorded. If the participant hung up before learning that the motorist had no more change, the response was recorded as a ‘‘premature hang-up.’’ The first finding from this study was direct and predicted. Conservatives showed a higher ‘‘helping’’ response to whites than to blacks (92% vs. 65%), whereas liberals helped whites somewhat, but not significantly, more than blacks (85% vs. 75%). By this measure, conservatives were more biased against blacks than were liberals. Additional inspection of the data, however, revealed an unanticipated finding. Liberals ‘‘hung up prematurely’’ much more often on blacks than they did on whites (19% vs. 3%) and especially often on a black male motorist (28%). Conservatives did not discriminate in

10

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

this way (8% vs. 5%). From the perspective of black callers, the consequence of a direct ‘‘not helping’’ response and of a ‘‘premature hang-up’’ was the same: they would be left without assistance. From the perspective of the participants, however, the consequences were diVerent. Whereas a ‘‘not helping’’ response was a direct, intentional form of discrimination because it should have been clear to participants that their help was needed, a ‘‘premature hang-up’’ was a more indirect form because participants disengaged from the situation before they learned of the other person’s dependence on them, and thus participants never overtly refused assistance. Indeed, to refuse help that is perceived to be needed clearly violates the social responsibility norm, whereas the appropriateness of hanging up prematurely is unclear. Therefore, both conservative and liberal whites discriminated against blacks but in diVerent ways.

III. Emergency Intervention Another of our early experiments (Gaertner & Dovidio, 1977) demonstrates how aversive racism can operate in fairly dramatic ways. The scenario for the experiment was inspired by an incident in the mid-1960s in which 38 people witnessed the stabbing of a woman, Kitty Genovese, without a single bystander intervening to help. What accounted for this behavior? Feelings of responsibility play a key role (see Darley & Latane´, 1968). If a person witnesses an emergency knowing that he or she is the only bystander, that person bears all of the responsibility for helping and, consequently, the likelihood of helping is high. In contrast, if a person witnesses an emergency but believes that there are several other witnesses who might help, then the responsibility for helping is shared. Moreover, if the person believes that someone else will help or has already helped, the likelihood of that bystander taking action is significantly reduced. We created a situation in the laboratory in which white participants witnessed a staged emergency involving a black or white victim. We led some of our participants to believe that they would be the only witness to this emergency, whereas we led others to believe that there would be other white people who also witnessed the emergency. We predicted that, because aversive racists do not act in overtly bigoted ways, whites would not discriminate when they were the only witness and the responsibility for helping was clearly focused on them. However, we anticipated that whites would be much less helpful to black than to white victims when they had a justifiable excuse not to get involved, such as the belief that one of the other witnesses would take responsibility for helping.

AVERSIVE RACISM

11

The results clearly reflected these predictions. When white participants believed that they were the only witness, they helped both white and black victims very frequently (more than 85% of the time) and equivalently. There was no evidence of blatant racism. In contrast, when they thought there were other witnesses and they could rationalize a decision not to help on the basis of some factor other than race, they helped black victims only half as often as white victims (37.5% vs. 75%). Thus, these results illustrate the operation of subtle biases in relatively dramatic, spontaneous, and life-threatening circumstances involving a failure to help, rather than an action intentionally aimed at doing harm. Therefore, this research shows that although the bias may be subtle and the people involved may be well intentioned, its consequences may be severe. Across a range of other studies using a number of helping paradigms, we found evidence that discrimination by whites against blacks occurs primarily when norms for appropriate behavior are weak or ambiguous (Frey & Gaertner, 1986) and tend to be more pronounced when the interaction involves potential threats to the traditionally superior status of whites relative to blacks (Dovidio & Gaertner, 1981, 1983a). Another series of studies provides evidence of the potential influence of aversive racism on another type of helping (Turner & Pratkanis, 1994), support for public policies designed to benefit blacks.

A. POLICY SUPPORT AYrmative action has been one of the most hotly debated policies in American politics over the past three decades, virtually since the inception of the program (Skrentny, 1996). As the recent Grutter v. Bollinger et al. Supreme Court decision demonstrates (Supreme Court of the United States, 539 U.S., 2003), the issues remain contentious today. One popular criticism of aYrmative action centers on negative reactions based on the perceived unfairness of these policies. The protest by many whites expressed around the Regents of the University of California v Bakke (1978) case was that the medical school admissions procedures were a form of ‘‘reverse discrimination’’ that violated their fundamental beliefs about procedural justice and fairness. That is, the commonly articulated reason for challenging the admissions procedure was that the policy was discriminatory and negated individual selection, evaluation, and advancement based on merit. Considerable theoretical and empirical support exists for the idea that procedural fairness is a critical factor in determining people’s response to decision-making procedures (Lind, Kurtz, Musante, Walker, & Thibaut, 1980). With respect to aYrmative action, the more weight given to

12

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

category-based criteria, such as race or sex, the less fair the procedure is perceived to be and the more negative the reactions to the policy and the persons involved. In addition, consistent with this reasoning, people who are more committed to principles of merit tend to be more opposed to aYrmative action when they believe that discrimination is no longer a problem but tend to be more supportive of aYrmative action when they recognize the persistent eVects of discrimination (Son Hing, Bobocel, & Zanna, 2002). In two studies, we therefore examined the importance of how aYrmative action is framed on whites’ reactions to aYrmative action. Whereas many people may initially feel that aYrmative action policies are unfair when characteristics such as race or sex are weighed in the decision at the micro level, they may come to perceive the procedure as fair if, at a macro level, they recognize the value of diversity—‘‘that individuals bring with them into the organization not merely diVerent amounts of the same things, but also diVerent kinds of things that make them valuable to an organization’’ (Clayton & Tangri, 1989, p. 180). Similarly, although preferential action may be seen as unfair in a specific case, the same action may be perceived as more fair if it is presented as a compensatory response to address historical inequities. Racism may also be a factor in attitudes toward aYrmative action. In general, people higher in self-reported racial prejudice are more opposed to aYrmative action (Frederico & Sidanius, 2002; Kravitz, 1995). We have proposed that aversive racism also plays a significant role in opposition to aYrmative action policies, particularly those designed to support blacks. In one of our studies that addressed the influence of aversive racism on opposition to aYrmative action (Murrell, Dietz-Uhler, Dovidio, Gaertner, & Drout, 1994), white respondents were questioned about their perceptions of fairness and support for four common ways of presenting aYrmative action policies. Two of these policies focused on micro-level actions varying in the degree to which the action places emphasis on nonmerit factors to address disparities (preferential treatment and reverse discrimination). The other two policies provided macro-level justifications in terms of achieving diversity or remedying historical injustices. We predicted that respondents would show less resistance to policy statements with explicit macro-level justifications than to policy statements that focus on the micro-level of implementation. In addition, to evaluate the possibility that resistance to aYrmative action may be an expression of racial bias, we assessed participants’ reactions to aYrmative action policies involving three target groups: blacks, elderly persons, and handicapped persons. To the extent that racial bias is a key factor in reactions to aYrmative action, white participants would be expected to exhibit more negative responses to policies targeted at blacks than at other groups. Moreover, based on the hypothesis that aversive racism motivates opposition to aYrmative action, we predicted that

AVERSIVE RACISM

13

attitudes toward aYrmative action would be particularly negative when blacks were the salient beneficiaries and white respondents could justify their opposition on the basis of some factor other than race, such as violations of procedural fairness. Our results underscore the importance of the manner in which a policy is framed in shaping public opinion. Programs that were framed in terms of macro-justice by remedying historical injustice (past discrimination) or increasing cultural diversity were more acceptable to respondents than were those that focused on specific implementation (i.e., preferential treatment and reverse discrimination). In addition, consistent with the hypothesis that racism contributes to resisting aYrmative action, policies directed at benefiting blacks yielded generally more negative responses than policies for persons with physical disabilities or elderly persons. Moreover, supportive of the specific predictions of the aversive racism framework, whites’ responses to aYrmative action were particularly negative when the group described as benefiting was blacks (vs. disabled or elderly persons) and the goal of the policy was presented as involving preferential treatment or reverse discrimination (vs. achieving diversity or compensating for past discrimination). A second experiment (Dovidio & Gaertner, 1996) further examined the eVects of framing and explicitly examined the mediating role of perceived fairness in responses to aYrmative action benefiting diVerent groups. In this experiment, aYrmative action was framed at a macro level (i.e., to correct for past injustice) or at a micro level (i.e., positive action for the individual). We also manipulated the salience of the group associated with the policy: blacks, handicapped persons, or Native Americans. Native Americans were included because pretesting indicated that they represented a minority group that did not evoke significant negative reactions among potential participants. Consistent with the results of Murrell et al. (1994), we found, in general, that attitudes toward the policy were more negative when it was not framed in terms of macro-justice than when it was, and that attitudes were more negative when the group benefiting was blacks rather than Native Americans or handicapped people. Furthermore, as expected, the condition in which the policy was not framed in terms of correcting for past discrimination and it was described as benefiting blacks produced a uniquely high level of resistance. In addition, these eVects were mediated by perceptions of perceived fairness. Thus, it was perceptions of unfairness of a policy that benefited blacks that provided the rationale for opposing aYrmative action. These types of biases also extend to other formal forms of decision making, such as juridic decisions and personnel selection. Decision making about legal issues is considered in the next section.

14

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

B. LEGAL DECISIONS Traditionally blacks and whites have not been treated equally under the law (Sidanius, Levin, & Pratto, 1998). Across time and locations in the United States, blacks have been more likely to be perceived by jurors as guilty (Fairchild & Cowan, 1997), more likely to be convicted of crimes, and, if convicted, sentenced to longer terms for similar crimes, particularly if the victim is white (see Robinson & Darley, 1995). Although some evidence indicates that disparities in judicial outcomes are declining over time (Sommers & Ellsworth, 2001), aversive racism appears to have a continuing, subtle influence. For example, Johnson, Whitestone, Jackson, and Gatto (1995) explored how the introduction of an apparently non–race-related factor suggesting guilt can diVerentially aVect juridic decisions in ways that discriminate against black defendants. In particular, in a laboratory simulation study, Johnson et al. (1995) examined the impact of the introduction of inadmissible evidence, which was damaging to a defendant’s case, on whites’ judgments of a black or white defendant’s guilt. No diVerences in judgments of guilt occurred as a function of defendant race when all the evidence presented was admissible. However, consistent with the aversive racism framework, the presentation of inadmissible evidence increased judgment of guilt when the defendant was black but not when the defendant was white. Furthermore, suggesting the unconscious or unintentional nature of the bias, participants’ self-reports indicated that they believed that the inadmissible evidence had less eVect on their decisions when the defendant was black than when the defendant was white. Johnson et al. (1995) conclude that these results ‘‘are clearly consistent with the modern racism perspective, which suggests that discriminatory behavior will occur only when it can be justified on nonracial grounds’’ ( p. 896). Several other studies of legal decision making have yielded evidence consistent with the proposition that whites’ biases against blacks will be more pronounced when they have an apparently non–race-related justification for judging a black defendant guilty or sentencing them more severely (Knight, Guiliano, & Sanchez-Ross, 2001). However, also consistent with the aversive racism framework, when testimony is included that suggests that racial bias may be involved in the allegations against a black defendant, whites no longer racially discriminate (Sommers & Ellsworth, 2000). We similarly found that providing white jurors with other types of justifications could also lead to discriminatory outcomes in capital sentencing (Dovidio, Smith, Donnella, & Gaertner, 1997). In particular, although aversive racists did not generally discriminate on the basis of race in their recommendations for the death penalty in a capital case, when they were

AVERSIVE RACISM

15

asked their judgments independently, they did recommend the death penalty for a black defendant significantly more strongly than for a white defendant when they learned that a black juror (but not a white juror) would vote for the death penalty. When a black juror also advocates the death penalty, it is easier for aversive racists to rationalize their vote for capital sentencing as not reflecting racial bias. Another study of simulated juridic decisions by Faranda and Gaertner (1979) demonstrated how traditional and aversive forms of racism can combine to shape perceptions of a defendant’s guilt. Specifically, this study investigated the hypothesis that, whereas the racial biases of those who are likely to have traditionally racist attitudes (high authoritarians) would reflect primarily anti-black biases, the racial biases of those who are likely to exhibit aversive racism (low authoritarianism) would mainly represent pro-white biases. Thus, complementing the work of Johnson et al. (1995), this experiment examined the extent to which high- and low-authoritarian–scoring white college students playing the role of jurors would follow a judge’s instruction to ignore inadmissible prosecution testimony that was damaging to a black or white defendant. As predicted, both high- and low-authoritarian participants displayed racial biases in their reactions to the inadmissible evidence, but they did so in diVerent ways. In their ratings of certainty of guilt, high authoritarians did not ignore the inadmissible testimony when the victim was black. They were more certain of the black defendant’s guilt when they were exposed to the inadmissible evidence than when they were not presented with this testimony. For the white defendant, however, high authoritarians followed the judge’s instructions appropriately. Low-authoritarian participants, in contrast, followed the judge’s instructions about ignoring the inadmissible testimony when the defendant was black. However, they were biased in favor of the white defendant when the inadmissible evidence was presented. That is, low authoritarians were less certain of the white defendant’s guilt when the inadmissible evidence was presented than when it was omitted. Thus, lowauthoritarian participants demonstrated a pro–in-group bias. Importantly, the anti–out-group bias of high authoritarians and the pro–ingroup bias of low authoritarians both disadvantage blacks relative to whites but in fundamentally diVerent ways.

C. SELECTION DECISIONS As noted at the beginning of this chapter, labor statistics continue to demonstrate fundamental disparities in the economic status of blacks relative to whites—a gap that has not only persisted but also, in some important

16

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

aspects such as family income, has widened in recent years (see Blank, 2001). Aversive racism may be one factor that contributes to disparities in the workplace by influencing both the access of blacks to the workplace and their performance in it. At the time of hiring, aversive racism can aVect how qualifications are perceived and weighed in a manner that systematically disadvantages black relative to white applicants. In particular, the aversive racism framework suggests that bias will not be expressed when a person is clearly qualified or unqualified for a position, because the appropriate decision is obvious. However, bias is expected when the appropriate decision is unclear, for example, when it is not clear whether the candidate’s qualifications meet the criteria for selection or when the candidate’s file has conflicting evidence (e.g., some strong and some weak aspects). In one study of hiring decisions (Dovidio & Gaertner, 2000), we presented college students with excerpts from an interview and asked them to evaluate candidates for a position in an ostensibly new program for peer counseling at their university. Specifically, white participants evaluated a black or white candidate who had credentials that were systematically manipulated to represent very strong, moderate, or very weak qualifications for the position. These findings were supportive of the aversive racism framework. When the candidates’ credentials clearly qualified them for the position (strong qualifications) or the credentials clearly were not appropriate (weak qualifications), there was no discrimination against the black candidate. However, when candidates’ qualifications for the position were less obvious and the appropriate decision was more ambiguous (moderate qualifications), white participants recommended the black candidate significantly less often than the white candidate with exactly the same credentials. Moreover, when we compared the responses of participants in 1989 and 1999, whereas overt expressions of prejudice (measured by items on a self-report prejudice scale) declined over this 10-year period, the pattern of subtle discrimination in selection decisions remained essentially unchanged (see Table I). In subsequent research (Hodson, Dovidio, & Gaertner, 2002), participants were asked to help make admissions decisions for the university. Given the social climate on college campuses today, it is possible that even higher prejudice-scoring students may be concerned about viewing themselves as prejudiced. Consequently, as we have observed among lower prejudiced participants in the past, these individuals may currently express their negative attitudes in subtle, indirect, and rationalizable ways—and, relative to the general population, these higher prejudice-scoring college students may actually be low to moderate in prejudice and not view themselves as racially prejudiced. Indeed, among a comparable sample of higher

17

AVERSIVE RACISM

TABLE I Recommendations and Attributions of Personal Characteristics and Interpersonal Orientation as a Function of Candidate Qualifications and Race Strength of recommendation* Condition Strong qualifications White candidate Black candidate

1988–1989 1998–1999

Both

Percent recommended 1988–1989 1998–1999 Both

6.74 (1.41) 7.32 (1.67)

6.21 (2.09) 6.52 (1.72) 7.00 (1.60) 7.18 (1.62)

89% 95%

79% 87%

85% 91%

Moderate qualifications White candidate 6.05 (1.73) Black candidate 5.06 (1.39)

5.69 (1.60) 5.91 (1.67) 4.53 (1.64) 4.82 (1.51)

75% 50%

77% 40%

76% 45%

Weak qualifications White candidate Black candidate

2.42 (1.68) 2.81 (1.66) 3.77 (1.69) 3.50 (1.68)

5% 12%

8% 15%

6% 13%

3.05 (1.65) 3.29 (1.69)

*Values are expressed as from Dovidio & Gaertner, 2000. Means with standard deviation presented in parentheses.

prejudice-scoring participants, only 15% regarded themselves as ‘‘prejudiced against blacks.’’ We found no anti-black bias among our higher and lower prejudicescoring college participants when applicants had uniformly strong or uniformly weak college board scores and records of high school achievement. However, when applicants were strong on one dimension (e.g., on college board scores) and weak on the other (e.g., high school grades), black applicants tended to be recommended less strongly than were white applicants among higher prejudice scoring-participants. Moreover, these participants systematically changed how they weighed the criteria to justify their decisions as a function of race. For black applicants, higher prejudicescoring college participants gave the weaker dimension (college board scores or grades) greater weight in their decisions, whereas for white applicants they assigned the stronger of the qualifications more weight. Analogously, Brief, Dietz, Cohen, Pugh, and Vaslow (2000) found that white interviewers who scored high on the Modern Racism Scale (McConahay, 1986) were particularly likely to discriminate against black relative to white applicants in hiring when a business-related justification for not hiring the candidate was available. Taken together, these findings suggest that when given latitude for interpretation, higher prejudice white college participants (whom, relative to the general population may be regarded as generally

18

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

moderate to low prejudiced, see Schuman et al., 1997), give white candidates the ‘‘benefit of the doubt,’’ a benefit they do not extend to blacks.

D. SUBTLE BIAS: A SUMMARY In summary, the behavior of aversive racists is characterized by two types of inconsistencies. First, aversive racists exhibit an apparent contradiction between their expressed egalitarian attitudes and their biased (albeit subtle) behaviors. Second, sometimes (in clear situations) they act in an unbiased fashion, whereas at other times (in ambiguous situations) they are biased unintentionally against blacks. Overall, we have oVered evidence across time, populations, and paradigms that illustrates how aversive racism—racism among people who are good and well intentioned—can produce disparate outcomes between blacks and whites. As noted earlier, although the bias of aversive racists may be subtle and unintentional, its consequences may ultimately be just as severe as oldfashioned racism—threats to the well-being of blacks and the restriction of opportunities. Furthermore, the racial biases of aversive racists are often manifested in terms of a pro–in-group rather than an anti–out-group bias (Gaertner et al., 1997). A pro–in-group bias is often less readily recognized, and when it is this type of bias is often less threatening to one’s nonprejudiced self-image than is overt bias against a black person. For instance, we found that white college students did not overtly associate negative characteristics more strongly with blacks than with whites, responses that might be interpreted as reflecting anti-black attitudes, but they did associate positive characteristics more strongly with whites than with blacks (Dovidio & Gaertner, 1991; Gaertner & McLaughlin, 1983). This is not the old-fashioned, overt type of bias associated with the belief about black inferiority but instead is a modern, subtle form of bias that reflects a belief about white superiority. In addition, pro–in-group bias is typically not encompassed in legal definitions of discrimination (Krieger, 1998). However, the distinction between anti-black and pro-white responses can have important implications. First, it may provide a more comprehensive and accurate understanding of contemporary racism. The conclusions drawn from considering only the anti–out-group portion of these attitudes in isolation of pro–in-group attitudes might misrepresent the overall phenomenon. For example, Crocker and Schwartz (1985) found that when looking at only out-group attitudes, people with low self-esteem appeared more prejudiced than those with high self-esteem. However, when both in-group and out-group attitudes were considered, people with high

AVERSIVE RACISM

19

self-esteem were more biased (i.e., evaluated the out-group more negatively than the in-group) than those with low self-esteem. Second, pro-white attitudes may be one example of the broader class of phenomena, which we considered earlier in our discussion of social categorization, in which people generally favor in-group members over out-group members. Conceiving of black–white relations within a more general context of intergroup relations (while still recognizing the unique cultural and historical characteristics of the conflict that have shaped sterotypes and status relations) has theoretical and practical advantages for identifying factors that perpetuate such biases and the factors that may increase intergroup harmony. In general, then, we have obtained substantial evidence indicating the existence of aversive racism and demonstrating how it aVects interracial behavior. However, because aversive racists are guarded about appearing prejudiced, to others and to one’s self, they may consciously or unconsciously alter their responses to appear nonprejudiced, particularly in contexts in which race or racial attitudes are salient. As a consequence, aversive racists often appear nonprejudiced, in an absolute sense, on self-report measures of prejudice. Recent advances in attitude measurement, particularly in terms of the assessment of attitudes and beliefs that are out of conscious awareness (e.g., implicit attitudes) have permitted a closer examination of how the conscious and unconscious forces hypothesized within the aversive racism framework operate. We consider these developments in the next section.

IV. Dissociated Attitudes Beginning with our earliest work on the aversive racism framework, we hypothesized that a dissociation commonly exists between whites’ conscious and unconscious racial attitudes and beliefs. Recent research in social cognition has yielded new techniques for assessing unconscious, as well as conscious, attitudes and stereotypes. These techniques thus provide direct evidence about the influence of factors previously only assumed to be involved in aversive racism. A. IMPLICIT PROCESSES Borrowing from work in cognition more generally, researchers have made a fundamental distinction between explicit and implicit processes (Devine, 1989; Greenwald & Banaji, 1995). Explicit attitudes and stereotyping operate in a conscious mode and are exemplified by traditional, self-report measures of these constructs. In contrast, implicit attitudes and stereotypes are

20

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

evaluations and beliefs that are automatically activated by the mere presence (actual or symbolic) of the attitude object. They commonly function in an unconscious and unintentional fashion. Implicit attitudes and stereotypes are typically assessed using response latency procedures, memory tasks, physiological measures (e.g., galvanic skin response), and indirect self-report measures (see Blair, 2001; Dovidio et al., 2001). We, along with other researchers using response-time measures based on the assumption that racial attitudes operate like other stimuli to facilitate responses and decision making about related concepts (e.g., doctor–nurse), have found consistent evidence of whites’ generally negative implicit (unconscious) attitudes toward blacks (e.g., Dovidio et al., 1986; Fazio, Jackson, Dunton, & Williams, 1995; Gaertner & McLaughlin, 1983; Greenwald, McGhee, & Schwartz, 1998; Wittenbrink, Judd, & Park, 1997). A study from Dovidio, Kawakami, Johnson, Johnson, and Howard (1997) illustrates a paradigm we have used to assess unconscious attitudes. In this technique, we do not mention race, which might prepare our participants to censor negative feelings. Instead, we presented black and white primes (schematic faces) subliminally. On each of the key experimental trials, we first presented a sketch of a black or white person very rapidly on a computer screen, and then we covered up (‘‘masked’’) the sketch with the large letter ‘‘P’’ within an oval (to indicate a trial on which participants would make decisions about people) or ‘‘H’’ (to indicate a trial on which participants would make decisions, in control trials, about houses) in the same area of the computer screen so that participants are unaware that the sketch even appeared. Thus participants were not cognizant of the fact that we were assessing their racial beliefs and feelings. On the trials in which the letter ‘‘P’’ was clearly visible to participants, they were asked to make a decision about whether the next word that appears could ever describe a person (i.e., P for person). Next, we displayed a characteristic, for example ‘‘good’’ or ‘‘bad,’’ and recorded how long participants took to make the decision. Faster response times are assumed to reflect greater association. We then examined whether the subliminal sketch of a black or white person would aVect their decision-making times. Using this subliminal procedure we found that our white participants had more positive associations with whites than with blacks and more negative associations with blacks than with whites, even though they were not aware of the schematic faces or that the study tested their racial attitudes. These findings converge with a substantial number of studies (see Blair, 2001; Dovidio et al., 2001) using a broad range of techniques (such as the Implicit Association Test; Greenwald et al., 1998) that reveal that the vast majority of white Americans harbor unconscious negative associations about blacks.

AVERSIVE RACISM

21

Moreover, supportive of the aversive racism framework, whites’ unconscious attitudes are largely dissociated from their conscious, self-reported attitudes but somewhat less so when the motivation to respond in socially desirable ways on self-reported measures of racial attitudes is reduced (Nier, 2003). The correlation between these diVerent types of attitudes is, on average, 0.24 (Dovidio et al., 2001). Nevertheless, the development of these new techniques thus allows us to examine the independent influence of conscious and unconscious attitudes to whites’ behaviors toward blacks, as well as their joint influence. We hypothesize that this disassociation between the conscious (explicit) and unconscious (implicit) attitudes of aversive racists can subtly shape the ways that whites and blacks interact and further contribute to the diVerent perceptions that whites and blacks develop about their situations. If whites are unaware of their negative implicit attitudes, they may also be unaware of how their behaviors in interracial interactions may be influenced by these racial biases. In contrast, blacks, who can observe the negative behaviors of whites with whom they are interacting, may form very diVerent impressions about whether racial bias is operating and the degree to which it is intentionally determined. Blacks (and other minority groups) may be vigilant to signs of bias and readily attribute these actions to intentional racism (Shelton, 2000; Vorauer & Kumhyr, 2001). We examine the implications of this aspect of our framework in the next section.

B. CONFLICTING ATTITUDES, MIXED MESSAGES We propose that the dissociation between the positive conscious attitudes and the negative unconscious attitudes of aversive racists fundamentally relates to the ways they interact with blacks. In particular, conscious and unconscious attitudes influence behavior in diVerent ways and under diVerent conditions (Dovidio, Kawakami, & Gaertner, 2002; Dovidio, Kawakami, et al., 1997; Fazio & Olson, 2003; Fazio et al., 1995; Wilson, Lindsey, & Schooler, 2000). Conscious attitudes shape deliberative, wellconsidered responses for which people have the motivation and opportunity to weigh the costs and benefits of various courses of action. Unconscious attitudes influence responses that are more diYcult to monitor and control (e.g., some nonverbal behaviors; see Chen & Bargh, 1997; McConnell & Leibold, 2001) or responses that people do not view as an indication of their attitude and thus do not try to control. For instance, we have found that whites’ unconscious negative attitudes predict nonverbal cues of discomfort (increased rate of blinking) and aversion (decreased eye contact) toward blacks (see also Word, Zanna, & Cooper, 1974), whereas whites self-reported, conscious

22

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

attitudes predict overt evaluations and indications of liking toward blacks (Dovidio, Kawakami et al., 1997). Thus, aversive racists, who have positive conscious attitudes and who want to be supportive of blacks but who also harbor unconscious negative attitudes or associations (see Karpinski & Hilton, 2000), are likely to convey mixed messages in interracial interactions. Given these conflicting signals, it is not surprising that blacks are likely to approach interracial interactions with anxiety, guardedness, and underlying mistrust (Hyers & Swim, 1998; Shelton, 2000).

C. INTERRACIAL INTERACTION AND IMPLICIT AND EXPLICIT ATTITUDES These potential communication obstacles and interaction problems are exacerbated by the fact that whites and blacks have fundamentally diVerent perspectives on the attitudes implied and the actions demonstrated by whites during these interactions. Whites have full access to their conscious attitudes and are able to monitor and control their more overt and deliberative behaviors. They do not have access to their unconscious attitudes or to their less monitorable behaviors. As a consequence, whites’ beliefs about how they are behaving or how blacks perceive them would be expected to be based primarily on their conscious attitudes and their more overt behaviors, such as the verbal content of their interaction with blacks, and not on their unconscious attitudes or less deliberative (i.e., nonverbal) behaviors. In contrast to the perspective of whites, the perspective of black partners in these interracial interactions allows them to attend to both the spontaneous (e.g., nonverbal) and the deliberative (e.g., verbal) behaviors of whites. To the extent that the black partners attend to whites’ nonverbal behaviors, which may signal more negativity than their verbal behaviors, blacks are likely to form more negative impressions of the encounter and be less satisfied with the interaction than are whites (Shelton, 2000). To investigate this possibility, we conducted another experiment (Dovidio et al., 2002). We assessed perceptions of interracial interactions by whites and blacks, and we related those perceptions to white participants’ conscious attitudes, measured on a self-report prejudice scale, and unconscious attitudes, assessed with a response-latency technique. Then we arranged interracial conversations with a black and a white dyad partner around a race-neutral topic. We videotaped the interactions and subsequently had one set of coders rate the nonverbal and verbal behaviors of white participants and another set of observers rate their global impressions of participants from a videotape recorded from their partners’ perspective.

AVERSIVE RACISM

23

Fig. 1. The relationships (correlations) between measures of prejudice and participant behavior and impressions. From Dovidio, Kawakami, & Gaertner, 2002.

As we hypothesized (see Fig. 1), white participants’ self-reported racial attitudes predicted their deliberative behaviors such as their verbal friendliness toward black relative to white partners, which in turn predicted white participants’ impressions of how friendly they behaved in interactions with the black relative to the white partner. Thus, whites participants’ conscious attitudes, controllable behaviors, and self-impressions were all consonant. Unconscious racial attitudes, measured with response latencies, did not predict these verbal behaviors or white participants’ impressions of how they behaved. However, as we also anticipated, we found that white participants’ unconscious racial attitudes reflected by their response latencies predicted biases in their nonverbal behaviors (as scored by our observers), which then predicted how they were perceived by their partners (see Fig. 1). Because white participants and their partners based their impressions on diVerent aspects of the participants’ attitudes, the conscious and unconscious attitudes were dissociated and their impressions of the interaction were generally uncorrelated (r ¼ 0.11). White participants typically reported that they found the interaction satisfying, and they expressed contentment with their contributions. Their black partners, however, reported being relatively dissatisfied with the exchange and were uneasy about their partners’ behaviors. Despite white participants’ good intentions, the impressions they made were not as good as they thought. Moreover, both dyad members assumed that their partner shared the impression of the interaction as they did. Taken together, our findings on the eVects of conscious and unconscious attitudes in interracial interaction suggest that the nature of contemporary biases can shape the everyday perceptions of white and black Americans in ways that interfere with the communication and trust that are critical to

24

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

developing long-term positive intergroup relations. These diVerent perspectives and experiences of whites and blacks in interracial interaction, which happen inadvertently and occur daily, can have summative eVects over time (Feagin & Sikes, 1994) and help to contribute to the climate of misperception and distrust that characterizes contemporary race relations in the United States. The majority of blacks in America today have a profound distrust for the police and legal system, and about a third are overtly distrustful of whites in general (Anderson, 1996). In addition, blacks commonly believe that conspiracies inhibit the progress of blacks (Crocker, Luhtanen, Broadnax, & Blaine, 1999). The mixed messages that aversive racists often convey can create fundamental miscommunication in interracial interaction and produce divergent impressions among interactants that can undermine their ability to interact eYciently in task-oriented situations as well as eVectively in social situations.

D. INTERRACIAL PERFORMANCE The diVerent and potentially divergent impressions formed by blacks and whites during interracial interactions can have significant impact on their eVectiveness in task-oriented situations. Cannon-Bowers and Salas (1999) have argued that eVective teamwork requires two types of skills, those associated with the technical aspects of the job and those associated with being a member of the team. For this latter factor, team competencies include the knowledge, skills, and attitudes required to work eVectively with others. In addition to manifesting itself in terms of diVerent impressions and perceptions, contemporary bias can influence personal relations and group processes in ways that unintentionally but adversely aVect outcomes for blacks. We examined these processes in interracial dyads in which a black participant was paired with a white student who was identified as a traditionally high-prejudiced person (who expressed their bias openly), an aversive racist (who expressed egalitarian views but who showed evidence of unconscious bias), or a low-prejudiced white (who held egalitarian views and showed little evidence of unconscious bias) (Dovidio, Gaertner, Kawakami, & Hodson, 2002). These participants engaged in a problem-solving task about challenges to college students. For example, in one task they were asked to identify the five most important things that incoming students need to bring to campus. Because there were no objective measures of the quality of their team solution, we focused on the quality of their interaction (as reflected in their perceptions of friendliness and trustworthiness and feelings of satisfaction) and on their eYciency (as indexed by their time to complete the task).

AVERSIVE RACISM

25

In general, whites’ impressions of their behavior were related primarily to their self-reported expressed attitudes, whereas blacks’ impressions of whites were related mainly to whites’ unconscious attitudes. Specifically, whites who expressed egalitarian ideals (i.e., low-prejudiced whites and aversive racists) reported that they behaved in more friendly ways than did those who expressed their bias openly (i.e., high-prejudiced whites). Black partners perceived only whites who showed no evidence of unconscious bias (i.e., lowprejudiced whites) to be more friendly than those who had unconscious biases (aversive racists and high-prejudiced whites). Of all three groups, blacks were least trustful of aversive racists. Our results further revealed that whites’ racial attitudes could be systematically related to the eYciency of the interracial teams. Teams with low-prejudiced whites solved the problem most quickly. Interracial teams involving high-prejudiced whites were next most eYcient. Teams with aversive racists were the least eYcient. Presumably, the conflicting messages displayed by aversive racists and the divergent impressions of the team members’ interaction interfered with the team’s eVectiveness. To the extent that blacks are in the minority in an organization and are dependent on high-prejudiced whites or aversive racists, their performance is likely to be objectively poorer than the performance of whites who predominantly interact with other whites. Thus, even when whites harbor unconscious and unintentional biases toward blacks, their actions can have eVects sometimes even more detrimental than those of overt racists on interracial processes and outcomes. Overall, we have oVered a range of evidence across time, populations, and paradigms that illustrates how aversive racism—racism among people who are good and well intentioned—can influence the nature of interracial interactions and directly or indirectly produce disparate outcomes between blacks and whites. As noted earlier, although the bias of aversive racists may be subtle and unintentional, its consequences may ultimately be just as debilitating, for example, by creating barriers to their advancement in employment settings, to blacks as old-fashioned racism. In the next section we examine strategies for combating this insidious type of bias.

V. Combating Aversive Racism When we describe our findings formally, in papers and presentations, and informally, a question often arises, ‘‘What can we do about subtle biases, particularly when we do not know for sure whether we have them?’’ Like a

26

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

mutating virus, racism may have evolved into diVerent forms that are more diYcult not only to recognize but also to combat. Traditional prejudice-reduction techniques have been concerned with changing conscious attitudes—old-fashioned racism—and obvious expressions of bias. Attempts to reduce this direct, traditional form of racial prejudice have typically involved educational strategies to enhance knowledge and appreciation of other groups (e.g., multicultural education programs), emphasize norms that prejudice is wrong, and involve direct (e.g., mass media appeals) or indirect (dissonance reduction) attitude change techniques (Stephan & Stephan, 2001). However, because of its pervasiveness, subtlety, and complexity, the traditional techniques for eliminating bias that emphasized the immorality of prejudice and illegality of discrimination are not eVective for combating aversive racism. Aversive racists recognize that prejudice is bad, but they do not recognize that they are prejudiced. We believe, however, that aversive racism can be addressed with techniques aimed at its roots at both individual and collective levels. At the individual level, strategies to combat aversive racism need to be directed at unconscious attitudes. Aversive racists’ conscious attitudes are already favorable, and may, in fact, be instrumental in motivating change. At the intergroup level, interventions may be targeted at processes that support aversive racism, such as ingroup favoritism (Gaertner & Dovidio, 2000).

A. ADDRESSING UNCONSCIOUS ATTITUDES AND BELIEFS Aversive racism is characterized by conscious (explicit) egalitarian attitudes and negative unconscious (implicit) attitudes and beliefs. Wilson et al. (2000) argue that systems of dual attitudes, such as those involved in aversive racism, typically arise developmentally. The person’s original attitudes, through repeated occurrence, practice, and ultimately overlearning, become unconscious and automatically activated (Wyer & Hamilton, 1998). Given the historic socialization of whites and the repeated exposure to negative images of blacks in the mass media, most whites develop negative attitudes and stereotypical belief about blacks that become internalized and habitualized relatively early in life (Devine, 1989). Aversive racists, however, also subsequently develop a strong conscious commitment to equality and to being nonprejudiced. Nevertheless, according to Wilson et al.’s (2000) model, the original attitude is not replaced. It is stored in memory and is implicit and unconscious, while the newer attitude is explicit and conscious. In general, explicit attitudes can change and evolve relatively easily, whereas implicit attitudes, because they are based in overlearning and habitual reactions, are much more diYcult to alter.

AVERSIVE RACISM

27

Just because they are unconscious and automatically activated does not mean that aversive racists’ unconscious negative attitudes are immutable and inevitable. If unconscious attitudes and stereotypes can be learned, we propose that they can also be unlearned or inhibited by equally well-learned countervailing influences. Devine and Monteith (1993) observed, ‘‘Although it is not easy and clearly requires eVort, time, and practice, prejudice appears to be a habit that can be broken’’ ( p. 336). We have found that with extensive practice, either imposed externally or self-motivated, it is possible to change implicit beliefs.

1. Imposed Practice With extensive practice, individuals can develop ‘‘auto-motive’’ control of their actions through frequent and persistent pursuit of a goal, such as to not be biased or not to stereotype (Bargh, 1990). As Monteith, Sherman, and Devine (1998) note, ‘‘Practice makes perfect. Like any other mental process, thought suppression processes may be proceduralized and become relatively automatic’’ ( p. 71). Consistent with this line of reasoning, we found in a series of studies (Kawakami, Dovidio, Moll, Hermsen, & Russin, 2000) that automatic stereotype activation can be reduced and eliminated with training to not stereotype members of a group. In particular, participants in this research practiced extensively to respond in ways either consistent with prevailing racial stereotypes (by indicating ‘‘yes’’ to stereotype-consistent pairings of black and white photographs and traits and responding ‘‘no’’ to stereotypeinconsistent pairings) or to negate racial stereotypes (by responding ‘‘no’’ to stereotype-consistent pairings and ‘‘yes’’ to stereotype-inconsistent pairings). At the end of the session, participants performed a response-latency task to assess their unconscious, automatically activated racial stereotypes. Whereas those participants who were in the condition in which they responded aYrmatively to conventional stereotypic associations showed equivalent evidence of unconscious racial stereotypes before and after the training exercise, those who practiced negating stereotypes demonstrated a significant decrease in unconscious stereotyping after training. These eVects of practice in negating stereotypes were also still evident for 24 hours after the training. Although such direct strategies appear to be promising, these kinds of intensive and time-consuming approaches may be limited in their general applicability. Alternative promising strategies, however, take advantage of aversive racists’ genuine interest in being nonprejudiced to motivate significant and enduring change.

28

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

2. Motivation and Self-Regulation Because aversive racists consciously endorse egalitarian values and truly want to be nonprejudiced, it may be possible to capitalize on their good intentions and induce self-motivated eVorts to reduce unconscious biases on becoming aware of them. Work by Monteith, Devine, and their colleagues (e.g., Devine & Monteith, 1993; Monteith & Voils, 1998) has revealed that when low-prejudiced people recognize discrepancies between their potential behaviour toward minorities (i.e., what they would do) and their personal standards (i.e., what they should do), they feel guilt and compunction, which produces motivations to respond without prejudice in the future. In their process model of prejudice reduction, Devine and Monteith (1993) further suggest that individuals who are committed to maintaining egalitarian standards learn to reject old, biased ways of responding and to adopt new, nonprejudiced ways. Over time and with practice, these people learn to reduce prejudicial responses and to respond in ways that are consistent with their nonprejudiced personal standards. Thus, this process of self-regulation, which is initiated by making people aware of their potential for racial bias, may produce changes in even unconscious negative responses when extended over time. We directly investigated this possibility (see Dovidio, Kawakami, & Gaertner, 2000). White participants, who were categorized as low or high in prejudice on the basis of their self-reported prejudice, completed a task making them aware of discrepancies between what they would do and what they should do (i.e., their personal standards) in interracial situations (Devine & Monteith, 1993). We assessed emotional reactions and, using a response-latency task, initial unconscious racial stereotyping. Three weeks later participants returned to the laboratory and completed the unconscious stereotyping tasks and another measure of the ‘‘would–should’’ discrepancy. We hypothesized that initial discrepancies between one’s actions (what one would do) and personal standards (what one should do) would generate stronger feelings of guilt and compunction and produce more self-initiated eVorts at change among low-prejudiced than among high-prejudiced participants. The eVects of this self-regulatory process were expected to be reflected in decreased discrepancies and unconscious stereotyping. As anticipated, greater discrepancies between what one would do and should do produced higher levels of guilt in the first session, and this relationship occurred primarily for low-prejudiced participants. These findings indicate the potential initiation of self-regulatory processes for low- but not high-prejudiced participants. When participants returned 3 weeks later, we found an overall greater alignment (i.e., smaller discrepancy) between what one would and should do—an indication that both high- and

AVERSIVE RACISM

29

low-prejudiced participants showed a decrease in overt expressions of bias. However, as hypothesized, low- and high-prejudiced whites diVered in terms of the extent to which they internalized these changes. Low-prejudiced whites who had larger initial discrepancies showed greater reductions in unconscious stereotyping (r ¼ 0.56); in contrast, for high-prejudiced whites the relationship was weaker ( 0.07) and nonsignificant. These findings demonstrate that the good intentions of aversive racists can be harnessed to produce self-initiated change in even unconscious biases with appropriate awareness, eVort, and practice over time. Recently Son Hing, Li, and Zanna (2002) have extended work along these lines by examining responses to being made aware of hypocrisy on participants’ subsequent interracial responses. This study was conducted in Canada with Asians as the target group. Participants were classified on the basis of the combination of their responses to an explicit, self-report measure of prejudice toward Asians and a measure of implicit associations (spontaneous completion of words as Asian stereotypes). Participants who were low in explicit prejudice but who showed implicit biases were identified as aversive racists; those low in both explicit and implicit bias were considered truly low prejudiced. Participants next wrote an essay, reinforcing their overt egalitarian orientation, about the importance of treating Asians fairly. Participants in the hypocrisy condition, which was designed to sensitize people to violations of their egalitarian principles, were asked to write briefly about two situations in which they reacted negatively or unfairly to an Asian person. Participants in a control condition were not asked to write about these situations. Participants then completed a mood questionnaire and, ostensibly after the study was completed, were asked to complete a survey for the university’s student government on how funding cuts should be allocated to various campus groups. They were informed that a 20% cut in funding was needed for the budget of 10 campus groups. The main dependent measure was the amount of funding recommended for the Asian Students’ Association. In general, the results supported the prediction that making people aware of violations of their egalitarian principles would primarily arouse guilt among aversive racists who actually harbor negative feelings toward Asians, and thus produce compensatory behavior in recommended funding among aversive racists but not among nonprejudiced participants. Aversive racists in the hypocrisy condition experienced uniquely high levels of guilt and displayed the most generous funding recommendations for the Asian Students’ Association. The funding recommendations of truly low-prejudiced participants were not aVected by the hypocrisy manipulation. Son Hing et al. (2002) concluded that making people aware of their biases is particularly eVective at reducing bias among people who explicitly endorse egalitarian

30

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

principles while also possessing implicit biases—the factors that characterize aversive racists. Strategies that emphasize intergroup processes, such as intergroup contact and social categorization, represent alternative, complementary approaches to these individual-level approaches. We examine one such approach in the next section.

B. REDIRECTING IN-GROUP BIAS One basic argument we have made in our research on aversive racism is that the negative feelings that develop toward other groups may be rooted, in part, in fundamental, normal psychological processes. One such process, identified in the classic work of Tajfel, Allport, and others, is the categorization of people into in-groups and out-groups, ‘‘we’s’’ and ‘‘they’s.’’ As we noted earlier, social categorization, particularly in terms of in-groups (‘‘we’s) and out-groups (‘‘they’s’’), is a fundamental process that contributes to aversive racism (Gaertner et al., 1997). In general, the mere categorization of people into ingroups and outgroups has a profound influence on social perception, aVect, cognition, and behavior. Because race is a fundamental type of social categorization in the United States, race is associated with strong ingroup biases. If bias is linked to fundamental, normal psychological processes, then attempts to ameliorate bias should be directed not at eliminating the process but rather at redirecting the forces to produce more harmonious intergroup relations. The process of social categorization is not completely unalterable. By shifting the basis of categorization from race to an alternative dimension we can potentially alter who is a we and who is a they, undermining a contributing force to aversive racism. Categories are hierarchically organized, with higher-level categories (e.g., nations) being more inclusive of lower-level ones (e.g., cities or towns). By modifying a perceiver’s goals, motives, perceptions of past experiences, expectations, as well as factors within the perceptual field and the situational context more broadly, there is opportunity to alter the level of category inclusiveness that will be most influential in a given situation. This malleability of the level at which impressions are formed is important because of its implications for altering the way people think about members of ingroups and outgroups, and consequently about the ways whites in general, and aversive racists in particular, respond to blacks. Because categorization is a basic process that is fundamental to intergroup bias, we have targeted this process as way of addressing the eVects of aversive racism. The next section explores how the forces of categorization

AVERSIVE RACISM

31

can be harnessed and redirected toward the reduction, if not the elimination, of racial bias. This approach is represented by the Common Ingroup Identity Model (Gaertner & Dovidio, 2000; Gaertner, Dovidio, Anastasio, Bachman, & Rust, 1993). 1. The Common In-Group Identity Model The Common In-Group Identity Model is rooted in the social categorization perspective of intergroup behavior and recognizes the central role of social categorization in both reducing and creating intergroup bias (Tajfel & Turner, 1979). Specifically, if members of diVerent groups are induced to conceive of themselves more as a single, superordinate group rather than as two separate groups, attitudes toward former out-group members will become more positive through processes involving pro–in-group bias. Thus, changing the basis of categorization from race to an alternative dimension can alter who we is and who they are, undermining a contributing force to contemporary forms of racism, such as aversive racism. Formation of a common identity, however, does not necessarily require groups to forsake their ethnic or other subgroup identities. It is possible for members to conceive of themselves as holding a ‘‘dual identity’’ in which both subgroup and superordinate groups are salient simultaneously. Substantial evidence across a variety of settings in support of the Common Ingroup Identity Model has been found (Gaertner & Dovidio, 2000). In one test of the model, which investigated the causal role of common group identity in reducing bias (Gaertner, Mann, Murrell & Dovidio, 1989), members of two separate laboratory-formed groups were induced through various structural interventions (e.g., seating arrangement) either to maintain their original group identities (i.e., conceive of themselves as diVerent groups) or to recategorize themselves as one superordinate group. As predicted, the manipulation to encourage recategorization of former out-group members within a common group identity produced more inclusive representations that ultimately mediated lower levels of intergroup bias, primarily by increasing the attractiveness of the former out-group members. Additional research also shows that interventions that have been demonstrated to reduce prejudice, such as cooperative interaction and appropriately structured intergroup contact, reduce bias, at least in part, by altering the intergroup cognitive representations. With respect to cooperative interaction, we brought two three-person laboratory groups, which had worked separately on a problem-solving task, together under conditions designed to vary independently the members’ representations of the six-person aggregate as one group or two groups and the presence or absence of intergroup cooperative interaction (Gaertner, Mann, Dovidio, Murrell, & Pomare, 1990).

32

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

Fig. 2. The eVect of cooperative interaction on improving attitudes toward an out-group occurs primarily by creating more inclusive, one-group representations Thick arrows indicate statistically significant paths (P < 0.05). From Gaertner et al., 1990.

As in our earlier experiment (Gaertner et al., 1989), our manipulation designed to influence representations of the aggregate as one group did produce stronger one-group representations, which in turn mediated lower degrees of intergroup bias in evaluations of original in-group and out-group members. Also, consistent with our hypothesis, when separate group identities were initially salient, cooperative interaction (see Fig. 2) decreased the extent to which the aggregate felt like two groups and increased the extent it felt like one. The inclusive, one-group representation then led to more favorable evaluations of out-group members, which contributed to reduced intergroup bias. Using survey techniques under more naturalistic circumstances, we also found converging support for the proposition that the features specified by the Contact Hypothesis (Allport, 1954; Williams, 1947) reduce intergroup bias, in part, because they transform members’ representations of the memberships from separate groups to a single, more inclusive group. Participants in these studies included students attending a multiethnic high school (Gaertner, Rust, Dovidio, Bachman & Anastasio, 1996), banking executives who had experienced a corporate merger involving a wide variety of banks across the United States (Bachman, 1993), and college students from blended families whose households are composed of two formerly separate families trying to unite into one (Banker & Gaertner, 1998). Consistent with the role of an inclusive group representation that is hypothesized in the Common In-group Identity Model, across all three

AVERSIVE RACISM

33

studies (1) conditions of intergroup contact that were perceived as more favorable predicted lower levels of intergroup bias, (2) more favorable conditions of contact predicted more inclusive (one group) and less exclusive (diVerent groups) representations, and (3) more inclusive representations mediated lower levels of intergroup bias and conflict (see Gaertner, Dovidio, Nier, Ward, & Banker, 1999). Recently a longitudinal study of stepfamilies found evidence supportive of the direction of causality between the constructs proposed by our model across time (Banker, 2002). Thus, across a variety of intergroup settings and methodological approaches we have found basically strong and consistent support for the Common Ingroup Identity Model. 2. Reducing Racial Biases: Experimental Evidence We have applied the general principles of the Common In-Group Identity to reducing racial biases in laboratory and field settings. Two studies reported by Nier, Gaertner, Dovidio, Banker, and Ward (2001) illustrate the eVectiveness of this approach for addressing whites’ biases toward blacks specifically. Another study (Houlette et al., 2004) explored a range of biases, including racial bias, among elementary school children. In a laboratory experiment (Nier et al., Study 1), white college students participated in a session with a black or white confederate. These students were induced to perceive of themselves as separate individuals participating in the study at the same time or as members of the same laboratory team. The participants evaluated their black partners significantly more favorably when they were teammates than when they were just individuals without common group connections. In contrast, the evaluations of the white partner were virtually equivalent in the team and individual conditions. Thus, inducing a common in-group identity was particularly eVective at producing positive responses toward blacks. The second study (Nier et al., Study 2) was a field experiment conducted at the University of Delaware football stadium before a game between the University of Delaware and Westchester State University. Black and white students approached fans from both universities just before the fans entered the stadium. These fans were asked if they would be willing to be interviewed about their food preferences. Our student interviewers wore either a University of Delaware or Westchester State University hat. By selecting white fans wearing clothing that identified their university aYliation, we systematically varied whether fans and our interviewers had a common or diVerent university identities in a context in which we expected university identities to be particularly salient. We predicted that making a common identity salient would increase compliance with the interviewer’s request, particularly when the interviewer was black.

34

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

Supportive of predictions from the Common In-Group Identity Model, white fans were significantly more cooperative with a black interviewer when they shared a superordinate university identity than when they did not (60% vs. 38%). For white interviewers, with whom they already shared racial group membership, the eVect was much less pronounced (43% vs. 40%). Thus, in field and laboratory settings, racial out-group members were accorded especially positive reactions when they shared common in-group identity with white participants relative to when the context did not emphasize their common group membership. These studies suggest the value of combating aversive racism at its roots, by strategically controlling the forces of ingroup favoritism that can produce subtle racial biases associated with aversive racism. In a recent study (Houlette et al., 2004), we attempted to evaluate these principles further in the context of the Green Circle intervention program, which is designed to combat a range of biases (based on weight and sex, as well as race and ethnicity) with young children. The guiding assumption of the Green Circle Program, which is practically and theoretically compatible with the Common In-Group Identity Model, is that helping children bring people from diVerent groups conceptually into their own circle of caring and sharing fosters appreciation of their common humanity, as well as respect for their diVerences. In particular, facilitators engage children in a variety of exercises designed to expand the circle. The facilitator points out that, ‘‘All of us belong to one family—the human family.’’ Paralleling the Common InGroup Identity Model, Green Circle assumes that an appreciation of common humanity will increase children’s positive attitudes toward people who would otherwise remain outside their circle of inclusion. First- and secondgraders either participated in the Green Circle Program or were in a control group of classes that did not yet have the program. In terms of outcomes, the Green Circle intervention motivated the children to be more inclusive with their most preferred playmate. Specifically, compared to children in the control condition who did not participate in Green Circle activities, those who were part of Green Circle showed significantly greater change in willingness to select other children who were diVerent than themselves in race and in sex as a child that they ‘‘would most want to play with.’’ These changes in the most preferred playmate involve a child’s greater willingness to cross group boundaries in making friends—a factor that is one of the most potent influences in producing more positive attitudes toward the out-group as a whole (Pettigrew, 1998). In addition, these intergroup friendships can have cascading eVects by reducing bias among peers. Making people aware that their friends have friends from another group also reduces prejudice toward the group as a whole (Wright, Aron, McLaughlin-Volpe & Ropp, 1997).

AVERSIVE RACISM

35

In summary, the experiments reviewed in this section show that creating a common group identity can combat a range of overt expressions of racial bias. In addition, other research has demonstrated that emphasizing common group membership can address other types of biases that are associated with aversive racism, such as orientations toward policies designed to benefit blacks and other traditionally disadvantaged groups. In a survey study of white adults, Smith and Tyler (1996, Study 1) measured the strength of respondents’ superordinate identity as ‘‘American’’ and also the strength of their identification as ‘‘white.’’ Regardless of whether they strongly identified with being white, those respondents with a strong American identity were more likely to base their support of aYrmative action policies that would benefit blacks and other minorities on concerns about fairness for diVerent groups rather than on self-interest or white group-interest. However, for those who identified themselves more strongly with being white than with being American, their position on aYrmative action was determined more strongly by concerns regarding the personal impact of these policies. This pattern of findings suggests that a strong superordinate identity (such as being American) allows individuals to support policies that would benefit members of other racial subgroups without giving primary consideration to their own instrumental needs (see also Huo, Smith, Tyler, & Lind, 1996). The next section considers how creating a common in-group identity can influence the basic motivational orientations and cognitive processes that form the basis for racial biases. 3. A Common In-Group Identity and the Motivational Orientations Within the aversive racism framework, we propose that the negative feelings, beliefs, and behaviors often are expressed subtly and indirectly— in ways that are not readily attributable (by others or themselves) to racial bias and thus do not threaten an aversive racist’s nonprejudiced self-image. From this perspective, a major motive of whites in interracial situations is to avoid wrongdoing. As we have demonstrated across a range of experiments, whites appear to monitor their interracial behaviors closely to avoid discriminating against blacks when norms for appropriate behaviors are clearly defined. Moreover, these attempts by aversive racists to avoid wrongdoing appear to involve significant conscious eVort. Richeson and Shelton (2003) found that whites high in implicit prejudice toward blacks performed more poorly on a cognitively demanding task after interacting with a black person than did whites low in implicit prejudice. Richeson and Shelton proposed that the cognitive eVort required by high implicitly prejudiced whites to monitor their interracial behavior depleted their cognitive resources, resulting in a

36

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

decrement in performance on the subsequent task. Thus, whites can, at least under some circumstances and with some eVort, successfully suppress negative beliefs, feelings, and behavior toward blacks when it is obvious that such expressions reflect racial bias. Although motivations to avoid thinking, feeling, or behaving in a prejudicial way can have positive interracial consequences, such as limiting social conflict, it can also have unintended negative consequences. Besides depleting aversive racists’ cognitive resources, eVorts to avoid wrongdoing and suppress prejudicial thoughts have two important potential costs for interracial interactions. First, a concern about avoiding wrongdoing increases anxiety that can motivate avoidance or premature withdrawal from the interaction (as evidenced by Liberal Party members hanging up prematurely more frequently on black than on white callers, as described earlier; see Gaertner, 1973). This avoidant reaction precludes the opportunity for meaningful, self-revealing exchanges between in-group and out-group members (Hyers & Swim, 1998). Second, in view of recent work on stereotype suppression and rebound (e.g., Bodenhausen & Macrae, 1996), it is possible that once this self-imposed suppression is relaxed, negative beliefs, feelings, and behaviors would be even more likely to occur than if they were not suppressed initially. In the search for strategies that could eliminate the indirect, rationalizable ways that aversive racists discriminate we have also considered the importance of establishing positive interpersonal and intergroup motivations rather than simply suppressing negative motivations. The Common In-group Identity Model, because it focuses on redirecting the forces of in-group favoritism, oVers such promise. Specifically, the recognition of a common ingroup identity potentially changes the motivational orientation or intentions of aversive racists from trying to avoid wrongdoing to trying to do what is right. Although this shift in orientation is subtle, it can have fundamental benefits. For instance, it may relieve intergroup anxiety (see Stephan & Stephan, 1985) and reduce the likelihood of negative consequences of eVortful attempts to avoid wrongdoing, such as the increased accessibility of negative thoughts, feelings, and behavior that occur when suppression is relaxed (Monteith et al., 1998; Wegner, 1994). Some preliminary evidence from our laboratory suggests the potential promise of a common in-group identity to alter motivation in just such a positive way (Dovidio, Gaertner, & Kawakami, 1998; Gaertner & Dovidio, 2000). In this experiment, white participants who were about to interact with a white or a black confederate were either asked to try to avoid wrongdoing, instructed to try to behave correctly toward the other person, informed that they were part of the same team with their partner and competing against a

AVERSIVE RACISM

37

team at a rival institution, or were given no instructions. The dependent measure of interest was the relative accessibility of negative thoughts, as assessed by changes in responses on a Stroop color-naming task after the interaction relative to responses on a baseline Stroop task administered before the interaction (see Lane & Wegner, 1995). A rebound eVect would be reflected in greater accessibility (operationalized in terms of longer colornaming latencies) of negative relative to positive words on the posttest Stroop task. We hypothesized that, because the primary motivation of aversive racists in interracial interaction is to avoid wrongdoing and thus to suppress negative thoughts and feelings, participants explicitly instructed to avoid wrongdoing and those given no instructions would show relatively strong accessibility of negative thoughts after interacting with a black confederate. In contrast, we expected participants instructed to behave correctly and those in the ‘‘same team’’ condition (who were hypothesized to adopt a positive orientation on their own) would escape such a rebound eVect. The results, while preliminary, are very encouraging. When the confederate was white, the experimental conditions did not diVer significantly in the accessibility of negative thoughts from one another or from baseline. When the confederate was black, however, the increased accessibility of negative relative to positive characteristics (from the pretest to the posttest) in the avoid wrongdoing and no instructions conditions was significantly greater than in the do right and same team conditions, in which there was an increase in the accessibility of positive relative to negative thoughts. The pattern of these findings suggests that the development of a common in-group identity can alter motivation in interracial situations from one of suppressing negative thoughts, feelings, and actions to one that is positive, more appetitive, and prosocial—and in a way that does not ironically result in further increases in negative thoughts. These findings are particularly encouraging to us because they illustrate the eVectiveness of the Common In-group Identity Model for addressing individual-level biases and particularly the underlying dynamics of aversive racism.

VI. Summary and Conclusions This chapter has described the concept of aversive racism, considered the factors contributing to aversive racism, demonstrated empirically how it aVects outcomes for blacks and shapes interracial interactions, and explored how it can be combated. Despite apparent consistent improvements in expressed racial attitudes over time, aversive racism continues to exert a

38

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

subtle but pervasive influence on the lives of black Americans. This bias is expressed in indirect and rationalizable ways that restrict opportunities for blacks while insulating aversive racists from ever having to confront their prejudices. Most of the work reviewed in this chapter has focused on the influence of contemporary racial biases of whites toward blacks because of the central role that racial politics has played in the history of the United States. Within the United States, this dynamic also relates to varying degrees to orientations toward women (Dovidio & Gaertner, 1983b; Rudman & Kilianski, 2000) and more recently to homosexuals (Hebl, Foster, Mannix, & Dovidio, 2002). Moreover, we propose that many of the principles of aversive racism also apply more generally to the responses of the majority group to minority groups in contexts in which egalitarian ideals are valued and discrimination is censured (Dovidio, Gaertner, Anastasio, & Sanitioso, 1992). For instance, Pettigrew and Meertens (1995) have found that whereas blatant prejudice in Europe is related to the unconditional exclusion or severe limitation of immigrants, subtle prejudice is associated with constraints in immigration, such as prerequisite educational levels, that can be justified on the basis of factors ostensibly unrelated to race or ethnicity. However, paralleling the scenario that we described for race relations, when conditions change and people feel threatened—such as after the terrorist attacks of September 11, 2001, in the United States—subtle biases may become more open and result not only in incidents of overt violence but also directly in attitudes and political actions against immigrants and immigration (Esses, Dovidio, & Hodson, 2002). Bias associated with aversive racism is an elusive phenomenon, and the situation plays a critical moderating role. When an interracial situation is one in which an action could be readily attributed to racial bias, aversive racists carefully monitor their interracial behaviors and do not discriminate. In fact, they may respond even more favorably to blacks than to whites as a way of aYrming their nonprejudiced self-images. But when the situation is ambiguous, norms for appropriate behavior are not clear, the circumstances permit a justification for negative behavior on the basis of some factor other than race, or aversive racists are not conscious of their actions, their bias is expressed. The challenge for addressing aversive racism thus resides in its elusiveness. Because aversive racists are unaware of their unconscious negative attitudes and their eVects, and truly embrace their egalitarian self-image, they are motivated to deny the existence of these feelings and not to recognize or take responsibility for the adverse impact of their behavior on blacks. The subtle processes underlying discrimination motivated by aversive racism can be identified and isolated under the controlled conditions of the laboratory;

AVERSIVE RACISM

39

however, at the societal and organizational levels, at which the controlled conditions of an experiment are rarely possible and multiple factors may shape decision making simultaneously, this process presents a substantial challenge to the equitable treatment of members of disadvantaged groups. Krieger (1995), in the Stanford Law Review, observed: ‘‘Herein lies the practical problem. . . . Validating subjective decisionmaking systems is neither empirically nor economically feasible, especially for jobs where intangible qualities, such as interpersonal skills, creativity, and ability to make sound judgments under conditions of uncertainty are critical’’ ( p. 1232). Thus the operation of aversive racism may go largely unnoticed and unaddressed in naturalistic settings. In addition, to the extent that discrimination reflects in-group favoritism (see also Gaertner et al., 1997), it is particularly diYcult to address legally. Krieger (1998) adds, ‘‘Title VII is poorly equipped to control prejudice resulting from in-group favoritism. . . . In-group favoritism manifests itself gradually in subtle ways. It is unlikely to trigger mobilization of civil rights remedies because instances of this form of discrimination tend to go unnoticed. If they are noticed, they will frequently seem genuinely trivial or be economically unfeasible to pursue. . . . For this reason as for others, we cannot expect existing equal opportunity tools adequately to prevent, identify, or redress this more modern form of discrimination’’ (pp. 1325–1326). It is apparent that new techniques are needed to address this and other contemporary forms of racism. Developing individual, intergroup, and societal-level interventions that not only control the expressions of aversive racism but also address the negative components of aversive racism has critical social implications. Aversive racism represents a latent form of bias whose expression is strongly moderated by social circumstances and norms. A change in conditions or norms can allow this bias to operate more directly and openly (see Rogers & Prentice-Dunn, 1981). For instance, research on interracial aggression has demonstrated that under normal circumstances whites are not more aggressive and harmful toward blacks than toward whites. Overt and unprovoked aggression toward blacks would readily be perceived as racist. However, when whites are first antagonized by another person’s aggressiveness, when they feel freed from prevailing norms through conditions that make them feel anonymous and deindividuated, or when norms change from censuring to supporting aggression, whites exhibit more aggressiveness toward blacks than toward whites (Donnerstein & Donnerstein, 1973; Donnerstein, Donnerstein, Simon, & Ditrichs, 1972; Kawakami, Spears, & Dovidio, 2002; Rogers & Prentice-Dunn, 1981). Latent racism also has important organizational implications. In corporate settings, racial discrimination in personnel selection decisions emerges

40

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

when norms, which typically sanction discrimination, change and an organizational authority condones discrimination (Brief, Buttram, Elliott, Reizenstein, & McCline, 1995). Thus, besides its subtle contemporary influence, if left unaddressed, aversive racism provides the seed for bias to emerge when conditions allow or encourage a more open expression of discrimination. Future research on aversive racism has several productive avenues to pursue. One direction involves further exploration of ways to identify aversive racists. Although there is debate about the meaning and validity of implicit associations, we believe that these techniques can provide valuable insights into the dynamics of aversive racism. Karpinski and Hilton (2000) have argued that the types of responses measured by implicit techniques, such as the Implicit Association Test (Greenwald et al., 1998), reflect environmental associations rather than personal endorsement of evaluations and stereotypes. In our earlier formulation of the aversive racism framework (Gaertner & Dovidio, 1986), we hypothesized that a broad range of cognitive, aVective, and motivational influences contributed to the unconscious forces involved in aversive racism. One of these was that exposure to sociocultural influences relating to institutional and cultural racism (Jones, 1997), the cultural transmission of stereotypes (Schaller, Conway, & Tanchuk, 2002), and the portrayal of blacks in the mass media (Devine, 1989) would create and support unconscious negative associations with blacks and influence behavior toward blacks in consequential ways. The facts that (1) these associations come to mind often automatically and without intentional control for many whites, as implicit techniques have demonstrated, and (2) that these associations predict subtle behavioral manifestations of bias (Dovidio et al., 2002; Son Hing et al., 2002), particularly in the absence of conscious personal endorsement, are consistent with our aversive racism framework. In addition, although many of the techniques used to measure implicit evaluation and stereotypes focus on processes reflective of ‘‘cold’’ cognition, with little involvement of aVective process, implicit racial evaluations are also related to fundamental aVective and motivational reactions. Phelps et al. (2000) demonstrated that responses on the Implicit Association Test, but not responses on a self-report measure of prejudice, predicted activation of the amygdala, a subcortical structure of the brain that plays a role in emotional learning and evaluation and is responsive to threats. Thus, although we still contend that unconscious aVective and cognitive processes contribute to aversive racism in ways beyond the eVects of implicit negative associations (Nail et al., 2003), we acknowledge that implicit associations are a key component and that the techniques developed to measure implicit evaluations and stereotypes are important tools for studying aversive racism and for distinguishing aversive racists from truly nonprejudiced people.

AVERSIVE RACISM

41

Before the development of such measures, we generally assumed that, given the pervasive psychological and social forces promoting bias, most white Americans who said they were not prejudiced were actually aversive racists. However, evidence of the prevalence of implicit negative racial attitudes and stereotypes (Blair, 2001), the dissociation between explicit attitudes and implicit associations (Dovidio et al., 2001), and the diVerent types of influences on behavior of implicit and explicit attitudes (Dovidio, Kawakami, & Gaertner, 2002; Fazio et al., 1995) suggest that the distinction between aversive racists and truly nonprejudiced people is a critical one that can now be examined directly. Indeed, work on implicit attitudes has already produced evidence that people who are high on implicit prejudice exhibit particularly strong motivations to suppress negative interracial behavior (Richeson & Shelton, 2003) and are particularly sensitive to threats to their egalitarian images (Son Hing et al., 2002). Recently Son Hing, Chung-Yan, Grunfeld, Robichaud, and Zanna (2004) also demonstrated that among people who report on explicit measures that they are low in prejudice, those who have negative implicit attitudes show the subtle pattern of discrimination that characterizes aversive racism, whereas those low on implicit as well as explicit prejudice behave in a consistent egalitarian manner. In addition, understanding the developmental factors that are associated with low levels of implicit biases (TowlesSchwen & Fazio, 2001) and the types of interventions that can successfully combat implicit biases (Kawakami et al., 2000) can help to address a crucial factor in aversive racism, unconscious negative attitudes and associations. A second potential direction for future research on aversive racism is to explore the potential evolution of social biases from overt and blatant forms to aversive forms. Norms against the expression of prejudice associated with certain types of stigmas vary considerably (Crandall & Eshleman, 2003). In general, the extent to which a person’s membership in a negatively viewed out-group (i.e., a stigmatized group) is perceived to be controllable is one of the strongest determinants of whether individuals will openly express negative feelings and beliefs and discrimination (Weiner, 1995). For instance, people who possess stigmas that are perceived to be more controllable (e.g., homosexuality, obesity, alcoholism) are regarded much more negatively and are more likely to be the targets of open discrimination. However, over time some of these group memberships become more socially acceptable and sanctions against the expression of prejudice become stronger. For instance, attitudes toward homosexuals have become more favorable in recent years, and increasingly local laws and organizational policies prohibit discrimination against gays and lesbians. Thus, studying attitudes and discrimination against homosexuals may provide a case study of how prejudices evolve from blatant to aversive types. Recent research (Hebl et al., 2002) has

42

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

already demonstrated that, at least in some circumstances, although discrimination against people perceived to be gay did not occur in open, formal ways, people did discriminate in more subtle, interpersonal ways (e.g., nonverbally). Similarly, studying how subtle biases (e.g., toward people of Middle Eastern descent) become more blatant in response to current events (e.g., terrorist attacks, wars) can provide further insight into the etiology of aversive racism. A third direction for future research involves the development of strategies and interventions to combat aversive racism more broadly, and to design strategies that will have sustained impact. As we have illustrated in this chapter, strategies that emphasize a common group identity may be eVective at decreasing discrimination and changing motivations from avoiding wrongdoing, which can have inadvertent negative consequences, to motivations to do what is right. Strengthening these types of positive motivations so that they become automatically activated (Kawakami et al., 2000; Moskowitz, Gollwitzer, Wasel, & Schaal, 1999) can counteract the eVect of the influence of implicit prejudice. However, in a society in which race has a special significance historically and socially, and racism has roots in both institutions and culture (Jones, 1997), it may be diYcult to sustain this common group identity and positive motivation. Hewstone (1996) has argued that, at a practical level, interventions designed to create a common, inclusive identity may not be suYciently potent to ‘‘overcome powerful ethnic and racial categorizations on more than a temporary basis’’ (p. 351). Nevertheless, creating perceptions of common group identity may form the foundation allowing other processes to operate with complementary eVects (Pettigrew, 1998). For instance, creating a common ingroup identity can reduce intergroup threat and anxiety, which can increase the likelihood of intergroup contact and lead to more personalized interactions. Personalized self-disclosing interactions, in turn, further reduce intergroup bias (Brewer & Miller, 1984; Miller, 2002) and can create the kinds of experiences that are associated with low levels of implicit prejudice (Towles-Schwen & Fazio, 2001). Thus, interventions to combat aversive racism may need to involve a ‘‘cocktail’’ of strategies that reciprocally address the intergroup, personal, and social-cognitive roots of racial biases. In conclusion, over the past 35 years we have argued that aversive racism is a fundamental and insidious form of racial bias that significantly aVects the lives of blacks in the United States. In this chapter, we have documented the eVect, examined the social and psychological processes that contribute to aversive racism, and have explored techniques to combat it. Aversive racism produces significant adverse outcomes for blacks (e.g., through bias in selection decisions, as our employment selection and college admissions studies show) and it subtly influences the nature of interracial interactions

AVERSIVE RACISM

43

in ways that contribute to the miscommunication and mistrust that have historically characterized race relations in the United States. Aversive racists display their biases subtly (e.g., in terms of nonverbal behaviors) and inconsistently (e.g., mainly in situations in which a negative response can be justified on the basis of a nonracial factor). As a consequence, blacks often receive conflicting messages in their interactions with aversive racists, and they perceive whites’ behaviors as unpredictable and thus their positive responses as insincere and untrustworthy (Crocker & Major, 1994). Moreover, because people tend to overattribute intentionality to another person’s actions (Jones & Harris, 1967), particularly for actions that are negative, blacks who feel discriminated against will likely assume that the white person’s behaviors were motivated by conscious, ‘‘old-fashioned’’ racism. Because an aversive racist does not discriminate with conscious intention and is not aware that he or she is discriminating on the basis of race, an aversive racist will be quick to deny evidence of personal prejudice. An aversive racist’s denial of intentional discrimination, although genuine, may then intensify racial conflict and distrust. Thus, it is important that people, including both whites and blacks, become aware of the existence and impact of aversive racism to understand the diVerent perspectives of members of diVerent racial groups, to facilitate more eVective communication and, ultimately, to take appropriate personal, social, and legal action to create a truly egalitarian society. Because aversive racists genuinely endorse egalitarian principles, once aware of their biases, they can help contribute to the solution rather than to the problem of racial tension, conflict, and inequality.

Acknowledgments Supported by NIMH Grant MH 48721. We are grateful to Adam Pearson for his helpful comments on an earlier version of the paper.

References Abrams, D. (1985). Focus of attention in minimal intergroup discrimination. British Journal of Social Psychology, 24, 65–74. Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950). The authoritarian personality. New York: Harper. Allport, G. W. (1954). The nature of prejudice. New York: Addison-Wesley. Anderson, J. (1996). Black and blue. New Yorker, April 29 and May 6, 62–64.

44

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

Bachman, B. A. (1993). An intergroup model of organizational mergers. Unpublished Ph.D. Dissertation, Department of Psychology, University of Delaware, Newark, DE. Banker, B. S. (2002). Intergroup conflict and bias reduction in stepfamilies: A longitudinal examination of intergroup relations processes. Unpublished Ph.D. Dissertation, Department of Psychology, University of Delaware, Newark, DE. Banker, B. S., & Gaertner, S. L. (1998). Achieving stepfamily harmony: An intergroup relations approach. Journal of Family Psychology, 12, 310–325. Bargh, J. A. (1990). Auto-motives: Preconscious determinants of thought and behavior. In E. Higgins & R. Sorrentino (Eds.), Handbook of motivation and cognition (Vol. 2, pp. 93–130). New York: Guilford. Blair, I. V. (2001). Implicit stereotypes and prejudice. In G. B. Moskowitz (Ed.), Cognitive social psychology: The Princeton Symposium on the Legacy and Future of Social Cognition (pp. 359–374). Mahwah, NJ: Erlbaum. Blank, R. M. (2001). An overview of trends in social and economic well-being, by race. In N. J. Smelser, W. J. Wilson, & F. Mitchell (Eds.), Racial trends and their consequences (Vol. 1, pp. 21–39). Washington, DC: National Academy Press. Blumer, H. (1958). Race prejudice as a sense of group position. Pacific Sociological Review, 1, 3–7. Bobo, L. (1999). Prejudice as group position: Micro-foundations of a sociological approach to racism and race relations. Journal of Social Issues, 55(3), 445–472. Bobo, L. (2001). Racial attitudes and relations at the close of the twentieth century. In N. J. Smelser, W. J. Wilson, & F. M. Mitchell (Eds.), Racial trends and their consequences (Vol. 1, pp. 264–301). Washington, DC: National Academy Press. Bodenhausen, G. V., & Macrae, C. N. (1996). The self-regulation of intergroup perception: Mechanisms and consequences of stereotype suppression. In C. N. Macrae, C. Stangor, & M. Hewstone (Eds.), Stereotypes and stereotyping (pp. 227–253). New York: Guilford. Brewer, M. B. (1979). Ingroup bias in the minimal intergroup situation: A cognitivemotivational analysis. Psychological Bulletin, 86, 307–324. Brewer, M. B., & Miller, N. (1984). Beyond the contact hypothesis: Theoretical perspectives on desegregation. In N. Miller & M. B. Brewer (Eds.), Groups in contact: The psychology of desegregation (pp. 281–302). Orlando, FL: Academic Press. Brief, A. P., Buttram, R. T., Elliott, J. D., Reizenstein, R. M., & McCline, R. L. (1995). Releasing the beast: A study of compliance with orders to use race as a selection criterion. Journal of Social Issues, 51(3), 177–193. Brief, A. P., Dietz, J., Cohen, R. R., Pugh, S. D., & Vaslow, J. B. (2000). Just doing business: Modern racism and obedience to authority as explanations for employment discrimination. Organizational Behavior and Human Decision Processes, 81, 72–97. Campbell, D. T. (1965). Ethnocentric and other altruistic motives. In D. Levine (Ed.), Nebraska symposium on motivation (Vol. 13, pp. 283–311). Lincoln, NE: University of Nebraska Press. Cannon-Bowers, J. A., & Salas, E. (1999). Team performance and training in complex environments: Recent findings from applied research. Current Directions in Psychological Science, 7, 83–87. Chen, M., & Bargh, J. (1997). Nonconscious behavioral confirmation processes: The selffulfilling consequences of automatic stereotype activation. Journal of Experimental Social Psychology, 33, 541–560. Clayton, S. D., & Tangri, S. S. (1989). The justice of affirmative action. In F. A. Blanchard & F. Crosby (Eds.), Affirmative action in perspective (pp. 177–192). New York: SpringerVerlag.

AVERSIVE RACISM

45

Crandall, C. S., & Eshleman, A. (2003). A justification-suppression model of the expression and experience of prejudice. Psychological Bulletin, 129, 414–446. Crocker, J., Luhtanen, R., Broadnax, S., & Blaine, B. E. (1999). Belief in U.S. government conspiracies against blacks among black and white college students: Powerlessness or system blame? Personality and Social Psychology Bulletin, 25, 941–953. Crocker, J., & Major, B. (1994). Reactions to stigma: The moderating role of justifications. In M. P. Zanna & J. M. Olson (Eds.), The psychology of prejudice: The Ontario symposium. Ontario symposium on personality and social psychology (Vol. 7, pp. 289–314). Hillsdale, NJ: Erlbaum. Crocker, J., & Schwartz, I. (1985). Prejudice and ingroup favoritism in a minimal intergroup situation: Effects of self-esteem. Personality and Social Psychology Bulletin, 11, 379–386. Darley, J. M., & Latane´, B. (1968). Bystander intervention in emergencies: Diffusion of responsibility. Journal of Personality and Social Psychology, 8, 377–383. Devine, P. G. (1989). Stereotypes and prejudice: The automatic and controlled components. Journal of Personality and Social Psychology, 56, 5–18. Devine, P. G., & Monteith, M. J. (1993). The role of discrepancy-associated affect in prejudice reduction. In D. M. Mackie & D. L. Hamilton (Eds.), Affect, cognition, and stereotyping: Interactive processes in intergroup perception (pp. 317–344). Orlando, FL: Academic Press. Donnerstein, E., & Donnerstein, M. (1973). Variables in interracial aggression: Potential ingroup censure. Journal of Personality and Social Psychology, 27, 143–150. Donnerstein, E., Donnerstein, M., Simon, S., & Ditrichs, R. (1972). Variables in interracial aggression: Anonymity, expected retaliation, and a riot. Journal of Personality and Social Psychology, 22, 236–245. Dovidio, J. F., Evans, N., & Tyler, R. B. (1986). Racial stereotypes: The contents of their cognitive representations. Journal of Experimental Social Psychology, 22, 22–37. Dovidio, J. F., & Gaertner, S. L. (1981). The effects of race, status and ability on helping behavior. Social Psychology Quarterly, 44, 192–203. Dovidio, J. F., & Gaertner, S. L. (1983a). Race, normative structure, and help-seeking. In B. M. DePaulo, A. Nadler, & J. D. Fisher (Eds.), New directions in helping (Vol. 2, pp. 285–302). New York: Academic Press. Dovidio, J. F., & Gaertner, S. L. (1983b). The effects of sex, status, and ability on helping behavior. Journal of Applied Social Psychology, 13, 191–205. Dovidio, J. F., & Gaertner, S. L. (1991). Changes in the nature and assessment of racial prejudice. In H. Knopke, J. Norrell, & R. Rogers (Eds.), Opening doors: An appraisal of race relations in contemporary America (pp. 201–241). Tuscaloosa, AL: University of Alabama Press. Dovidio, J. F., & Gaertner, S. L. (1996). Affirmative action, unintentional racial biases, and intergroup relations. Journal of Social Issues, 52(4), 51–75. Dovidio, J. F., & Gaertner, S. L. (1998). On the nature of contemporary prejudice: The causes, consequences, and challenges of aversive racism. In J. Eberhardt & S. T. Fiske (Eds.), Confronting racism: The problem and the response (pp. 3–32). Newbury Park, CA: Sage. Dovidio, J. F., & Gaertner, S. L. (2000). Aversive racism and selection decisions: 1989 and 1999. Psychological Science, 11, 319–323. Dovidio, J. F., Gaertner, S. L., Anastasio, P. A., & Sanitioso, R. (1992). Cognitive and motivational bases of bias: The implications of aversive racism for attitudes toward Hispanics. In S. Knouse, P. Rosenfeld, & A. Culbertson (Eds.), Hispanics in the workplace (pp. 75–106). Newbury Park, CA: Sage. Dovidio, J. F., Gaertner, S. L, & Kawakami, K. (1998, October). Multiple attitudes and contemporary racial bias. Paper presented at the annual meeting of the Society for Experimental Social Psychology, Lexington, KY.

46

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

Dovidio, J. F., Gaertner, S. L., Kawakami, K., & Hodson, G. (2002). Why can’t we just get along? Interpersonal biases and interracial distrust Cultural Diversity & Ethnic Minority Psychology, 8, 88–102. Dovidio, J., Kawakami, K., & Beach, K. (2001). Implicit and explicit attitudes: Examination of the relationship between measures of intergroup bias. In R. Brown & S. L. Gaertner (Eds.), Blackwell handbook of social psychology: Intergroup relations (Vol. 4, pp. 175–197). Oxford, UK: Blackwell. Dovidio, J. F., Kawakami, K., & Gaertner, S. L. (2000). Reducing contemporary prejudice: Combating explicit and implicit bias at the individual and intergroup level. In S. Oskamp (Ed.), Reducing prejudice and discrimination (pp. 137–163). Hillsdale, NJ: Erlbaum. Dovidio, J. F., Kawakami, K., & Gaertner, S. L. (2002). Implicit and explicit prejudice and interracial interaction. Journal of Personality and Social Psychology, 82, 62–68. Dovidio, J., Kawakami, K., Johnson, C., Johnson, B., & Howard, A. (1997). The nature of prejudice: Automatic and controlled processes. Journal of Experimental Social Psychology, 33, 510–540. Dovidio, J. F., Smith, J. K., Donnella, A. G., & Gaertner, S. L. (1997). Racial attitudes and the death penalty. Journal of Applied Social Psychology, 27, 1468–1487. Duckitt, J. (1992). The social psychology of prejudice. Westport, CT: Praeger. Elvira, M. M., & Zatzick, C. D. (2002). Who’s displaced first? The role of race in layoff decisions Industrial Relations, 41, 329–361. Esses, V. M., Dovidio, J. F., & Hodson, G. (2002). Public attitudes toward immigration in the United States and Canada in response to the September 2001 ‘‘Attack on America.’’ Analysis of Social Issues and Public Policy, 2, 69–85. Esses, V. M., Dovidio, J. F., Jackson, L. M., & Armstrong, T. M. (2001). The immigration dilemma: The role of perceived group competition, ethnic prejudice, and national identity. Journal of Social Issue, 57, 389–412. Fairchild, H. H., & Cowan, G. (1997). The O. J. Simpson trial: Challenges to science and society. Journal of Social Issues, 53(3), 583–591. Faranda, J., & Gaertner, S. L. (1979, March). The effects of inadmissible evidence introduced by the prosecution and the defense, and the defendant’s race on the verdicts by high and low authoritarians. Paper presented at the annual meeting of the Eastern Psychological Association, New York. Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995). Variability in automatic activation as an unobtrusive measure of racial attitudes: A bonafide pipeline? Journal of Personality and Social Psychology, 69, 1013–1027. Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition research: Their meaning and uses. Annual Review of Psychology, 54, 297–327. Feagin, J. R., & Sikes, M. P. (1994). Living with racism: The black middle-class experience. Boston, MA: Beacon Press. Fein, S., & Spencer, S. J. (1997). Prejudice as self-image maintenance: Affirming the self through derogating others. Journal of Personality and Social Psychology, 73, 31–44. Ferguson, C. K., & Kelley, H. H. (1964). Significant factors in over-evaluation of own groups’ products. Journal of Abnormal and Social Psychology, 69, 223–228. Frederico, C. M., & Sidanius, J. (2002). Racism, ideology, and affirmative action revisited: The antecedents and consequences of ‘‘principled objections’’ to affirmative action Journal of Personality and Social Psychology, 82, 488–502. Frey, D. L., & Gaertner, S. L. (1986). Helping and the avoidance of inappropriate interracial behavior: A strategy that perpetuates a nonprejudiced self-image. Journal of Personality and Social Psychology, 50, 1083–1090.

AVERSIVE RACISM

47

Gaertner, S. L. (1973). Helping behavior and racial discrimination among liberals and conservatives. Journal of Personality and Social Psychology, 25, 335–341. Gaertner, S. L., & Dovidio, J. F. (1977). The subtlety of white racism, arousal, and helping behavior. Journal of Personality and Social Psychology, 35, 691–707. Gaertner, S. L., & Dovidio, J. F. (1986). The aversive form of racism. In J. F. Dovidio & S. L. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 61–89). Orlando, FL: Academic Press. Gaertner, S. L., & Dovidio, J. F. (2000). Reducing intergroup bias: The Common Ingroup Identity Model. Philadelphia, PA: The Psychology Press. Gaertner, S. L., Dovidio, J. F., Anastasio, P. A., Bachman, B. A., & Rust, M. C. (1993). The common ingroup identity model: Recategorization and the reduction of intergroup bias. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 4, pp. 1–26). New York: John Wiley & Sons. Gaertner, S. L., Dovidio, J. F., Banker, B., Rust, M., Nier, J., Mottola, G., & Ward, C. (1997). Does racism necessarily mean anti-blackness? Aversive racism and pro-whiteness In M. Fine, L. Powell, L. Weis, & M. Wong (Eds.), Off white (pp. 167–178). London: Routledge. Gaertner, S. L., Dovidio, J. F., Nier, J., Ward, C., & Banker, B. (1999). Across cultural divides: The value of a superordinate identity. In D. Prentice & D. Miller (Eds.), Cultural divides: Understanding and overcoming group conflict (pp. 173–212). New York: Russell Sage Foundation. Gaertner, S. L., Mann, J. A., Dovidio, J. F., Murrell, A. J., & Pomare, M. (1990). How does cooperation reduce intergroup bias? Journal of Personality and Social Psychology, 59, 692–704. Gaertner, S. L., Mann, J., Murrell, A., & Dovidio, J. F. (1989). Reducing intergroup bias: The benefits of recategorization. Journal of Personality and Social Psychology, 57, 239–249. Gaertner, S. L., & McLaughlin, J. P. (1983). Racial stereotypes: Associations and ascriptions of positive and negative characteristics. Social Psychology Quarterly, 46, 23–30. Gaertner, S. L., Rust, M. C., Dovidio, J. F., Bachman, B. A., & Anastasio, P. A. (1996). The Contact Hypothesis: The role of a common ingroup identity on reducing intergroup bias among majority and minority group members. In J. L. Nye & A. M. Brower (Eds.), What’s social about social cognition? (pp. 230–360). Newbury Park, CA: Sage. Gallup (2002). Poll topics & trends: Race relations. Washington, DC: The Gallup Organization. http://www.gallup.com/poll/topics/race.asp. Greenwald, A., & Banaji, M. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102, 4–27. Greenwald, A., McGhee, D., & Schwartz, J. (1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74, 1464–1480. Hamilton, D. L., & Trolier, T. K. (1986). Stereotypes and stereotyping: An overview of the cognitive approach. In J. F. Dovidio & S. L. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 127–163). Orlando, FL: Academic Press. Hebl, M. R., Foster, J. B., Mannix, L. M., & Dovidio, J. F. (2002). Formal and interpersonal discrimination. A field study of bias toward homosexual applicants. Personality and Social Psychology Bulletin, 28, 815–825. Hewstone, M. (1996). Contact and categorization: Social psychological interventions to change intergroup relations. In C. N. Macrae, M. Hewstone, & C. Stangor (Eds.), Foundations of stereotypes and stereotyping (pp. 323–368). New York: Guilford.

48

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

Hodson, G., Dovidio, J. F., & Gaertner, S. L. (2002). Processes in racial discrimination: Differential weighting of conflicting information. Personality and Social Psychology Bulletin, 28, 460–471. Houlette, M., Gaertner, S. L., Johnson, K. M., Banker, B. S., Riek, B. M., & Dovidio, J. F. (2004). Developing a more inclusive social identity: An elementary school intervention. Journal of Social Issues, 60, 35–55. Howard, J. M., & Rothbart, M. (1980). Social categorization for in-group and out-group behavior. Journal of Personality and Social Psychology, 38, 301–310. Huo, Y. J., Smith, H. H., Tyler, T. R., & Lind, A. E. (1996). Superordinate identification, subgroup identification, and justice concerns: Is separatism the problem. Is assimilation the answer? Psychological Science, 7, 40–45. Hyers, L. L., & Swim, J. K. (1998). A comparison of the experiences of dominant and minority group members during an intergroup encounter. Group Processes and Intergroup Relations, 1, 143–163. Insko, C. A., Schopler, J., Gaertner, L., Wildschut, T., Kozar, R., Pinter, B., Finkel, E. J., Brazil, D. M., Cecil, C. L., & Montoya, M. R. (2001). Interindividual-intergroup discontinuity reduction through the anticipation of future interaction. Journal of Personality and Social Psychology, 80, 95–111. Jenkins, R. R. (2001). The health of minority children in the year 2000: The role of government programs in improving the health status of America’s children. In N. J. Smelser, W. J. Wilson, & F. F. Mitchell (Eds.), Racial trends and their consequences (Vol. 2, pp. 351–370). Washington, DC: National Academy Press. Johnson, J. D., Whitestone, E., Jackson, L. A., & Gatto, L. (1995). Justice is still not colorblind: Differential racial effects of exposure to inadmissible evidence. Personality and Social Psychology Bulletin, 21, 893–898. Jones, E. E., & Harris, V. A. (1967). The attribution of attitudes. Journal of Experimental Social Psychology, 3, 1–24. Jones, J. M. (1997). Prejudice and racism (2nd ed.). New York: McGraw-Hill. Karpinski, A., & Hilton, J. L. (2000). Attitudes and the Implicit Association Test. Journal of Personality and Social Psychology, 81, 774–788. Katz, I., & Hass, R. G. (1988). Racial ambivalence and value conflict: Correlational and priming studies of dual cognitive structures. Journal of Personality and Social Psychology, 55, 893–905. Katz, I., Wackenhut, J., & Hass, R. G. (1986). Racial ambivalence, value duality, and behavior. In J. F. Dovidio & S. L. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 35–59). Orlando, FL: Academic Press. Kawakami, K., Dovidio, J. F., Moll, J., Hermsen, S., & Russin, A. (2000). Just say no (to stereotyping): Effects of training in trait negation on stereotype activation. Journal of Personality and Social Psychology, 78, 871–888. Kawakami, K., Spears, R., & Dovidio, J. F. (2002). Disinhibition of stereotyping: Context, prejudice and target characteristics. European Journal of Social Psychology, 32, 517–530. Kelman, H. C. (2001). Reflections on social and psychological processes of legitimization and delegitimization. In J. T. Jost & B. Major (Eds.), The psychology of legitimacy: Emerging perspectives on ideology, justice, and intergroup relations (pp. 54–73). New York: Cambridge University Press. Kleinpenning, G., & Hagendoorn, L. (1993). Forms of racism and the cumulative dimension. Social Psychology Quarterly, 56, 21–36. Knight, J. L., Guiliano, T. A., & Sanchez-Ross, M. G. (2001). Famous or infamous? The influence of celebrity status and race on perceptions of responsibility for rape Basic and Applied Social Psychology, 23, 183–190.

AVERSIVE RACISM

49

Kovel, J. (1970). White racism: A psychohistory. New York: Pantheon. Kravitz, D. A. (1995). Attitudes toward affirmative action plans directed at blacks: Effects of plan and individual differences. Journal of Applied Social Psychology, 25, 2192–2220. Krieger, L. H. (1995). The content of our categories: A cognitive bias approach to discrimination and equal employment opportunity. Stanford Law Review, 47, 1161–1248. Krieger, L. H. (1998). Civil rights perestroika: Intergroup relations after affirmative action. California Law Review, 86, 1251–1333. Lane, J. D., & Wegner, D. M. (1995). The cognitive consequences of secrecy. Journal of Personality and Social Psychology, 69, 237–253. Lind, A. E., Kurtz, S., Musante, L., Walker, L., & Thibaut, J. W. (1980). Procedure and outcome effects on reactions to adjudicated resolution of conflicts of interest. Journal of Personality and Social Psychology, 39, 643–653. Massey, D. S. (2001). Residential segregation and neighborhood conditions in U. S. metropolitan areas. In N. J. Smelser, W. J. Wilson, & F. Mitchell (Eds.), Racial trends and their consequences (Vol. 1, pp. 391–434). Washington, DC: National Academy Press. McConahay, J. B. (1983). Modern racism and modern sexism: The effects of race, racial attitudes, and context on simulated hiring decisions. Personality and Social Psychology Bulletin, 9, 551–558. McConahay, J. B. (1986). Modern racism, ambivalence, and the modern racism scale. In J. F. Dovidio & S. L. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 91–125). Orlando, FL: Academic Press. McConnell, A. R., & Leibold, J. M. (2001). Relations among the Implicit Association Test, discriminatory behavior, and explicit measures of racial attitudes. Journal of Experimental Social Psychology, 37, 435–442. Miller, N. (2002). Personalization and the promise of contact theory. Journal of Social Issues, 58, 387–410. Monteith, M. J., Sherman, J., & Devine, P. G. (1998). Suppression as a stereotype control strategy. Personality and Social Psychology Review, 1, 63–82. Monteith, M. J., & Voils, C. (1998). Proneness to prejudiced responses: Toward understanding the authenticity of self-reported discrepancies. Journal of Personality and Social Psychology, 75, 901–916. Moskowitz, G. B., Gollwitzer, P. M., Wasel, W., & Schaal, B. (1999). Preconscious control of stereotype activation through chronic egalitarian goals. Journal of Personality and Social Psychology, 77, 167–184. Murphy-Berman, V. A., Berman, J. J., & Campbell, E. (1998). Factors affecting health-care allocation decisions: A case of aversive racism? Journal of Applied Social Psychology, 28, 2239–2253. Murrell, A. J., Dietz-Uhler, B. L., Dovidio, J. F., Gaertner, S. L., & Drout, C. (1994). Aversive racism and resistance to affirmative action: Perceptions of justice are not necessarily color blind. Basic and Applied Social Psychology, 15, 71–86. Myrdal, G. (1944). An American dilemma: The Negro problem and modern democracy. New York: Harper. Nail, P. R., Harton, H. C., & Decker, B. P. (2003). Political orientation and modern versus aversive racism: Tests of Dovidio and Gaertner’s (1998) Integrated Model. Journal of Personality and Social Psychology, 84, 754–770. Nier, J. A. (2003). How dissociated are implicit and explicit racial attitudes? A bogus pipeline approach. Unpublished manuscript, Department of Psychology, Connecticut College, New London, CT.

50

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

Nier, J. A., Gaertner, S. L., Dovidio, J. F., Banker, B. S., & Ward, C. M. (2001). Changing interracial evaluations and behavior: The effects of a common group identity. Group Processes and Intergroup Relations, 4, 299–316. Operario, D., & Fiske, S. T. (1998). Racism equals power plus prejudice: A social psychological equation for racial oppression. In J. Eberhardt & S. T. Fiske (Eds.), Confronting racism: The problem and the response (pp. 33–53). Newbury Park, CA: Sage. Otten, S., & Moskowitz, G. B. (2000). Evidence for implicit evaluative in-group bias: Affectbased spontaneous trait inference in a minimal group paradigm. Journal of Experimental Social Psychology, 36, 77–89. Park, B., & Rothbart, M. (1982). Perception of out-group homogeneity and levels of social categorization: Memory for the subordinate attributes of in-group and out-group members. Journal of Personality and Social Psychology, 42, 1051–1068. Pettigrew, T. F. (1959). Regional differences in anti-Negro prejudice. Journal of Abnormal and Social Psychology, 59, 28–36. Pettigrew, T. F. (1998). Intergroup Contact Theory. Annual Review of Psychology, 49, 65–85. Pettigrew, T. F., & Meertens, R. W. (1995). Subtle and blatant prejudice in Western Europe. European Journal of Social Psychology, 25, 57–76. Phelps, E. A., O’Connor, K. J., Cunningham, W. A., Funayama, E. S., Gatenby, J. C., Gore, J. C., & Banaji, M. R. (2000). Performance on indirect measures of race evaluation predicts amygdala activation. Journal of Cognitive Neuroscience, 12, 729–738. Regents of the University of California v. Bakke (1978). U. S. Law Weekly, 46, 4896. Richeson, J., & Shelton, J. N. (2003). When prejudice does not pay: Effects of interracial contact on executive function. Psychological Science, 14, 287–290. Robinson, P. H., & Darley, J. M. (1995). Justice, liability, and blame. Boulder, CO: Westview Press. Rogers, R. W., & Prentice-Dunn, S. (1981). Deindividuation and anger-mediated interracial aggression: Unmasking regressive racism. Journal of Personality and Social Psychology, 41, 63–73. Rosenfeld, R. A. (1998). Race and sex differences in career dynamics. In K. A. Myers & C. D. Anderson (Eds.), Feminist foundations: Toward transforming sociology. Gender and society readers (Vol. 3, pp. 113–151). Thousand Oaks, CA: Sage. Rudman, L. A., & Kilianski, S. E. (2000). Implicit and explicit attitudes toward female authority. Personality and Social Psychology Bulletin, 26, 1315–1328. Sawires, J. N., & Peacock, M. J. (2000). Symbolic racism and voting behavior on Proposition 209. Journal of Applied Social Psychology, 30, 2093–2099. Schaller, M., Conway, L. G. III, & Tanchuk, T. L. (2002). Selective pressures on the once and future contents of ethnic stereotypes: Effects of the communicability of traits. Journal of Personality and Social Psychology, 82, 861–877. Schuman, H., Steeh, C., Bobo, L., & Krysan, M. (1997). Racial attitudes in America: Trends and interpretations. Cambridge, MA: Harvard University Press. Sears, D. O., Henry, P. J., & Kosterman, R. (2000). Egalitarian values and contemporary racial politics. In D. O. Sears, J. Sidanius, & L. Bobo (Eds.), Racialized politics: The debate about racism in America (pp. 75–117). Chicago, IL: University of Chicago Press. Shelton, J. N. (2000). A reconceptualization of how we study issues of racial prejudice. Personality and Social Psychology Review, 4, 374–390. Sherif, M. (1966). Group conflict and cooperation: Their social psychology. London: Routledge and Kegan Paul. Sidanius, J., Levin, S., & Pratto, F. (1998). Hierarchical group relations. Institutional terror, and the dynamics of the criminal justice system. In J. Eberhardt & S. T. Fiske (Eds.), Confronting racism: The problem and the response (pp. 136–165). Newbury Park, CA: Sage.

AVERSIVE RACISM

51

Skrentny, J. D. (1996). The ironies of affirmative action: Politics, culture, and justice in America. Chicago, IL: University of Chicago Press. Smedley, B. D., Stith, A. Y., & Nelson, A. R. (Eds.) (2003). Unequal treatment: Confronting racial and ethnic disparities in health care. Washington, DC: National Academy Press. Smith, H. J., & Tyler, T. R. (1996). Justice and power: When will justice concerns encourage the advantaged to support policies which redistribute economic resources and the disadvantaged to willingly obey the law? European Journal of Social Psychology, 26, 171–200. Sommers, S. R., & Ellsworth, P. C. (2000). Race in the courtroom: Perceptions of guilt and dispositional attributions. Personality and Social Psychology Bulletin, 26, 1367–1379. Sommers, S. R., & Ellsworth, P. C. (2001). White juror bias: An investigation of prejudice against black defendants in the American courtroom. Psychology, Public Policy, and Law, 7, 201–229. Son Hing, L. S., Bobocel, D. R., & Zanna, M. P. (2002). Meritocracy and opposition to affirmative action: Making concessions in the face of discrimination. Journal of Personality and Social Psychology, 83, 493–509. Son Hing, L. S., Chung-Yan, G., Grunfeld, R., Robichaud, L., & Zanna, M. P. (2004). Exploring the discrepancy between implicit and explicit prejudice: A test of aversive racism theory. In J. P. Forgas, K. Williams, & W. von Hippel (Eds.). Social Motivation: Conscious and unconscious processes (in press). Son Hing, L. S., Li, W., & Zanna, M. P. (2002). Inducing hypocrisy to reduce prejudicial responses among aversive racists. Journal of Experimental Social Psychology, 38, 71–78. Stephan, W. G., & Stephan, C. W. (1985). Intergroup anxiety. Journal of Social Issues, 41(3), 157–175. Stephan, W. G., & Stephan, C. W. (2001). Improving intergroup relations. Thousand Oaks, CA: Sage. Supreme Court of the United States (2003). Grutter v. Bollinger et al., Certiorrari to the United States Court of Appeals for the Sixth Circuit, 539, U.S., 1–5. Tajfel, H. (1969). Cognitive aspects of prejudice. Journal of Social Issues, 25(4), 79–97. Tajfel, H. (1970). Experiments in intergroup discrimination. Scientific American, 223, 96–102. Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin & S. Worchel (Eds.), The social psychology of intergroup relations (pp. 33–48). Monterey, CA: Brooks/Cole. Towles-Schwen, T., & Fazio, R. H. (2001). The origins of racial attitudes: Correlates of childhood experiences. Personality and Social Psychology Bulletin, 27, 162–175. Turner, J. C. (1985). Social categorization and the self-concept: A social cognitive theory of group behavior. In E. J. Lawler (Ed.), Advances in group processes (Vol. 2, pp. 77–122). Greenwich, CT: JAI Press. Turner, M. E., & Pratkanis, A. R. (1994). Affirmative action as help: A review of recipient reactions to preferential selection and affirmative action. Basic and Applied Social Psychology, 15, 43–69. U.S. Merit Systems Protection Board (1997). Fair and equitable treatment: A progress report on minority employment in the federal government. Washington, DC: Office of Policy and Evaluation. Vorauer, J. D., & Kumhyr, S. M. (2001). Is this about you or me? Self- versus other-directed judgments and feelings in response to intergroup interaction Personality and Social Psychology Bulletin, 27, 706–719. Whaley, A. L. (1998). Racism in the provision of mental health services: A social-cognitive analysis. American Journal of Orthopsychiatry, 68, 47–57. Wegner, D. (1994). Ironic processes of mental control. Psychological Review, 101, 34–52.

52

JOHN F. DOVIDIO AND SAMUEL L. GAERTNER

Weiner, B. (1995). Judgments of responsibility: A foundation for a theory of social conduct. New York: Guilford. Wilder, D. A. (1981). Perceiving persons as a group: Categorization and intergroup relations. In D. L. Hamilton (Ed.), Cognitive processes in stereotyping and intergroup behavior (pp. 213–257). Hillsdale, NJ: Erlbaum. Williams, R. M., Jr. (1947). The reduction of intergroup tensions. New York: Social Science Research Council. Wilson, T. D., Lindsey, S., & Schooler, T. Y. (2000). A model of dual attitudes. Psychological Review, 107, 101–126. Wittenbrink, B., Judd, C., & Park, B. (1997). Evidence for racial prejudice at the implicit level and its relationship with questionnaire measures. Journal of Personality and Social Psychology, 72, 262–274. Worchel, S., Rothgerber, H., Day, E. A., Hart, D., & Butemeyer, J. (1998). Social identity and individual productivity with groups. British Journal of Social Psychology, 37, 389–413. Word, C. O., Zanna, M. P., & Cooper, J. (1974). The nonverbal mediation of self-fulfilling prophecies in interracial interaction. Journal of Experimental Social Psychology, 10, 109–120. Wright, S. C., Aron, A., McLaughlin-Volpe, T., & Ropp, S. A. (1997). The extended contact effect: Knowledge of cross-group friendships and prejudice. Journal of Personality and Social Psychology, 73, 73–90. Wyer, N., & Hamilton, D. (1998). The balance between excitation and inhibition in stereotype use. In Wyer (Ed.), Stereotype activation and inhibition. Advances in social cognition (Vol. 11, pp. 227–242). Mahwah, NJ: Erlbaum.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

Eliot R. Smith Gu¨n R. Semin

We propose a new integration of social psychology and situated cognition, which we term socially situated cognition (SSC). This new approach rests directly on recent developments in psychology and cognitive science captured by the label situated cognition, but it also has roots going back to William James and Frederick Bartlett. The approach involves a shift in the guiding metaphor regarding cognition and action, from computation to biology, and highlights four core assumptions that are common to social psychology and the situated cognition perspective. (1) Cognition is for the adaptive regulation of action, and mental representations are action oriented. (2) Cognition is embodied, drawing on our sensorimotor abilities and environments as well as our brains. (3) Cognition and action are the emergent outcome of dynamic processes of interaction between an agent and an environment. (4) Cognition is distributed across brains and the environment (e.g., through the use of tools) and across social agents (e.g., when information is discussed and evaluated in groups). With regard to each of these themes, we review and integrate relevant social psychological research and suggest ways in which the theme can be advanced by rethinking current assumptions. Our overall goals are to make social psychology part of the interdisciplinary integration emerging around the concept of situated cognition, and to advance these four themes as high-level conceptual principles that can organize seemingly disparate areas of research and theory within social psychology itself. For more than a century, social psychology has developed as the scientific discipline devoted to the study of social cognition and social behavior: of attitudes, social influence processes, person perception, group stereotypes 53 ADVANCES IN EXPERIMENTAL SOCIAL PSYCHOLOGY, VOL. 36

Copyright 2004, Elsevier Inc. All rights reserved. 0065-2601/04 $35.00

54

¨ N R. SEMIN ELIOT R. SMITH AND GU

and prejudice, interaction in groups, and personal relationships, among other topics. Particularly since the 1960s, social psychology has been dominated by experimental methods and by theoretical approaches emphasizing cognitive processes (Devine, Hamilton, & Ostrom, 1994; Jones, 1985). In parallel with the rise of the cognitive perspective throughout psychology, social psychologists focused on analyses of mental representation and process as pathways to understanding social psychological phenomena. The cognitive approach led to the development of robust, sophisticated, and cumulative bodies of theory concerning such issues as the nature of mental representations, the impact of accessible representations on judgments, and the factors aVecting people’s use of simple heuristic cues versus more systematic information processing. In more recent years, social psychologists have broadened their focus beyond cognitive processes by directing increased attention to issues of motivation and aVect, personal relationships, group memberships, and cultural diVerences as they have worked toward a more complete understanding of social behavior. The emphasis is now on motivational constraints and situational eVects on cognition—with motives and situations taken not just as additional information to be processed, but as fundamental regulators of cognition. Paralleling these developments within social psychology, a powerful new intellectual movement has arisen across many areas of psychology and cognitive science, which has been termed situated cognition (e.g., Agre, 1997; Barsalou, 1999b; Brooks, 1999; Chapman, 1991; Clark, 1997; Greeno, 1998; Hendriks-Jansen, 1996; Norman, 1993; Steels & Brooks, 1995; Yeh & Barsalou, 2000). A convergent but nonoverlapping school of thought that emphasizes the significance of situated action and cognition is found in sociocultural psychology (e.g., Cole, 1996; Wertsch, 1998). While the sociocultural school traces its intellectual roots to Vygotski’s work (e.g., 1962/ 1986), the fundamental ideas of the situated cognition movement go back in various forms to Dewey, Mead, and particularly William James and Frederick Bartlett. It is no accident that these same individuals are also frequently named as important forebears of social psychology, for the central thesis of this chapter is that the fundamental assumptions of situated cognition are quite compatible with those of social psychology—so much so that the term socially situated cognition (SSC) seems quite appropriate. Despite this fundamental conceptual similarity, for which we will argue in detail, situated cognition and social psychology have had little or no contact, although neighboring subdisciplines (such as developmental, cognitive, and educational psychology) have been deeply aVected by the situated cognition perspective. We believe that vigorous attempts at integration and mutual recognition—from both sides of this conceptual divide—will pay rich dividends in future theoretical and empirical progress.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

55

Situated cognition theorists view cognition as an adaptive process that emerges from the interaction between an agent and the world, both physical and social. This is an idea that can be found in Mead’s (1934) focus on ‘‘symbolic social processes’’ and is fundamental to the symbolic interactionist viewpoint, which makes ‘‘joint action’’ (Blumer, 1969) a central focus. Today the idea is continued in sociocultural psychology, an intellectual tradition strongly influenced by Vygotsky (e.g., 1978) that emphasizes the situated action level of analysis to account for the generation of meaningful activity in social contexts (Cole, 1966; Lave & Wenger, 1991; Wertsch, 1998). The same idea of cognition as a situated interactive process is also found in other recent thinking in cognitive and developmental psychology and cognitive science, disciplines that are conceptually closer to mainstream experimental social psychology than is the sociocultural school (e.g., Barsalou, 1999b; Clancey, 1995, 1997a; Clark, 1997, 1999b; Greeno, 1998; Kirshner & Whitson, 1997; Norman, 1993; Yeh & Barsalou, 2000). An early appearance of the new perspective was in the area of robotics within artificial intelligence (AI), where researchers seeking to build robots that interact with the world ran into the limitations of the traditional mentalist approach to cognition and sought a workable alternative (Agre & Chapman, 1990; Brooks, 1991). The movement then spread: ‘‘What originally seemed nothing more than a cute idea turned out to have profound ramifications and changed the entire research disciplines of artificial intelligence and cognitive science. It is currently beginning to exert its influence on psychology, neurobiology, and ethology as well as engineering’’ (Pfeifer & Scheier, 1999, p. xii). Although that influence has not yet touched the mainstream of social psychology, we believe that the central assumptions of situated cognition are largely compatible with those of social psychology. Our discussion of these central assumptions serves the two major goals of this chapter. First, there has been a relative mutual neglect and ignorance between workers in social psychology and situated cognition over the past decade or so. We will attempt to show that social psychology deals directly with the most central themes of situated cognition, and therefore belongs in the emerging interdisciplinary integration around those ideas—in many ways, at its very heart. Second, we will argue that the key themes of situated cognition are at the same time the most fundamental themes of social psychology as well. The themes can therefore conceptually bring together and interrelate seemingly disparate and unrelated bodies of research and topic areas within social psychology, as well as pointing toward new directions for conceptual and empirical exploration. Thus, the emerging interdisciplinary integration carries the potential for theoretical growth and development within social psychology. The first and most central postulate of situated cognition is that the function of cognition is the control of adaptive action. The insight that

56

¨ N R. SEMIN ELIOT R. SMITH AND GU

cognition is for action dates back to William James (1890) but has often been overlooked in more recent years, especially in the heyday of the cognitive or information-processing approach (see Fiske, 1992). However, its truth is unquestionable (Barsalou, 1999b; Dewey, 1929/1958; Mead, 1934). The second theme is that cognition depends on our bodily properties and environments, as well as our brains. Therefore cognition must be regarded as embodied, dependent on and intertwined with (not apart from) sensorimotor processes. Third, adaptive action must be closely tuned to the immediate environment, and therefore cognition must be situated, interactive, and flexible. Cognition emerges from moment-by-moment interaction with the environment rather than proceeding in an autonomous, invariant, contextfree fashion. And crucially, the relevant environment is always social, as well as being defined by human artifacts, physical spaces, and tasks. Fourth, cognition is distributed—not encapsulated within our brains, but extended and empowered by tools (both physical and conceptual, crucially including language) and by social resources such as other people and groups. These four themes represent the organizing framework of this chapter. With regard to each of them, we will first show that social psychologists have produced bodies of relevant research and theory—our field’s potential exports, if we may be so bold, to others interested in situated cognition. At the same time, for each theme we will argue that in some respects social psychology has not yet fully appreciated all the insights of situated cognition. In many cases, these point to new areas for theoretical and empirical exploration or suggest modifications of some typical social psychological assumptions. Despite its conceptual kinship to social psychology, the situated cognition approach calls for rethinking several assumptions that have been prevalent within social psychology. Most basic is a shift in the guiding metaphor for the analysis of cognition and action, from computation to biology. The computational or information-processing framework analyzes cognition in terms of representational structures and algorithmic processes, drawing on the fundamental concepts and principles of computer science (Marr, 1982; Newell & Simon, 1972; Vera & Simon, 1993). This framework is evident in most prominent theoretical models within social psychology, as they postulate the construction of schemas or associatively linked structures to represent stimulus information, which are then stored in memory and reactivated to aVect judgments and social behavior (e.g., Wyer & Srull, 1989). In contrast, the starting point for the biological metaphor is that all cognition and action constitute an adaptive regulatory process that ultimately serves survival needs. The biological metaphor also invites us to consider cognition and action as embodied—constrained and directed by the nature of our bodies. Because of the importance of adaptation to specific and

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

57

varying situations, cognition and action are considered as the emergent outcome of dynamic processes of interaction between an agent and an environment. Finally, SSC emphasizes that cognition is not purely an inner process within an individual brain but is distributed across the environment (e.g., through the use of tools and artifacts) and across social agents (e.g., when information is shared and evaluated in groups; Caporael, 1997). This new approach does not require a complete rejection of computational ideas such as representations, but appears likely to fundamentally transform our understanding of their nature and their role in the overall adaptive process (Clark, 1997). The biological metaphor is not meant to direct social psychologists toward neuroscience as a substitute for psychological theory nor toward the types of analysis characteristic of some strands of current-day evolutionary psychology (e.g., analyses of mate choice or of sex diVerences in social behavior). Both neuroscience and evolutionary psychology are valuable levels of analysis, but our focus is on the psychological level. Within that level the biological metaphor calls attention to the relation of all cognition and action to fundamental self-regulatory processes and to the demands of adaptive action. This is an evolutionary approach in a broad sense, the type of ‘‘evolutionary view of human cognition inextricably embedded in the social structural context in which it occurs’’ that Caporael (1997, p. 277) has advocated. A fuller understanding of both cognition and action will result when we integrate the implications of the biological approach and the inherently situated nature of social action, with the strengths of the sophisticated, process-oriented cognitive social psychological research tradition of the last couple of decades.

I. Cognition is for Action Cognitive scientist Stan Franklin concluded from his analysis of biological and artificial minds: ‘‘The overriding task of Mind is to produce the next action (minds are the control structures of autonomous agents)’’ (1995, p. xx). The first and primary principle of the socially situated cognition perspective is that cognition evolved for the control of adaptive action, not for its own sake. The fundamental evolutionary demands on cognition are the organism’s survival and reproduction, which (for humans) always take place in a social context (Caporael, 1997; Fiske, 1992; Simpson & Kenrick, 1997). In the situated cognition approach, intelligence is not identified with detached thought but with adaptively successful interaction with other agents and with the world. Mind is viewed in action-oriented terms, as

58

¨ N R. SEMIN ELIOT R. SMITH AND GU

containing ‘‘inner structures that act as operators upon the world via their role in determining actions’’ (Clark, 1997, p. 47). Theorists in this perspective vary in the extent to which they reject traditional conceptions from the computation approach, such as mental representations (Agre, 1997; Beer, 1995). Some follow Gibson (1966) in at least rhetorically denying that representations exist or are useful concepts for analyzing cognition (e.g., Brooks, 1986; Thelen & Smith, 1994). Others acknowledge the existence of representations but see them as diVering fundamentally from the text-like abstractions portrayed by classic cognitive science (e.g., Clark, 1997). In contrast, the information-processing perspective emphasizes cognition for its own sake, often losing sight of its links to action and its dynamic and adaptive functions. The prevailing implicit model is of cognition as reception: the person is treated as a passive observer of ongoing stimuli, much like a couch potato watching television. (This is certainly the mode in which participants are placed in most social cognition studies within social psychology.) Cognition becomes the construction and manipulation of inner representations, rather than real interaction with the world, conversation, and so forth. As an illustration of the close links between cognition and adaptive action, consider one of the early classic studies of neuroscience. In the 1950s, Warren McCulloch and his colleagues analyzed the information transmitted along the optic nerve from the frog’s eye to the brain (in McCulloch, 1988). What they found was not, as might have been expected, a topologically organized map of visual space, with diVerent nerve fibers responding to light impinging on diVerent retinal areas. Instead, certain neurons responded optimally to local convexities—to the light patterns produced by small objects on an underlying surface. Other neurons responded to a sudden overall dimming of the illumination. The researchers termed these types of fibers bug detectors and hawk detectors, respectively, labels that underline their action-oriented properties. Thus, very early in the visual system (in the retina itself ) the frog computes not just a maplike image of the surroundings that is neutral with respect to action, but specific aVordances for adaptive action—for extending the tongue to eat a bug or for jumping to escape a predator looming overhead.

A. SOCIAL PSYCHOLOGICAL CONTRIBUTIONS TO THE THEME In many respects, social psychology has long accepted and built on the idea that cognition is for action in a social world. As a result, the field has developed several bodies of theory and research that illustrate the manner in

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

59

which cognition is intrinsically shaped by demands for action (Fiske, 1992). We briefly describe this work under several headings: relations of motivation and cognition, the methods by which the demands of action shape cognition and perception, as well as mental representation, and findings illustrating the intimate connections between cognition and overt behavior. 1. Relations of Motivation and Emotion to Cognition If cognition is for action, then cognition, emotion, and motivation all are parts of an overall self-regulatory process. Despite the conventional opposition of ‘‘heart versus mind,’’ which can be traced back to ancient Greek philosophy, an emotionless cognitive agent would be not smart and rational, but totally nonfunctional (Clark, 1997; Picard, 1997; Sloman, 1997). For example, patients with brain damage aVecting their emotional systems are drastically impaired in rational decision making (Damasio, 1994) despite verbal abilities and tested ‘‘intelligence’’ that remain relatively unimpaired. The social psychological literature contains many additional examples of the ways that motivation and aVect regulate cognition, including eVects of mood on memory and cognitive processing (Singer & Salovey, 1988). Emotion, cognition, and motivation are all equally functionally indispensable and inseparable parts of a self-regulatory system, subserving adaptive action. Specifically, there have been numerous empirical demonstrations of pervasive eVects of motivation on cognitive processes. The very definitions of traits, the basic concepts that make up our impressions of people and groups, are altered in self-serving ways by our own perceived standing on those traits (Dunning & Cohen, 1992). Impressions we form of close relationship partners are aVected by motives to perceive our relationships in a positive light (Murray et al., 2000). Stereotype activation or suppression may be governed by our wish to view a member of a stereotyped group in a positive or negative light (for example, because they have praised or criticized us; Sinclair & Kunda, 1999). Several other examples could equally well be cited (Brewer, 1991; Brewer & Harasty, 1996; Gollwitzer & Bargh, 1996; Higgins & Sorrentino, 1990; Sorrentino & Higgins, 1986). Summarizing all of this evidence into a set of coherent theoretical principles is still a work in progress. However, Kruglanski (1996) has argued that cognition and motivation are in large measure two aspects of the same thing, viewed through diVerent lenses and described in diVerent languages. Kruglanski outlines how motivational constructs are useful for understanding the initiation and termination of processing. For example, a person may rest content with a quick heuristic response that is self-enhancing or

60

¨ N R. SEMIN ELIOT R. SMITH AND GU

otherwise motivationally congenial, but continue with more extensive processing if the initial response appears less desirable. In other words, virtually all cognitive activity is performed (or not performed) for motivational reasons—little ‘‘motivation-free’’ cognition may exist. Within the course of processing, Kruglanski (1996) argues that motivational eVects operate at every stage, including information encoding, memory retrieval, information integration, and judgment formation. But conversely, not only does motivation aVect cognition; goals and motives themselves are mentally represented as knowledge structures and operate according to the basic principles of cognition (e.g., spread of activation, construct accessibility eVects from recent use). Thus, not only does cognition intrinsically have a motivational aspect, but motivation necessarily has a cognitive aspect (Kruglanski, 1996; Sorrentino & Higgins, 1986). 2. Pragmatic Concerns Involving Action Shape Perception and Cognition a. Social Perceivers are ‘‘Good enough’’. A prominent research tradition in the 1970s and 1980s emphasized the susceptibility of social perceivers to ‘‘errors and biases’’ (e.g., Nisbett & Ross, 1980). Current thinking is much diVerent, as Fiske (1992) has noted. We now consider social perceivers as strongly driven by pragmatic concerns, and as striving for and generally attaining suYcient accuracy to suit their everyday needs for adaptive action. For example, accuracy in perceiving others is generally found to be surprisingly high (Kenny & Albright, 1987), even when based on just a few minutes of data (Ambady et al., 2000; Krauss, Freyberg, & Morsella, 2002). Other researchers have noted that accuracy in perceiving others may not even be the most important goal in social interaction, compared to goals such as making the interaction proceed smoothly and predictably (e.g., Snyder & Cantor, 1998). Other aspects of social perception that were initially viewed as ‘‘biases,’’ such as greater sensitivity to negative information than to positive, are now also conceptualized as adaptive and pragmatically useful (Peeters, 1991; Skowronski & Carlston, 1989; Wentura et al., 2000). Even perceivers’ use of cognitive shortcuts and heuristics, once derided as lazy and errorprone, is now more often viewed as adaptive (e.g., Chaiken, Liberman, & Eagly, 1989). In many ways, then, social perceivers can be viewed as functionally attuned to pragmatic concerns—to the demands of action—in the ways that they seek and process information. b. Time Pressures and Cognitive Processes. One of the main adaptive constraints on cognition is that we often must generate behavior under conditions of time pressure, when leisurely consideration is impossible. Carrying on a dyadic conversation, playing basketball, or participating in a group discussion are examples of social cognition under time pressure.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

61

Social psychologists have developed dual-process theories in many domains that assume that people possess a repertoire of simple processing short-cuts (‘‘heuristics’’) on which they can rely when the time or inclination for extensive processing is absent (see Smith & DeCoster, 2000 for a review). Theorists assume that in many real-life situations a good strategy is to trade oV some increased accuracy, which might be attained by extensive information gathering and eVortful consideration, for the sake of increased speed and eYciency (e.g., Chaiken, Liberman, & Eagly, 1989). As attitude researchers have observed, if we stopped to carefully consider our attitude toward every product in a supermarket, a simple shopping trip would take days; this is a circumstance in which readily activated attitudes toward a limited number of products dramatically ease our task. These dual-process theories also emphasize another point: cognition is not always time pressured. Sometimes we can devote extensive consideration over hours, days, or weeks to developing a plan, elaborating an abstract idea, or making a decision. In fact, the ability to perform such detached or ‘‘oV-line’’ cognition may be one of the fundamental traits that sets human beings apart from other animals, whose cognition is probably much more tightly linked to current situations and tasks (Wilson, 2002). c. Action Relevance and Cognitive Style. Researchers, including Gollwitzer (1990, 1996), have found that the demands of a decision fundamentally aVect methods of processing information. For example, Gollwitzer (1990) has studied how people consider and weigh information about a decision they will make in the near future versus a decision made in the recent past. Prior to a decision, people adopt a ‘‘deliberative’’ mindset in which information is considered in a relatively open and unbiased fashion. Information seeking is a key goal. Once the decision is made, however, people switch to an ‘‘implemental’’ mindset in which information is processed with more biases (e.g., with an optimistic focus on the probability of the action’s success) and information seeking is considerably narrowed. Action relevance once again is found to aVect cognitive processes. d. Action Relevance and Person Perception. Consider a perceiver who receives information about another person or a social group in a context detached from action (because the others are fictitious, far away, long-dead historical characters, etc.). Now consider the perceiver receiving the same information about a person or group with whom he or she is about to meet and interact. What diVerences might this type of action relevance make in person perception processes? Neuberg and Fiske (1987) found that perceivers rely less on stereotypes and devote more eVort to forming individuated impressions when they expect to meet the target person. Similarly, Newtson and Enquist (1976) find that action relevance changes the way perceivers perceptually segment a videotaped behavioral sequence. And Carlston

62

¨ N R. SEMIN ELIOT R. SMITH AND GU

(1994), based on his AST model of person impressions, advanced hypotheses about the types of diVerences that action relevance (which he terms behavioral involvement) may create in impressions of others—producing relatively less emphasis on visual appearance and abstract trait information and more on aVective responses and concrete behaviors. Thus, in general, action relevance has a variety of eVects on person perception processes, all of which serve as reminders that, if perceiving is for doing, studies that put participants in a passive receptive or ‘‘TV-viewing’’ mode may not capture a full picture of social cognitive processes. e. Action Relevance and Memory. Even memory is profoundly aVected by action relevance. Two older but highly influential studies emphasize this point. Pichert and Anderson (1977) had people read a brief story about a walk through a house, from the perspective of either a prospective homebuyer or a prospective burglar. Some details in the story (e.g., a leaky roof ) were relevant to the homebuyer but not the burglar, and others (e.g., an expensive television) were relevant only to the burglar. When the participants recalled the story later, they were more likely to recall details that were relevant from their assigned perspective, in the sense of having some implications for future action. Zeigarnik (1927), in the early days of social psychology, conducted studies based on her observation that servers in a restaurant could often remember each diner’s order flawlessly until the party finished the meal and paid, but thereafter could not recall the slightest amount. Marsh, Hicks, and Bink (1998) have recently verified this finding with controlled experimental methods. The message of this work is that once information no longer has action relevance, it tends not to be remembered. f. Action Relevance and Communication. The SSC perspective holds that communication is action oriented, intended to accomplish something in the social world. This point is obvious considering many types of speech acts, such as requests, commands, persuasive attempts, or questions: the speaker is overtly trying to change the recipient’s beliefs, attitudes, or behaviors. Although the goal is not always so overtly clear, other types of speech also have implications for social action. Conversations often involve the exchange of information that will facilitate adaptive action in future situations. Even idle gossip involves the sharing and reinforcing of group norms, and often functions to strengthen the social bond within an in-group or dyad that is sharing the gossip. Research shows that speech acts are geared toward the attainment of situated goals in cooperative or competitive relationships (Semin, de Montes, & Valencia, 2003). Findings regarding ‘‘audience design’’ (Krauss & Fussel, 1996) or message modulation (Semin et al., 2003) support the conception of communication as socially situated action. In this conception, pragmatics (rather than syntactic structure) becomes the central focus of

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

63

concern with regard to language use. In contrast, the traditional perspective regards communication essentially as the neutral, unbiased translation or output of inner representations. The situated aspect of communication and its sensitivity to motivations, relationships, and other contextual influences are lost or deemphasized. (These issues will be reviewed more fully in the later section on cognition as distributed.) 3. Pragmatic Concerns and the Demands of Action Shape Mental Representation Because cognition is for the control of action, an agent must represent objects in the world in terms of their relations to the agent and their aVordances for action, not necessarily their ‘‘objective’’ qualities (Barsalou, 1999a; Gibson, 1966; Glenberg, 1997). Abstract, action-neutral representations simply describe a situation, leaving the implications for action implicit (‘‘Button A is connected to the switch that applies current to the motor;’’ a city map from the conventional overhead perspective). In contrast, actionoriented representations are exemplified by instructions or directions (‘‘Press button A to start the motor;’’ ‘‘go straight ahead three blocks then turn left.’’). As another example, a driver might construct and use a representation of ‘‘the car I am currently passing,’’ an example of the action-oriented or ‘‘deictic’’ representations that have been investigated by Clancey (1997a) and Agre and Chapman (1990). Deictic representations emphasize the object’s role in the agent’s current activity, and its features that are relevant to that activity, such as its relative position, speed, and so on, while neglecting specific individuating features (color, license plate number, etc.). Therefore deictic representations aVord a type of generalization or abstraction: ‘‘The car I am currently passing’’ will identify a diVerent car from one minute to the next. Deictic representations are not solely ‘‘in the head’’ but are patterns of interactive activity, including the agent and the external object. Perception of the object’s action-relevant qualities goes hand in hand with tuning the agent’s actual behavior to accomplish the current goal. Deictic representations are useful not for modeling the world but for letting the agent participate in the world (Chapman, 1991, p. 31). Connectionist networks are particularly well suited for learning and using deictic representations of this type (Chapman, 1991; Clark, 1997). Instead of constructing abstract inner representations of objects in the world, connectionist networks allow signals to flow through them, linking perception directly to adaptive action (Ballard et al., 1997). Action-oriented deictic representations contrast with traditional assumptions that representations are action neutral and identify specific individuated objects (that car is a blue Honda, license number 123ABC). The traditional perspective assumes that representations are abstract, static, and objective.

64

¨ N R. SEMIN ELIOT R. SMITH AND GU

The implicit model is textual representations (propositions) or equivalent symbolic structures (schemas, etc.). These descriptions are relatively ‘‘objective:’’ independent of the person’s viewpoint, current situation, and goals. Of course, people do seem to construct and use objective or action-neutral representations, at least at times. That is, one might be able to recall that the last car one passed was a blue Honda, despite the action-irrelevance of that information. The notion that cognition is for action does not require the extreme prediction that no action-irrelevant information is ever considered or represented (after all, it might become action-relevant in some unpredicted way in the future), only to the prediction that immediately action-relevant information is the most likely to be processed under normal circumstances. a. Attitudes. Attitudes have been considered perhaps the most characteristically social psychological concept (Allport, 1954). They have been conceptualized, however, perhaps too narrowly as purely mental representations. In reality attitudes are deictic representations. They capture relationships between the agent and the attitude object, which have implications for the way the agent perceives the object and acts toward it, as well as for the way the person thinks about it. Research shows, for example, that attitudes can be automatically activated to color perceptions and influence judgments and behaviors toward the objects (Fazio et al., 1986). Attitudes are a preeminent example of action-oriented representations, because they specify not just the nature of the object but how to behave toward it (e.g., whether to approach or avoid it). b. Person Impressions. Person impressions as well as attitudes are actionoriented representations, according to Carlston’s (1994) AST model. This model proposes that impressions of other people include several diVerent types of information that reflect contributions from four underlying representational systems: visual, verbal, aVective, and action. For example, an impression may include images of the person’s appearance (visual system) and traits believed to characterize the person (verbal system). Importantly, the impression also includes representations of the emotions felt toward the person (aVective system), and the perceiver’s own behaviors toward the person (action system), such as giving the person hugs or teasing him. Evidence of several types (reviewed in Carlston, 1994) supports the general postulates of the AST model, as well as the specific proposition that impressions contain aVective and behavioral responses. Thus, impressions are action-oriented representations. Impressions of others are also regulated by the type of relationships we have with them. We form impressions based not only on a target person’s characteristics, but also on the type of relationship we have with the person. Research supporting this point (e.g., Baldwin, 1992; Fiske et al., 1991; Holmes, 2000) suggests that relational interdependence and its action implications are integral to the way we represent people. For example, studies

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

65

demonstrate that other people with whom one has the same type of relationship (as categorized by Fiske’s model, 1991) tend to be confused with each other. This pattern of confusions holds independent of the targets’ personal characteristics, such as age, race, or personality traits (Fiske & Haslam, 1996). The confusion data suggest, then, that we mentally represent others in terms of the types of relationships that we have with those people and therefore the types of actions that we perform toward them, such as communal sharing, market-oriented bargaining, and so forth. 4. Close Connections Between Cognition and Action Perceiving another person performing a behavior or having the concept of a behavior activated through priming methods leads to the actual performance of the behavior (Bargh et al., 1996; Chartrand & Bargh, 1999). In other words, representations of behaviors are action oriented in the sense that activating those representations—even in the course of perceiving another person or for similar extraneous reasons—tends to lead to the production of the behavior (Prinz, 1990). We use our bodies in the process of perceiving behaviors. Another aspect of the connection between cognition and action has been explored by researchers working with one of the most characteristic and important theories of social psychology, cognitive dissonance theory (Festinger, 1957). The insight from this theory is that the connection between cognition and behavior is bidirectional. For example, disliking a particular toy is likely to lead a child to decline to play with that toy. Dissonance researchers have shown, conversely, that when the child does not play with an available toy for other reasons, it tends to lead to dislike for the toy. Similarly, having a particular position on an issue is likely to incline people to write an essay advocating that viewpoint. Dissonance researchers find that when people write an essay advocating a position they do not favor, under the correct circumstances it leads them to adopt the corresponding attitude (see Cooper & Fazio, 1984 for a review). All these findings illustrate the same fundamental principle, that cognition and behavior are so closely tied that it is diYcult to change one without changing the other.

B. TAKING THE THEME FURTHER As illustrated by the areas of research just reviewed, social psychologists generally accept the core idea that cognition is for action. Yet they may not have thought through and accepted all the implications of that idea. Here are some areas where the SSC perspective suggests we might be more

66

¨ N R. SEMIN ELIOT R. SMITH AND GU

radical in rethinking basic assumptions, or more creative in exploring their implications. 1. Language Comprehension Social psychologists generally recognize that person perception or attitude formation is ultimately driven by action relevance. But what of seemingly more objective, neutral, and passive activities such as text comprehension? Barsalou (1999a) has argued that even text comprehension is best understood as a tool for adaptive situated action. He makes an evolutionary argument, that types of social organization that stress passive ‘‘archival storage’’ of texts (such as formal education) are extremely recent developments—in contrast to the use of language to coordinate various cooperative forms of social action. Thus, in evolutionarily relevant contexts, situated action is much more important than archival storage of verbal information (Barsalou, 1999a, p. 65). The claim that language comprehension is for situated action does not mean that language always describes situations that are immediately present. Using a hypothetical group of hunters as an example, Barsalou (1999a) describes several types of relationship between language and concrete situations. Language can describe an immediate situation (as when a group of hunters discuss a herd of antelope while viewing them), or an absent situation that had been perceived earlier. Language can describe a situation that is similar (although not identical) to one that the listener has experienced (as when a hunter who has seen other antelope herds is told about the herd sighted today), or a situation that has not been experienced as a whole but is made up of familiar components. Language can also refer to future events (e.g., a discussion of plans for tomorrow’s hunt among individuals who have previously experienced similar hunts or components from other types of hunts). The basic point is that even when language does not refer to an immediately present situation, its use may still be in the service of preparation for situated action. Note that, as Caporael (1997) has argued in her ‘‘core configurations’’ model, this type of language use presupposes not only the small face-to-face groups currently engaged in talking about specific action, but also larger groups or ‘‘macrodemes’’ (size of at least several hundred—for example, the population of a small village) that are suYciently large to stabilize and standardize language (Caporael, 1997, p. 286). Does this argument hold true only for hypothetical prehistoric hunting bands? Barsalou admits that no rigorous research exists to answer the question, but notes: ‘‘Over the course of writing this commentary . . . I have been informally observing the conversations around me. So, for example, I have recently heard discussions of good places to shop for fresh produce, good restaurants with short waits, where best to park on campus, how best

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

67

to treat children with sinus infections, the best route for getting to work, neighborhoods where houses are still aVordable, how best to ensure that house remodelers finish on time, sports to watch during the NBA strike, and so on’’ (1999a, p. 68). Such conversations are generally for the support of situated action either in the immediate situation or a later one. One potential objection to this claim centers on the role of abstract concepts. When people discuss or think about truth, justice, or morality, are they still essentially geared toward eVective situated action? Although the point seems less clear than with concrete concepts that are more closely linked to immediate situations, the answer may still be ‘‘yes’’ (Barsalou, 1999b). Abstract concepts are equally important for situated action as are concrete concepts, even though they refer to diVerent aspects of situations. For example, the abstract property of ‘‘ownership’’ has major implications for what one can do with an object. LakoV and Johnson (1999) and Goldstone and Barsalou (1998) take this analysis further, showing how the most important abstract concepts are underlain by implicit metaphors based on perceptual-motor processes, the body, and situated action. Thus, it is at least arguable that everyday language comprehension can best be viewed as preparation for situated action, rather than for action-neutral ‘‘archival storage’’ (Barsalou, 1999a, but see Wilson, 2002). In line with this emphasis, Glenberg (1997; Glenberg & Robertson, 1999) has noted that much of the background knowledge on which text comprehension relies is experiential and perceptual, rather than itself being verbal and abstract. For social psychologists, one major implication of this line of thinking is methodological. When we give research participants textual materials, photographs, or other materials, we have to be aware of what types of action they anticipate as they process the information—we cannot assume that they simply strive to form representations that are as precise and accurate as possible given cognitive limitations. For example, participants will no doubt represent material diVerently if they think they will soon meet the people or groups described in the materials than if they do not anticipate a meeting, and will represent it diVerently if they expect a cooperative interaction with those persons than if they anticipate competition. Aspects of these predictions have been explored (e.g., Neuberg & Fiske, 1987), but the potential implications are far more general.

II. Cognition is Embodied The second major theme of socially situated cognition is that cognition is embodied, drawing on our physical bodies and particularly our sensorimotor capabilities as well as our brains. Our experiences of the world originate

68

¨ N R. SEMIN ELIOT R. SMITH AND GU

from bodily interactions. Thus, the evolved architectures of our body and brain constitute sources of regularity and constraint for cognition, aVect, motivation, and behavior. Embodied cognition is one of the central and converging issues of current interest in philosophy, cognitive science, psychology, robotics, and neuroscience (Almassy, Edelman, & Sporns, 1998; Ballard, Hayhoe, Pook, & Rao, 1997; Brooks, 1991; Clark, 1997, 1999b; Damasio, 1994; Glenberg, 1997; Goldstone & Barsalou, 1998; LakoV, 1987; LakoV & Johnson, 1999; Smith, Thelen, Titzer, & McLin, 1999; Varela, Thomson, & Rosch, 1991; Wilson, 2002). One aspect of embodiment is suggested by studies of how we perform natural tasks. Findings suggest that people routinely use bodily (sensorimotor) capabilities to support and enable cognition. For example, mental imagery draws on our sensory capabilities (Kosslyn, 1994), and is an important tool in many kinds of problem solving and text comprehension (Glenberg & Robertson, 1999; Johnson-Laird, 1983). Ballard et al. (1997) argue on the basis of their research that ‘‘at time scales of approximately one-third of a second, the momentary disposition of the body plays an essential role in the brain’s symbolic computations. The body’s movements at this time scale provide an essential link between processes underlying elemental perceptual events and those involved in symbol manipulation and the organization of complex behaviors’’ (p. 723). For example, eye movements can act as ‘‘pointers’’ in the computational sense, directing cognitive processes to analyze or act on specific objects and ignore others. This is not the classic type of ‘‘pointer’’ (an inner data structure), but a bodily action that simultaneously aVects perceptual processes. In this approach, cognition is treated as distributed across neural, bodily, and environmental features. Bodily movements, such as hand gestures, also play a role in cognitive processes including memory retrieval. For example, when people’s hand gestures are restricted, they are less able to retrieve words based on semantic cues or to recall earlier-presented words (Frick-Horbury, 2002; Frick-Horbury & Guttentag, 1998). Gestures appear to be used as a code that is represented with items in memory and can facilitate their retrieval. Another example of the embodiment principle comes from work on categorization—one of the fundamental enabling mechanisms of cognition. Recent work by Barsalou and colleagues (Barsalou, 1999b; Goldstone & Barsalou, 1998) has provided convincing evidence that categorization is grounded in perception, rather than being based on conceptual features that are abstracted from the concrete and perceptual. For example, perceptually driven impressions of similarity are key in many types of categorization, and perceptual features aVect performance even on many cognitive tasks that have no explicit perceptual components. In a diVerent way, Edelman’s (1987) work also underlines the importance of embodiment in the process

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

69

of category learning. He argues that forming discrete object categories based on real sensory inputs (as opposed to abstract feature vectors used in models or simulations) is extraordinarily diYcult. Sensory inputs are continually changing due to distance, orientation, lighting conditions, the agent’s own movements, etc. Instead of the near-impossible task of mapping these sensory inputs to inner representations, categorization can best be considered as a problem of sensorimotor coordination (Edelman, 1987; Pfeifer & Scheier, 1997). The agent’s movements, instead of adding additional variation to sensory inputs, actually allow the formation of cross-modal associations (e.g., between an object’s visual appearance and its tactile qualities), which are at the heart of concept formation (Pfeifer & Scheier, 1997; Thelen & Smith, 1994). Thus, categorization depends on the embodiment of cognition (i.e., on sensorimotor capabilities). Damasio’s work (1994) makes a similar although perhaps broader point: that our cognitive abilities depend on bodily processes of emotion. Damasio (1994; see also Clancey, 1999) argues that emotions not only serve a directive function or ‘‘point us in the proper direction.’’ Instead, emotions actually organize the perception and categorization of objects. Emotion is ‘‘not just a response to objects, but an organizer playing a causal role in categorization of objects’’ (Clancey, 1999, p. 218). The emotional feelings that help us categorize objects and the concepts of those objects form and develop side by side, rather than concepts forming first and emotions being mere passive reactions. In contrast to the new picture of cognition as embodied, the traditional perspective identifies cognition with abstract, disembodied information processing. Perceptual and motor systems are seen as mere ‘‘transducers’’ converting between abstract, amodal inner symbols (the stuV of real ‘‘cognition’’) and the external world. The mind/body separation is taken to an extreme, as the role of bodily constraints and needs in shaping cognitive abilities and processes is essentially ignored.

A. SOCIAL PSYCHOLOGICAL CONTRIBUTIONS TO THE THEME 1. Embodiment of Cognitive Processes a. Attitudes. Consistent with the embodiment assumption, recent work in social psychology (Neumann & Strack, 2000) underlines the close connections of attitudes to sensorimotor systems and therefore to our bodies. For instance, motor movements (isometric flexion and extension of the upper arm) have been shown to influence evaluative judgments

70

¨ N R. SEMIN ELIOT R. SMITH AND GU

(e.g., Cacioppo, Priester, & Berntson, 1993). Similarly, horizontal or vertical head movements influence evaluations (Wells & Petty, 1980) and recognition performance (Fo¨rster & Strack, 1996). In all these cases, bodily movements associated with approach or approval (arm movements pulling something toward the person or nodding the head vertically) cause evaluations to be more positive, or facilitate recognition of evaluatively positive material. Conversely, movements associated with avoidance or disapproval (pushing something away from one or shaking the head horizontally) cause evaluations to be more negative, or facilitate recognition of negative material. Other studies (Neumann & Strack, 2000) present words on a computer screen overlaid on a background of a rotating spiral pattern. Rotation in one direction gives an appearance that one is moving toward the screen, while rotation in the other direction generates the appearance of moving away. When the screen appears to be moving toward one, positive words can be more quickly classified, and when the screen appears to be moving away, negative words are given faster responses. All of these findings suggest that evaluations are not just ‘‘in the head’’ but involve the perceiver’s whole body, being linked to movement toward or away from objects in a real physical sense. b. Stereotypes. Stereotypes of social groups are also, like attitudes, aVected by bodily states and processes. In an early study of this issue, Bodenhausen (1990) provided information about an Anglo or Hispanic individual who was accused of an assault to participants who were either ‘‘morning people’’ or ‘‘night people.’’ When tested at their nonoptimal time of day (e.g., in the morning for night people), the participants relied more on their ethnic stereotypes, judging that the Hispanic was more likely to have committed the crime. As we will describe later, other types of bodily states such as emotions and moods also influence stereotyping. c. Creative Thought. Friedman and Foerster (2000, 2002) show that approach and avoidance motor movements aVect the creativity of responses that people generate, indicating once again close connections between bodily movements and mental processes. 2. The Perception-Behavior Link Interest in the ‘‘ideomotor’’ or perception-behavior link has a long history in psychology, dating back to William James (1890). Some research (reviewed earlier) by Bargh et al. (1996) and Chartrand and Bargh (1999) shows that subtly activating a concept such as ‘‘polite’’ can actually influence people’s behavior in the direction of being more polite. The same conclusion is supported by studies of the so-called mirror neurons found in the motor cortex in monkeys (Rizzolatti & Arbib, 1998). These neurons are found to be

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

71

active both when the monkey performs some specific action (e.g., scratching its head) and also when the monkey observes another performing the same action. This finding suggests that perception of others’ actions involves a motoric code, including at least some of the same components as actually performing the action. Perhaps the ‘‘ideomotor’’ eVects described by Chartrand and Bargh (1999) and others involve similar internal mechanisms, with embodied action being intertwined with perception from the very beginning. 3. AVect-Cognition Interfaces Social psychological researchers have identified a number of ways in which aVect (either general mood or specific emotions) influences cognition. Early work (as summarized, for example, by Isen, 1987) focused on mood and specifically on a positive/negative mood dichotomy. A number of simple bivariate relationships were identified, such as positive mood increasing creativity or negative mood increasing scrutiny of information. Later researchers, inevitably, identified moderators and boundary conditions for many of these eVects. We now know, for example, that mood can not only influence the amount of processing (e.g., through alterations in motivation or cognitive capacity) but also regulates processing in very fundamental ways. Considerable evidence supports the mood-as-information model of Schwarz and Clore (1996) and related proposals by Bless and others (Bless, 2000; Bless & Schwarz, 1999). In these models, positive mood is a signal that the situation is safe and that processing can rely on general rules based on repeated past experiences. In contrast, negative mood signals that a situation is problematic, implying a need for eVortful or vigilant processing of information (as conceptualized in the various dual-process models; see Smith & DeCoster, 2000). In these models, then, mood takes its place as an integral component of an overall regulatory system that governs many aspects of cognitive processing. Research interest in the eVects of specific emotions on cognition is more recent. Lerner and Keltner (2000), working from the perspective of appraisal/ emotion theory, find that when people experience a given emotion, they make judgments in line with the appraisals that are linked to that emotion. For example, an appraisal of uncertainty is part of the emotion of fear (compared to other negative emotions such as disappointment or anger) and so when people are experiencing fear, they judge events as more uncertain than when they are experiencing disappointment (Tiedens & Linton, 2001). This line of research and theory even suggests a reinterpretation of some of the work on mood and cognition, in that moods have often been induced through specific emotions (e.g., by having participants read stories, watch films, or recall

72

¨ N R. SEMIN ELIOT R. SMITH AND GU

autobiographical episodes that are either happy or sad). The specific appraisals associated with the happy or sad emotions, rather than the valence (positive or negative) of the resulting mood state, might thus be responsible for some or all of the cognitive processing changes identified by research (Tiedens & Linton, 2001). An additional and more general eVect of emotion on cognition has been identified by Niedenthal and her associates (1999). When people are in an emotional state, they categorize objects diVerently than they do in a neutral state. Specifically, the researchers present participants with three objects, such that one word (e.g., joke) is semantically related to a second (speech) and is related by emotional connotation to the third (sunbeam). Participants are asked to pick which of the latter two words is most similar to the first. In a neutral state people generally choose the semantically related word, but in an emotional state (whether sadness, happiness, or some other emotion) people see the emotionally related words as most similar. All of these types of research support the notion that emotion and aVect are closely and fundamentally interrelated with our cognitive processes, rather than constituting separable systems. This is what would be expected from a viewpoint that emphasizes cognition as an adaptive control mechanism for behavior (with emotions being part of the overall process) rather than one that views cognition as computation, disembodied and abstracted from the organism’s requirements for adaptive living. 4. Evolved Psychological Mechanisms The constraints of embodiment impact psychological processes in another way, which has been explored by evolutionary psychology theorists. The demands of social living pose certain recurrent problems (e.g., mate choice, child rearing, alliance formation). This fact sets the conditions for psychological mechanisms specific to these problems to evolve over time (Caporael, 1997; Simpson & Kenrick, 1997). Thus, evolutionary psychologists believe that specific aspects of our psychological makeup can be understood in terms of their adaptiveness in an evolutionary timeframe. For example, preferences for mating partners with smooth skin and symmetrical facial features are thought to have evolved because these are cues to relatively good genes and freedom from parasites and infections. Males’ preferences for younger females are believed to have evolved because youth is correlated with fertility. On a diVerent front, the fact that people have relative diYculty with certain logical problems (such as the Wason selection task) unless they are framed in terms of violations of social norms is argued to show that we have evolved psychological mechanisms specifically for detecting ‘‘cheaters’’ such as norm violators (Cosmides, 1989). Many of these claims are controversial, and

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

73

critics note that the evolutionary explanations are not the only plausible accounts for the observations (Wood & Eagly, 2002). Still, the evolutionary story points out another very broad potential eVect of embodiment on psychology—through the evolution of specific psychological mechanisms that have over the millennia contributed to survival and reproduction. B. TAKING THE THEME FURTHER As just reviewed, social psychologists have devoted extensive research attention to two central issues of the embodiment of cognition: (1) the role of perceptual/motor systems (broadly related to approach and avoidance) in evaluations and (2) the interrelations of emotion and cognition. Although many issues remain to be explored within those domains, in this section we briefly mention other research directions related to embodiment that have scarcely been explored by social psychologists to date. 1. Role of Perceptual Attributes in Stereotyping Recent research by Livingston and Brewer (2002) and Blair et al. (2002) shows that stereotyping is aVected by continuously varying perceptual attributes (such as skin tone and facial features) as well as by a target’s category membership. Such findings falsify a model in which perceptual features are used only to access a social category (e.g., African-American or female), which in turn triggers the attribution of category-stereotypic characteristics. Instead, perceptually based processes aVect stereotyping beyond the eVects of category membership per se. Further investigation of this phenomenon might usefully apply the model developed by Barsalou and his associates (Barsalou, 1999b; Goldstone & Barsalou, 1998) regarding the important role of perception in all kinds of conceptual processing. 2. Role of Motor Movements in Perception and Cognition Two research reports from areas outside social psychology illustrate further directions for exploration of the role of motor processes in perception and cognition, even aside from the approach/avoidance dimension. Boroditsky and Ranscar (2002) noted that the abstract concept of time can be conceptualized with two diVerent embodied metaphors. A person can be thought of as moving forward through time, or as staying still while time moves toward the person. The investigators noted that a statement like ‘‘the meeting was moved forward one day’’ is ambiguous, with its interpretation depending on the current metaphor (is a planned Tuesday meeting that was ‘‘moved forward a day’’ now on Monday or on Wednesday?). In a series of

74

¨ N R. SEMIN ELIOT R. SMITH AND GU

studies, they demonstrate that people’s current bodily experience of space and movement (e.g., position in a cafeteria line, imminent departure on a journey) systematically aVects their conceptualization of time and hence their interpretation of the ambiguous statement. In a study by Bekkering and Neggers (2002), participants either pointed at or reached out to grasp target items defined by a conjunction of color and position among an array of distractor items. Because the grasping movement itself depends on position (whereas position is irrelevant to pointing), they predicted that participants would show more sensitivity to this attribute, indicated by fewer position errors in the grasping condition. Their predictions were confirmed. Color errors were equivalent between the conditions, as predicted, because color is relevant to neither pointing nor grasping. Thus, the type of action that is required for a target systematically tunes perceptual attention. Motor planning is intertwined with perception from the very beginning. These studies simply illustrate some of the ways in which bodily experiences and behaviors influence our perception and conceptualization— beyond the issues of evaluation (and related approach/avoidance actions) and emotion-cognition interactions that social psychologists have studied to date. It would not be surprising if the embodied nature of cognition showed itself in eVects of bodily movements and experiences on most or all of the types of behavior in which social psychologists are most centrally interested: in the ways we perceive other people, think about them, categorize them, feel attracted to them, form relationships with them, accept or resist influence from them, and so forth.

III. Cognition is Situated: Emergent from Interaction of Agent and Task and Social Environment Since the passing of behaviorism, psychology and the cognitive sciences generally have been dominated by mentalist assumptions (Agre, 1997; Clancey, 1997a). Researchers focus on understanding ‘‘high-level’’ processes such as memory, thought, and decision making rather than ‘‘peripheral’’ processes such as perception and motor action. Cognition is considered a form of symbolic computation, with inner processes operating on stored mental representations of knowledge about the self and the world. Mainstream approaches within social psychology have often shared these presuppositions. Thus, people’s thoughts and actions have been investigated as a function of their mental representations—their attitudes, impressions of other people and stereotypes of social groups, scripts and schemas, and so

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

75

on. The processes that operate on these inner representations are also assumed to be inside the head—systematic or heuristic processing of persuasive messages, activation or suppression of stereotypes, attributional inferences, and so forth. Social psychology is centrally concerned with the eVects of the social environment on individual representations and processes, of course, but the environment has been conceptualized essentially as providing ‘‘inputs’’ to what remains very much an inner processing story. As Philip Agre has written, Mentalism provides a simple formula that gives plausible answers to all questions of psychological research: put it in the head. If agents need to think about the world, put analogs of the world in the head. If agents need to figure out what might happen, put simulations of the world in the head. The tacit policy of mentalism, in short, is to reproduce the entire world inside the head. The sophisticated structures and processes [assumed by the mentalist approach] are not geared to living in the world; they are geared to replacing it (1997, p. 51, italics in original).

As part of a rethinking of these standard mentalist assumptions, the SSC approach assumes that social cognition and social action take place within a socially meaningful environment. The defining features or aVordances (Gibson, 1966, 1979) of that environment are resources for, and constraints on, cognition. One implication of this point is that fine-grained interaction or coupling between the agent and the environment often obviates the need for detailed inner representations. The mentalist assumption is that we navigate through our social environment by drawing on inner representations such as scripts. We might be assumed to have scripts for eating in a restaurant or playing a baseball game, which specify what actions to perform in what sequence. Despite the familiarity and seemingly obvious intuitive validity of that claim (Schank & Abelson, 1977), a moment’s thought should make it clear that a script cannot possibly suYce to guide meaningful social behaviors. A meal cannot be successfully ordered without monitoring the waiter’s nonverbal signals (such as eye contact) that indicate appropriate conversational turns. Catching a fly ball cannot be done by following a generic ‘‘fielder’’ script, but only by monitoring the path of the ball as one runs to the spot where it can be caught. In other words, we certainly have conceptual knowledge about restaurant meals and baseball games, which we can draw on to comprehend simple story-like texts (which was actually the goal of Schank and Abelson’s [1977] original development of the script concept). But because of the inherently variable nature of the real social world, scripts or other inner representations can never suYce to direct appropriate and adaptive behavior. Behavior emerges from continual sensori-motor interaction with the world.

76

¨ N R. SEMIN ELIOT R. SMITH AND GU

Summarizing this point, William Clancey holds that Human behavior is situated because all processes of behaving, including speech, problem-solving, and physical skills, are generated on the spot, not by mechanical application of scripts or rules previously stored in the brain. Knowledge can be represented, but it cannot be exhaustively inventoried by statements of belief or scripts for behaving. Knowledge is a capacity to behave adaptively within an environment; it cannot be reduced to (replaced by) representations of behavior or the environment (1995, p. 229).

The socially situated cognition perspective requires a shift in theoretical focus: explanations of behavior cannot be based solely on the individual’s internal representations, but on the interaction of the individual with the social and physical situation. The idea that behavior flows from the inner states of the individual is deeply embedded in Western culture—not only in current theories of psychology. The image of the autonomous individual, who chooses action based on reasoning about his or her own preferences, attitudes, beliefs, and values, is very diYcult to discard for a number of reasons (Clark, 1999a, pp. 16–19; LakoV & Johnson, 1999, pp. 553–557). However, we must understand that ultimately this is a meta-theoretical assumption, not a scientific theory. If it is considered as the starting point for psychological theory, it has serious limitations. Simple examples suYce to illustrate what is meant by behavior arising from interaction. Mataric’s (1991) robot, Toto, provides an early and important example in the situated robotics movement (Brooks, 1991). Toto is a small mobile robot with a compass, sonar-based distance sensors, and wheels. Simple logic circuitry creates connections between the sonar sensors and wheel motors. As Toto moves, if the sensed distance to the nearest wall becomes too small, the motor speed changes so Toto veers farther away. Conversely, if too far from the wall the motor speed changes make Toto approach it. The resultant behavior is following the wall around the room, one or two feet away from it. Note that this behavior does not reflect only Toto’s internal mechanisms but also the typical properties of walls (vertically oriented, tall enough, good reflectors of sonar, etc.). Note also that Toto has no explicit inner representation of what a wall is, let alone an explicitly represented goal to follow walls. Rather, the behavior emerges from the interaction between Toto’s specific hard-wired proclivities and the physical properties of walls. No conceptual analysis of the behavior can reasonably give a more central role to the agent’s wiring than to the environmental properties (Agre & Chapman, 1990). In the situated cognition approach, the environment is a recipient of action as well as a supplier of inputs. In fact, a central insight of situated

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

77

cognition is that actions change inputs. As Toto moves, the sensed distance to the wall changes. In a social interaction, a person’s conversational comments shape what the partner says next. The environment is not a passive source of ‘‘inputs’’ but is interactive and responsive to the agent’s actions, in a process of ‘‘continuous reciprocal causation’’ (Clark, 1997). As Maturana and Varela (1987, p. 169) note, if the nature of the organism-environment interaction diVers for diVerent activities, there is no fixed set of ‘‘inputs’’ and ‘‘outputs’’ that furnishes a useful description across the board. Thus, the theoretical focus must be on the interaction of agent and environment, rather than spotlighting the agent and especially his or her internal representations. This focus denies our natural tendency as observers to attribute behavior, especially goal-directed, adaptive behavior, to inner characteristics (plans, scripts, etc.) of the agent (Heider, 1958). These attributions are common even when the behavior actually emerges from interaction of the agent and the environment. Braitenberg (1984) has furnished compelling illustrations of this fact, by asking readers to imagine simple vehicles constructed of sensors and motors. For example, a vehicle with a light sensor controlling the speed of a motor will move quickly in the vicinity of bright light and slow down or stop in dark areas. Although the vehicle has no goals, emotions, or plans, observers are inclined to say that it ‘‘hates’’ light, ‘‘prefers’’ darkness, and ‘‘looks for’’ dark areas in which to ‘‘hide.’’ In fact, as far back as the 1940s, Heider and Simmel (1944) noted people’s tendencies to attribute goals and desires to even simple geometric shapes shown moving and interacting with each other. A. SOCIAL PSYCHOLOGICAL CONTRIBUTIONS TO THE THEME 1. Communication as Socially Situated Action Currently in social psychology, and consistent with the situated cognition perspective, communication is conceptualized as social action and not as the mere transmission of information from a source to a recipient, like a fax machine. There is a long-standing research tradition in social psychology that has shown the adaptive responsiveness of communicative, cognitive, and evaluative processes to the situated properties of communication contexts (Higgins & Semin, 2001). As summarized below, this research supports four conclusions. First, communicators systematically take into account their audiences and formulate their messages by shaping them to the characteristics of their audience and their situation. The addressee is a full participant in how a message is shaped by the communicator (e.g., Krauss & Fussel, 1996; Zajonc, 1960) and

78

¨ N R. SEMIN ELIOT R. SMITH AND GU

how the message is delivered (i.e., speech accommodation theory, see Giles and Coupland, 1991). Second, audiences treat communication as social action and do not simply decode the information transmitted in a message. They take the communicator and his or her characteristics into account in deciding why and how a message is formulated. They assume that the communicator shaped the message for the purposes of the audience and adjust their judgments to what they assume the communicator’s true beliefs may be (see Higgins, 1992; Krauss and Fussell, 1996). A third line of research shows how communicators flexibly and strategically use language to influence an audience (e.g., Maass, 1999; Semin et al., 2003; Wigboldus, et al., 2000). The fourth line of research shows that audience responses can influence a communicator himself or herself. Throughout these diverse research lines, the adaptation of communicative acts to the characteristics of the audience, the communicator, and their respective goals is the main theme, attesting to the socially situated nature of human communication. a. Audience EVects on Communication. When people communicate they do not simply encode some information and then transmit it to a destination. Speakers take into account the social context of their communication when they are modulating a message (Semin, 2000b). The classic study demonstrating this was reported by Zajonc (1960) in what he termed the process of cognitive tuning in communication. This study showed that the communicative role (transmitter vs. recipient) and the purpose of the communication influenced both the perceiver’s cognitive representation and the nature of the communication, showing that both impression formation and message qualities were shaped by the communicative characteristics of the situation (see also Zajonc & Adelman, 1987). This theme is also found in the classic research on ‘‘saying is believing’’ initiated by Higgins and Rholes (1978). In these studies, speakers’ relationships were experimentally shaped to promote positive self-presentation or intimacy to a listener. Interdependence between communicator and recipient influenced not only the message people wrote, but also the communicator’s own beliefs. These and other studies (Higgins & McCann, 1984; Higgins, McCann & Fondacaro, 1982; McCann, Higgins, & Fondacaro, 1991) show that participants distort their messages in a way that is consistent with an audience’s attitudes. Thus, when another person is the topic of their message, speakers encode the same information about that target person diVerently depending on whether they think the audience likes or dislikes that person. Moreover, their impressions as measured at a later time are evaluatively consistent with the content of their message. Similarly, information that a speaker knows is shared with the audience influences what they include and exclude in their message. b. Contextual EVects on Message Interpretation. The subtlest of cues, if they provide information about a potential audience, can influence supposedly

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

79

fundamental and automatic attribution processes, as Norenzayan and Schwarz (1999) have shown. They demonstrated the context sensitivity of the fundamental attribution error by using the subtlest of cues, namely a research letterhead that read either ‘‘Institute for Social Research’’ or ‘‘Institute of Personality Research.’’ Participants gave causal explanations for a mass murder based on a newspaper report. They used more situational explanations and fewer dispositional ones in the former case, whereas the ‘‘Personality’’ context resulted in more dispositional causes: a stronger fundamental attribution error. Norenzayan and Schwarz (1999) suggest that the subtle letterhead manipulation of the communicative context influenced participants’ perceptions of what is epistemically relevant to the researchers. Again, this study shows that processes considered to be as stable, fundamental, and context-free as the fundamental attribution error can be incredibly responsive to subtle aspects of social situations. c. Communicative Relevance Shapes Messages. Another set of studies show that the communicative relevance of a message contributes to the shape of what people say and think. An early example is a study by Carlsmith, Collins, and Helmreich (1966), which replicated and extended the classic Festinger and Carlsmith (1959) dissonance experiment. In the original experiment participants were paid $1 or $20 to inform a naive subject that a dull task that the participant had just performed was interesting and exciting. As is well known, when these participants are later asked to report how much they liked the dull tasks, participants who were paid $1 reported that they liked it more than those who were paid $20. Carlsmith et al. (1966) replicated this experiment, but asked half of the participants to write an anonymous essay (instead of informing a naive subject) describing the dull tasks as interesting. Only the experimenter would see this essay. While the original dissonance findings were replicated for participants asked to inform a naive subject, the anonymous essay condition yielded liking proportional to the financial incentive. ‘‘The response alternatives of the task-liking questions are associated with diVerentially evaluated situated identity attributions and subjects choose the most favorable one in each condition’’ (Alexander & Knight, 1971, p. 68). The situated responses when the audience is a naive subject versus an experimenter introduce diVerent communicative relevancies and consequently diVerent cognitive processes and responses to the ‘‘liking’’ variable. Research by Semin et al. (2003) contrasted the influence on a message of the communicative context with that of psychological processes (expectations). Earlier research (see Maass, 1999) on the Linguistic Intergroup Bias has shown that in intergroup contexts, the linguistic properties of messages vary systematically as a function of cognitive (expectations) or motivational processes (e.g., threat to ingroup identity). The research by Semin et al.

80

¨ N R. SEMIN ELIOT R. SMITH AND GU

(2003) showed that such systematic variations in the linguistic properties of messages only occur if participants expect that the message has a function (the addressee will actually read it). However, when the message has no communicative function, because participants are told nobody will read it, the message does not show any systematic variation in its linguistic properties. These results suggest that the bias is not due to context-free, automatic inner processes related to invariant linguistic expressions of diVerent types of knowledge. Instead, cognitive and motivational processes are strategic and responsive to social (i.e., communicative) contexts. d. Speaker Accommodation to Audiences. Several studies have shown that speakers adapt many verbal and nonverbal behaviors (e.g., linguistic, prosodic, nonvocal features including speech rate, pausal phenomena, phonological variants, smiling, gazing) to audience characteristics. For instance, Levin and Lin (1988) have shown how Watergate figure John Dean, when testifying before a U.S. Senate committee, adapted his median word frequency (a sign of formality) by converging to the median word frequency of the respective senators interviewing him. Similarly, Coupland (1984) showed how a travel agent converged phonologically to her individual clients, thus adapting to each client’s communicative behaviors. These types of convergence increase perceived attractiveness, perceived supportiveness, and interpersonal involvement (Giles et al., 1987). For instance, Buller and Anne (1988) showed that when slow and fastspeaking participants were addressed at their own rates, they judged such interviewers to be closer and more intimate. Similarly, Putnam and Street (1984) showed that speakers who converge toward their interviewers in terms of speech rate and response latency are reacted to favorably in terms of perceived social attractiveness. More recently, similar behavioral convergences have been observed under the new label of the ‘‘Chameleon eVect’’ (Chartrand & Bargh, 1999, see section above). The communication research has also found that in situations of intergroup conflict, speakers may adopt divergence strategies, accentuating diVerences in communicative behaviors between self (ingroup) and others (outgroup). These can take very diverse forms (see Coupland & Giles, 1991). For instance, Bourhis et al. (1979), in experimentally controlled neutral and ethnically threatening encounters, have shown that when an outgroup speaker was demeaning toward the ingroup and used a strong outgroup accent, it led to polarization of ratings of ethnic in- and outgroup identity (see also Bourhis & Giles, 1977). 2. Contextual Influences on the Self Social situations (including psychological experiments) are events in which people (participants and experimenters alike) convey a particular identity by their choice of actions (GoVman, 1959, 1961; Strauss, 1959). Such situated

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

81

actions are determined by the demands of a social context, which confers a specific meaning to actions and constrains the way actions express the individual’s identity. Any social situation entails a range of possible identities. The functional role in which participants find themselves in an experimental (or other) context therefore shapes processes of socially situated cognition. ‘‘The social context of the situation establishes the relative salience of attribution dimensions for impression formation. This implies that the basic contours of a social situation can be described in terms of the configuration of situated identity dimensions that are important for shaping an actor’s image’’ (Alexander & Knight, 1971, p. 76). In line with these assumptions, McGuire, McGuire, and Cheever (1986, McGuire & McGuire, 1986) studied self-conceptualizations in contrasting social contexts (school and home) by examining the types of verbs that were used in the spontaneous (oral or written) self-descriptions provided by 7 to 17 year olds. Their analyses of the diVerent narratives led to the conclusion that ‘‘even so individualistic and intimate an aspect of social cognition as the phenomenal sense of self takes diVerent forms depending on whether it is evoked in the context of the family or of school, two major social worlds of childhood’’ (p. 260). For example, the findings showed that the sense of self is more passive in the family context and more dynamic or active in the school context. Furthermore, McGuire’s work shows that self-characteristics that are salient to a person are a function of the distinctiveness of this characteristic in a social context. Basically, this research shows that the types of categories (e.g., those relating to gender, ethnicity) that people deploy in spontaneous self-descriptions depend on the degree to which it is peculiar in one’s situated reference group. Thus, people are significantly more likely to mention their weight, their birthplace, their gender, or ethnic membership if they are heavier than the rest of your reference group, born in an exotic town, or are in a minority in their reference group regarding gender or ethnicity (e.g., McGuire & McGuire, 1980, 1981; McGuire, McGuire, Child, & Fujioka, 1978; McGuire & Padawer-Singer, 1976). Crocker et al. (1991, Experiment 2) showed the contextual sensitivity of self-esteem by creating a context in which participants who are generally known to be targets of prejudice (black students) received positive or negative feedback from an evaluator on essays they had written. They were much more likely to attribute the feedback (positive or negative) to prejudice when the evaluator could see them (thus identify that the source of the essay was a black person) but not when they could not see them. This pattern did not emerge for white participants, irrespective of whether the evaluator could or could not see them. The theme of the situated impact of feedback on self-esteem is also demonstrated in studies on the self-esteem of overweight women (e.g., Quinn & Crocker, 1999). They are shown to be highly

82

¨ N R. SEMIN ELIOT R. SMITH AND GU

susceptible to contextually activated (salient) ideologies. In other words, this research line by Crocker and colleagues suggests that the self-esteem of stigmatized individuals is a situated construction rather than an enduring individual facet (Crocker, 1999). Some social contexts are externally given (e.g., school vs. home contexts, situations entailing communication or no communication, subtle cues such as the letterhead on a questionnaire). Other social contexts are those we create by our own actions as we publicly structure situations. Thus, another way of looking at social context is by considering the impact that our own public behaviors, as self-generated social contexts, have on our evaluations of our selves. There is a substantial research literature showing that externally induced behaviors influence the self concept, suggesting an ‘‘internalization’’ of the overt behavior (e.g., Fazio, EVrein, & Falender, 1981; Jones, Rhodewalt, Berglas, & Skelton, 1981; Markus & Kunda, 1986; Rhodewalt & Agustsdottir, 1986). The prominent explanations for this eVect have invoked automatic, context-free cognitive processes, for instance that the self-concept changes ‘‘resulted from a change in the accessibility of particular self-conceptions’’ (Markus & Kunda, 1986, p. 865) that were activated by the self-presentation. Similarly, it has been argued that biased scanning, increasing the cognitive salience of specific traits, is responsible. However, in a series of studies, Tice (1992) showed that identical behaviors are much more likely to impact the self-concept when they are performed in an interpersonal context than when they are performed privately. Thus, when participants were asked to portray behaviors regardless of whether they possessed the trait, they internalized these traits when the self-presentations were highly identifiable but not when in relatively private conditions. This eVect was shown not only on self-report measures, but also behavioral ones. These and other studies (e.g., Schlenker, Dlugolecki, & Doherty, 1994) show how the eVect of behavior on self-evaluations depends on specific social contexts. The eVect cannot be attributed to context-free, automatic inner processes such as an increase in accessibility of traits related to the self-presentation. 3. Contextual Influences on Responses to Others Social contexts aVect not only the self-concept and self-esteem, but also responses to other people and social groups. In particular, social contexts that lead people to think of themselves as members of a particular social group (an in-group)—such as the presence of out-group members or a situation of intergroup conflict—have pervasive eVects (Turner et al., 1987). Fellow in-group members are liked, at an automatic or implicit level as well as on explicit self-report ratings (Mullen, Brown, & Smith, 1992;

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

83

Perdue, Dovidio, Gurtman, & Tyler, 1990), and as a consequence are treated favorably. In contrast, however, out-group members come to be seen as homogeneous (Linville et al., 1989) and are disliked and behaviorally discriminated against (Billig & Tajfel, 1963; Opotow, 1990). Thus, as shifting social circumstances—changing alliances, newly emerging intergroup conflicts—alter the salient lines of social categorization, entire configurations of judgments and behaviors toward other people will be likely to shift accordingly. Importantly, as a recent review by Blair (2002) has demonstrated, eVects of social contexts on responses to other people (including attitudes and stereotypes) extend to implicit measures (response times, priming tasks, IAT measures) as well as explicit (self-report) measures. Social contexts alter our perceptions and reactions to others even at automatic, nonconscious levels, not only in terms of the overt reports that we make. In addition to shared group memberships or group identification, other types of social contexts aVect people’s responses to others. Social power influences many types of processes, including communication (Giles & Wiemann, 1987) and stereotyping (Fiske, 1993); people in positions of power stereotype their subordinates to a greater extent than the reverse. In addition, interdependence, as in close relationships, aVects the ways people think about and evaluate their interdependent partners (Murray, Holmes, & GriYn, 1996; Murray et al., 2000). In all these ways, social psychologists have explored how the social ties and connections that link people aVect their thoughts, feelings, and evaluations. 4. Situations as Cues for Behaviors A powerful illustration of the role of the environment in driving action is provided by Gollwitzer (1999). These studies compare participants who form intentions to perform a specific action, either under instructions to think in particular about when and where the action will be done or without such instructions. For example, I might decide that on Election Day I will vote at the local fire station at 5 pm on my way home after work. Participants who form such specific situated intentions actually perform the planned behavior at remarkably higher levels than do participants who form the equivalent intention, but without reference to particular places or times. The diVerence is that the time and place—situational cues—can assist our memory and turn the behavior into more directly situated action, rather than leaving it to be driven only by relatively abstract inner plan representations. Clancey (1997a, pp. 97–100) analyzes computer interface designs in these terms, discussing the way that items on a screen (e.g., windows) can act as external reminders, or cues to intentions, facilitating the use of the computer system.

84

¨ N R. SEMIN ELIOT R. SMITH AND GU

5. Self-Attributions of Agency As sense-making and narrative-constructing creatures, we construct rationales and explanations for many things that we observe—especially those such as our own behaviors that is fundamentally important to us. When people introspect or reflect on their own behavior, they are often led to conclude that their inner representations and characteristics—their beliefs, goals, personality traits, values, and so on—are what cause them to behave in the ways that they do. These beliefs about the sources of our own behaviors are powerful and subjectively compelling. Yet, as we have just argued, the causes of behavior more frequently actually reside in the situation rather than within the person. How can we solve this puzzle? Social psychological research dating back to Festinger (1957) and Bem (1967) has demonstrated that our subjective experiences of our inner characteristics are often post-hoc rationalizations for our actions. That is, action comes first, and then through self-perception or dissonance processes, we infer corresponding inner characteristics such as traits, goals, or attitudes. Thus, the intuitive picture of a match between our internal states and our overt behaviors, which naturally seems like the strongest evidence that our behaviors were generated by those states, actually may not demand that conclusion at all (see Wegner & Wheatley, 1999). The observer-inferred (or self-inferred) traits, attitudes, or goals may seem to justify keeping the analytic focus on the individual. But the SSC perspective directs us to consider both the actual causation of the behavior, which may have been induced by specific though subtle environmental stimuli (e.g., Chartrand & Bargh, 1999), and the way the behavior leads to the post-hoc inference of corresponding traits.

6. What Aspects of Situations Matter? As stated previously, cognition and behavior are, from the SSC perspective, caused by social situations in interaction with properties of the person. But the term situation is infinitely extensible—what particular aspects of situations are crucial for human social behavior? Social psychologists have much to contribute to answering this question. First, social psychologists have identified a small number of basic social goals and motives (e.g., Stevens & Fiske, 1995) that structure adaptive behaviors in social situations. For example, people seek to maintain self-esteem, to build close bonds of connectedness with others, to predict and control their environment, and so on. Hence, aspects of situations that cue these goals or signal that one or more of the goals are relevant are expected to be particularly influential over thought and behavior. Second, social psychological

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

85

research on personal relationships and group processes tells us about additional crucial situational factors. Personal relationships involve processes of commitment and interdependence (e.g., Agnew et al., 1998) and depend in important ways on the partners’ perceptions of each other (e.g., Murray et al., 2000). Group interaction depends on processes of conformity and norm formation (e.g., Sherif, 1936) and on social comparisons with relevant outgroups (e.g., Tajfel et al., 1971). Thus, these bodies of social psychological research translate rather directly into specific predictions about what aspects of situations will matter for social behavior. Cues about the relative status of an out-group, for example, can be expected to aVect people’s behaviors when in groups, whereas other aspects of a situation, such as the color of the walls, should not matter in general. In summary, the general thrust of many diverse studies is to show that environmental contexts—and particularly features of the communication context, including the relationship of the individual to partners, communicators, audiences, or fellow group members—are among the most important regulators of cognition and action. Human beings are capable of detached (‘‘oV-line,’’ nonsituated) cognition, contemplating situations that occurred long ago, imagining fictional or artistically constructed worlds, or planning for distant future events (Wilson, 2002). When we exercise these abilities, immediate social environments have less of a hold on our thinking (although we would not expect their eVects to totally vanish). In most everyday situations, socially situated action emerges from the interaction of the agent and the environment and demonstrates an exquisite sensitivity to the relational and motivational implications of the environment. These studies’ demonstrations of context sensitivity contradict the notion that mental processes, such as the fundamental attribution error or attitude change through dissonance reduction, are invariant properties of our perceptual or cognitive systems. Instead, these processes are found to depend on specific features of the social environment (Carlsmith et al., 1966; Norenzayan & Schwarz, 1999), suggesting that they emerge as a function of the relational and communicative context of action.

B. TAKING THE THEME FURTHER As the above review shows, social psychologists have developed several bodies of research and theory that are highly relevant to the assumptions of the situated cognition perspective. Yet in some ways, social psychology as a field has not fully adopted the assumptions of this emerging perspective. This section lists several ways in which some common assumptions demand rethinking.

86

¨ N R. SEMIN ELIOT R. SMITH AND GU

1. Conceptual Knowledge is Situation Specific If cognition and action depend on the details of specific situations, it follows that conceptual knowledge should not be represented in abstract, situationfree form but instead should be organized in terms of specific situations. This implies that conceptual knowledge takes diVerent forms in diVerent situations, consistent with the picture of situated cognition. Yeh and Barsalou (2000) have summarized a variety of evidence supporting this assumption. For example, people recall lists of words better in the same situation (e.g., the same physical room) in which they previously studied them than they do in diVerent situations. Words also can be recognized better when they are presented in the same sentence context at test as at study. People can verify that specific properties are true of an object or concept much faster when the properties are relevant to the situation in which the object is presented than in a situation that makes the property irrelevant. Yeh and Barsalou (2000) summarize these and other forms of evidence for the idea that the agent’s current situation activates related conceptual knowledge, at the same time as concepts activate related situational information. Active conceptual knowledge always takes situation-specific forms. The authors argue that these aspects of situated cognition are highly functional, on the basis that using situations during conceptual processing is likely to improve prediction of the environment. This is because concepts are likely to continue to be relevant in situations where they were relevant on past occasions. In contrast, social psychologists have often assumed that inner representations are abstract and context free—stored as prototypes, schemas, or rules, divorced from the specifics of the situations in which the knowledge was acquired and used.

2. Methodological Implication: Study How People Use Environments to Support Social Behavior Because the environment can be part of our cognitive processes, it is natural to assume that we learn to actively manage the environment to ease memory or processing loads. As Kirsh (1995) has described, we do exactly this. We remind ourselves to buy more laundry detergent by putting the empty detergent box near the door so we will notice it when we go out. We physically cover up or hide things that represent tasks that cannot be done yet, avoiding distraction. We leave tools and materials where we can readily find them again. We physically group together a number of similar small items (e.g., pencils and pens) rather than leaving them individually scattered about, so that they are more readily perceptually locatable and identifiable.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

87

All these and other tricks let us use our environment to oZoad memory (e.g., we do not have to recall information that is right in front of us) and choice (e.g., we do not have to decide what to do next if materials for the next task are right on top), saving cognitive resources. The SSC approach prompts seeking explanations of behavior in such interactions between an agent’s inner ‘‘machinery’’ and aspects of the environment. Research using this approach assumes that ‘‘machinery and dynamics [i.e., behavior patterns] constrain one another: only certain kinds of machinery can participate in given kinds of dynamics in a certain environment. For example, an agent that always puts its tools back where they belong may need simpler means of finding those tools than an agent that leaves them lying about’’ (Agre, 1997, p. 62). Given this conceptualization, the principle of ‘‘machinery parsimony’’ becomes a key methodological guideline: ‘‘postulate the simplest machinery that is consistent with the known dynamics’’ (p. 62). This principle leads directly to a research focus on the ways that environments provide useful resources (supports, scaVolding) that can simplify an agent’s tasks. ‘‘Far from the Cartesian ideal of detached contemplation, real agents lean on the world. The world is its own best representation and its own best simulation’’ (Clark, 1997, p. 63, italics in original). For example, social psychologists have often uncritically adopted the mentalist assumption that routine behaviors such as eating in a fast-food restaurant are driven by internal representations or ‘‘scripts’’ (Schank & Abelson, 1977). Considering the potential supports for behavior provided by the environment opens a profoundly diVerent alternative viewpoint. The next time you are in a fast-food restaurant, notice that there are signs saying ‘‘Order Here.’’ Lines of customers may be waiting to be served. The cashier will ask, ‘‘May I take your order?’’ and at a later time state, ‘‘That will be $3.47, please.’’ In other words, simple language skills and a few generalpurpose algorithms (stand at the end of a line; pay when asked for money) would suYce to guide one through the restaurant behavior. No overall script knowledge is necessary because of the plentiful environmental supports. Thus, one specific direction for research opened up by the socially situated cognition perspective is investigation of the ways social and physical environments support social behaviors that had been traditionally attributed to inner processes and representations. 3. Methodological Implication: Study Social Behavior in Context If human behavior is sensitive to social situations and contexts, it follows that the situation cannot be ignored when social behavior is being studied. Sometimes the social psychological laboratory is regarded as a sterile, virtually

88

¨ N R. SEMIN ELIOT R. SMITH AND GU

context-free setting for studying behavior, and thus superior to other more specific and limiting contexts (such as organizations or other field settings). In our view, this is a mistake. The laboratory is a social situation, and thus many aspects of it (such as the communicative relationships between experimenters and participants) aVect participants’ responses, just as they do in any social situation. As noted earlier, for example, the fundamental attribution error may be specific to contexts that suggest that the experimenter is interested in personal dispositions as potential attributional responses (Norenzayan & Schwarz, 1999; Schwarz, 1994, 2000). Theoretically informed assumptions about socially situated action invite us to rethink experimental contexts, the goals of participants in our experiments, and the stimulus materials we use. The true strength of the laboratory is not its supposed insulation of behavior from context eVects, but its flexibility in allowing experimenters to construct very diVerent types of contexts, suited to test diVerent types of theoretical hypotheses. But a methodological implication of this way of thinking is that laboratory researchers should be mindful of the contexts and situations they are creating. The social context of the psychological experiment was a major concern of the field once before, in the 1960s and early ’70s. The focus was on demonstrating the contribution of unintended factors such as ‘‘demand characteristics,’’ or participants’ understandings of the communicative and other goals of the researcher, to experimental outcomes (Orne, 1962; Rosenthal & Rosnow, 1969; inter alia). This entire debate addressed, in our terms, the socially situated nature of experimentation. The lack of understanding (at that time) of the socially situated nature of cognitive and behavioral processes made it possible for these sources of variance to remain unnoticed. Such factors include the types of demands that are conveyed implicitly to participants, the influence of experimenter’s expectancies on outcomes, and the participant’s role (e.g., of being a good ‘‘subject’’). There is another way in which contexts have been overlooked and neglected in our methodological practice. Self-report measures are a mainstay of social psychological research, and they often ignore context by asking people (for example) to report their attitudes toward various objects without any context being specified, or to report their standing on various broad personality traits or aVective states, again without context. Thus, a participant might be asked to indicate how favorably he or she evaluates Asians, or to state to what extent he or she is generally honest or happy. Although such measures are so often used and familiar that it may be diYcult to see the problem, in fact by failing to specify a context they require the participant to develop one on his or her own. Should Asians be evaluated in the context of competitors for a limited number of jobs, for example? However, self-report

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

89

questionnaires need not ask general, context-free questions. We could ask participants specifically how they evaluate Asians as potential job competitors, how honest they are with close family members, or how happy they are when engaging in leisure activities. Or we could use experience-sampling methodology, beeping participants at selected times and asking them to report how happy they are at the moment, or how honest they were in their most recent social interaction. Or we could use scenario methodologies, describing specific situations and asking people to indicate how they would evaluate a social group member in that situation, or how honest or happy they would be. Such methodologies seem to be required to assess theories that emphasize the role of context in people’s thoughts, feelings, and behaviors. Reliance on general self-report questions that ask people to make summary judgments without specifying context does not simplify theory by ignoring an irrelevant but complicating factor. Rather, such reliance may complicate theory by introducing arbitrary variation as diVerent participants select diVerent contexts to anchor the questions and give them meaning, as well as by denying researchers the opportunity to systematically investigate the consequences of varying contexts on whatever is being measured.

IV. Cognition is Distributed Spatially and Temporally Across Tools, People, and Groups The fourth major theme of the SSC approach is that cognition is distributed. Cognition not only takes place within individual brains, but also makes use of tools and other artifacts, aspects of the environment, other people, and groups. Indeed, the social psychologist Thomas M. Ostrom argued in 1984 that all cognition is inherently social, thus challenging a purely individual-centered view of information processing. The evolution of human society in general and individual functioning in society cannot be understood properly unless we conceive of knowledge as a cumulative process that is distributed and preserved by diverse means. Such means include physical tools (such as compasses, hammers, and calculators), the structuring of the physical environment (road signs, the architecture of restaurants, and post oYces), and the distribution of knowledge across people and groups (car mechanics, navigators, programmers). Of course, for agents to be able to ‘‘lean on the world’’ (Clark, 1997), they must be able to access, coordinate, and synchronize distributed knowledge to solve specific problems. This means that agents have to have access to shared or common knowledge and to tools (e.g., language) that enable ‘‘social coupling’’

90

¨ N R. SEMIN ELIOT R. SMITH AND GU

(Semin, 2000a). This is not a mere passive uptake of knowledge but a twoway interaction, either between agents or between an agent and the physical environment or physical tools. These constitute some of the central themes addressed in this section. A paradigm example of distributed cognition is Hutchins’ (1995b) analysis of ‘‘how a cockpit remembers its speeds.’’ The cockpit is that of a commercial airliner, and the speeds that need to be remembered are those at which specific types of flight maneuvers can take place (e.g., flaps can be extended). The relevant memory system includes the pilot and copilot, who maintain a division of labor between them (one performing the demanding task of flying the airplane on a runway approach, the other calling out relevant speeds). The memory system goes beyond these two interacting individuals to include external mechanical memory aids as well, such as movable indicators (‘‘speed bugs’’) that mark specific zones on the face of an air-speed indicator on the control panel. Hutchins persuasively shows that all of these elements can be usefully analyzed as a single memory system—that the cockpit (not the pilot) remembers his speeds. The idea that cognitive processes such as memory can extend outside the head may seem odd if one holds to a definition of the ‘‘cognitive’’ that draws the line a priori and says that only brains can take part in cognitive processes. Yet such a definition is problematic. By focusing only on unaided brains (i.e., ruling out external artifacts like a calculator), it rules out language too—also an artifact. In fact, this definition of cognition is even more restrictive than a definition that equates cognition with information processing; certainly other things besides brains (e.g., calculators and computers) can process information. The situated cognition perspective does not state that brains are irrelevant, but holds that cognition as preparation for adaptive action emerges from the interaction of brains, bodies, and the world (Clark, 1999b). There are numerous examples of the way agents ‘‘lean on the world’’ to simplify their inner processing requirements. Most have been investigated by cognitive scientists and others outside of social psychology. These examples include the distribution of complex tasks across group members (Hutchins, 1995a), the fact that direct observation can often usefully replace complicated deductions (Brooks, 1991; Kirsch, 1995), the provision by cultures of linguistic and representational schemes that support complex forms of cognition (Latour, 1986), and the ways common artifacts (pencils, forks, scissors) provide readily perceptible cues to their functions (aVordances; see Agre, 1997, p. 225). In social psychology, study of the distributed nature of cognition has focused on how cognition extends outside the isolated individual and includes people or groups. Complementarily, there is a long-standing social psychological interest in the socially shared nature of knowledge. We turn first to these domains.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

91

A. SOCIAL PSYCHOLOGICAL CONTRIBUTIONS TO THE THEME Social psychology has made substantial contributions toward understanding the socially distributed nature of knowledge, by elucidating how cognition is distributed over individuals and groups and how it is possible for people to lean on such social resources. Moreover, the issue of how distributed knowledge is accessed and used has also been a focus. Recent years have seen increasingly articulated social psychological analyses of what it means for cognition to be social, distributed, and shared (Caporael, 1997; Levine, Resnick, & Higgins, 1993; Resnick, Levine, & Teasley, 1991; Schwartz, 1996) and what group cognition may entail (Kerr et al., 2000; Ruscher & Duval, 1998; Tindale & Kameda, 2000; see for a general review: Thompson & Fine, 1999). Some of these concerns have been of long-standing interest in social psychology. For instance, the issue of how socially shared reality emerges in social interaction has long been at the core of social psychology (e.g., Asch, 1955; Festinger, 1950; Sherif, 1936), continuing to today (e.g., Hardin & Higgins, 1996). There is a growing acknowledgement across these diverse theorists of the importance of understanding what it means for human interaction to form a common reality. Nevertheless, these approaches still remain ‘‘an orientation or perspective’’ (Thompson and Fine, 1999, p. 280) rather than an integrated theory with a body of systematic empirical results. Social psychological contributions cluster around two broad subjects: (1) that cognition is socially distributed and preserved and (2) that cognition is socially shared. These two subjects are complementary. Although it may be self-evident that knowledge is distributed, it is equally important to understand how distributed knowledge is coordinated and used. Knowledge distributed across multiple individuals has to be somehow synchronized or linked to be used. Socially shared cognition provides the means to couple distributed knowledge to implement situated action, such as the cooperative activities of a group of experts (e.g., a heart surgery team). The following sections presents these twin themes: the socially distributed nature of cognition and socially shared cognition. 1. Cognition is Socially Distributed and Preserved Many important tasks are diYcult, time consuming, or even impossible for unaided individuals. In such cases, we lean on knowledge that exceeds the capacities of the individual. This gives rise to a number of complex problems that we solve by means of distinctive tools (artifacts). For instance, we have compasses, computers, and calculators. As discussed later, these

92

¨ N R. SEMIN ELIOT R. SMITH AND GU

tools allow us to solve problems without having to fully apprehend the knowledge that has gone into their production. All we need to know is the causal relationship between the tool and the goal we wish to achieve. For tasks that exceed the capabilities of an individual, we can use the concerted action of a number of individuals—as in performing heart surgery or navigating a large vessel. In such instances, we lean on others to solve problems. The ways in which such tasks are performed show a number of distinctive characteristics. Obviously, complex tasks that supersede individual capabilities mean that each individual in such groups or teams possesses unique and specialized knowledge. We do not need to know anything about the specialized knowledge that the other members of the team possess. However, we need to coordinate this socially distributed knowledge to execute complex tasks eYciently and eVectively. Although each team member does not need to know the specialized knowledge that is distributed, it is essential for the execution of the task that each has shared knowledge by means of which it is possible to deploy specialized knowledge. Shared knowledge is therefore akin to knowing the causal relationship between a tool and the goal that has to be achieved by the use of that tool, rather than knowing the knowledge that has gone into the shaping of that tool. For instance, a heart surgery team consists of diverse experts (the surgeon, the anesthetist, etc.), each with his or her specialist and unshared knowledge bases relevant to the performance of the operation. Yet, the team also shares knowledge about their joint activity, as well as how to coordinate and synchronize such activity. Thus, while some knowledge is distributed across diverse individuals to constitute the scaVold for joint action, other types of knowledge (when to do what, the general purpose, etc.) must be shared to coordinate and synchronize the sequential execution of diverse tasks. Interpersonal communication is of the essence for a smooth and relatively flawless execution of joint action. Cooperating individuals must frequently exchange requests, instructions, corrections, and so on to ascertain that the activities are coordinated and matched. Thus communication provides the means by which the progress of the task and the coordination of actions are kept in constant check and the sharedness of joint reality monitored. A further feature of shared knowledge is time. Because the performance of shared tasks always has a sequentially organized structure, the coordination of the activities of a number of people takes place over a temporal horizon. One example of socially distributed cognition is carefully analyzed in Hutchins’ work (1995a) on a task that exceeds the capabilities of an individual: navigating a large Navy ship. The task is cyclical and involves processing complex, socially distributed information. It is performed by several individuals who play discrete roles (reading a timepiece, identifying a landmark, communicating a bearing, etc.) and using physical and computational tools

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

93

(charts, protractors, compasses, etc.). The task is accomplished by a sequence of actions that occur as diVerent individuals coordinate their activities and draw on each other’s specialized information to establish knowledge that supersedes each individual’s unique knowledge. In all these cases (heart surgery or navigating a large vessel), the human organism interacts with external entities (other people or physical objects) to perform a task. If the task is a cognitive one (navigation, decision making, heart surgery), then we can say that cognitive processes ‘‘extend out beyond the individual’’ in the sense that the causal role played by aspects of the external environment are fully equal to the causal role played by the agent’s internal states and processes. ‘‘Remove the external component and the system’s behavioral competence will drop, just as it would if we removed part of its brain’’ (Clark & Chalmers, 1995, p. 4). The manner in which people provide external scaVolds for each other has been investigated in Wegner’s (1995) work on transactive memory systems. This work shows how memory becomes progressively specialized, socially shared, indexed, and complementary in interdependent groups. When interdependent individuals (e.g., a couple) communicate with each other and share information, they coordinate their respective memory expertise that gives rise to a qualitatively diVerent memory system. Individuals can refer to information stored by another person as a type of external memory. Put another way, the cognitive powers of the individual mind are enhanced by socially available and accessible scaVolds. One has access to the information available in another ‘‘system’’ (i.e., individual memory) by knowing that the other has the relevant item and by addressing that individual to retrieve it. Thus, this interdependent system is able to supply a more elaborate memory than that of any individual member (Wegner, 1995). Transactive memory is a system that is irreducible, operates at the group level, and depends on a distribution of specializations within this system, as in the case of partners (Wegner, 1995). Other social psychological research examining how distributed knowledge is used has been conducted with decision-making groups. This research has predominantly used groups of individuals who are not explicitly diVerentiated in terms of their knowledge or expertise. (It is important to note that this stands in contrast to the groups involved in navigating a large vessel or performing heart surgery, where individuals are socially diVerentiated and marked for their expert roles.) In undiVerentiated groups the coordination of available expertise (distributed knowledge) has been repeatedly shown not to be eYcient, due to a phenomenon known as the common knowledge eVect (Gigone & Hastie, 1993, 1997; Tindale & SheVey, 2002). These studies show that if an item of decision-making information is widely shared prior to group discussion, then this shared information has a greater impact on the

94

¨ N R. SEMIN ELIOT R. SMITH AND GU

group judgment. These and other types of studies show that under specific circumstances groups do not eVectively share all the information relevant to solving a problem (see also Stasser, 1992; 1999; Stasser & Stewart, 1992). 2. Socially Shared Nature of Cognition Deploying knowledge that is distributed across the members of a group or a team requires coordination or synchronization through ‘‘social coupling’’ in a two-way interaction (Semin, 2000a). Social coupling is achieved by means of communication, involving the use of diverse social conventions and in particular the use of language. Communication constitutes purposeful social interaction, takes place in a social context, and is regulated by social rules and conventions (e.g., regarding language use) that are deployed to establish a shared reality and to attain individual goals (e.g., Austin, 1962; Grice, 1975; Higgins, 1981, 1992; Krauss & Fussell, 1996; Searle, 1969). Synchronization between speaker and addressee requires conventions to regulate the sequential flow of conversation. Conversationalists concurrently access a set of tacit conventions or maxims. These maxims are derived from the unspoken principle of cooperation, by means of which intended meaning is achieved in communication (Grice, 1975, 1978). The roles of speaker and addressee reverse in a turn-taking process regulated by conversational conventions signaling turns (Sacks, SchegloV, & JeVerson, 1974). These are but some of the conventions that contribute to establishing socially shared meanings or realities. These conventions have evolved to regulate the speaker–addressee relationship. A substantial amount of social psychological research addresses how social coupling and shared reality are achieved. The synchronization of conversation between speaker and addressee involves monitoring the perspective of the addressee (e.g., Fussell & Krauss, 1989a,b, 1991; Krauss & Fussell, 1988; Shober, 1998), which influences how a message is shaped. Studies by Krauss and colleagues (e.g., Fussell & Krauss, 1989a,b) using a referential communication paradigm examine message design as a function of addressee characteristics. These types of studies illustrate how perspective taking influences the linguistic features of messages and how these in turn influence communicative accuracy. Considerable groundwork is necessary to achieve intersubjectivity (Rommetweit, 1974; Schu¨tz, 1962). One of the ways in which this is accomplished is by monitoring whether or not common ground is established with the addressee (e.g., Clark & Shober, 1992; Clark, Schreuder, & Buttrick, 1983). To this end a number of typically linguistic strategies are used to coordinate joint reference to objects and events in a communicative setting (e.g., Clark, 1992, 1996; Krauss & Fussell, 1996).

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

95

Language is the main vehicle in the coordination and synchronization of two-way interaction. Language is distinctive in that it constitutes a stable type of knowledge that is shared by all to establish an intersubjective (Schu¨tz, 1962) or shared reality. Whereas meaning is initially unshared and subjective, word meanings, syntactic rules, and conversational conventions must be shared to create an ‘‘objective’’ or intersubjective reality (Rommetweit, 1974; Schu¨tz, 1962) without which communication could not be accomplished. In order to be able to communicate an intention, experience, idea, wish, or desire, I must access a medium that is ‘‘objective,’’ namely shared and detached from each and every person. In Vygotski’s terms: ‘‘In order to transmit some experience or content of consciousness to another person, there is no other path than to ascribe the content to a known class, to a known group of phenomena, and as we know this necessarily requires generalization. Thus, it turns out that social interaction necessarily presupposes generalization and the development of word meaning, i.e., generalization becomes possible with the development of social interaction’’ (Vygotski, 1956/1987, pp. 48–49). As a whole, communication involves drawing on shared resources thereby enabling the coordination and synchronization of those processes by which it is possible to share or pool knowledge distributed across members of a group or a team or to convey potentially novel and unique meanings (Semin, 1995, 1996). Finally, for recurrent social or task situations (e.g., navigating a large vessel or performing heart surgery) participants have a representation of the overall cycle of activities to be performed, including the respective distributed roles and temporal ordering constraints for role-related activities. Such generalized knowledge, which has a managerial function, does not require detailed knowledge possessed by the respective specialists in a team or group. We do not need to know the workings of the post oYce to use it eVectively to send a letter to its destination, and the post oYce does not need to know the specific purpose behind our letter. Nevertheless, our roles are coupled to achieve the aim of delivering the letter at the designated address by means of diverse environmental markers/devices (see next section for detail), such as post boxes in public places, our representation of the postal system and its general function, and so forth. Many important tasks also have external agents who maintain the representation of an overall cycle of activities. These agents (captains, managers) may have no specialized function within the task, but may be responsible for coordinating distributed knowledge by prompting individual actors when necessary to perform the right act at the right moment, and so forth. Thus, diverse means are available to coordinate cognition that is distributed across agents. These range from socially shared representations (Moscovici, 1984) of specific situations (e.g., navigating a large vessel) to socially designated

96

¨ N R. SEMIN ELIOT R. SMITH AND GU

roles for diverse agents, including that of a master of ceremonies (captain or manager) who oversees the coordination of the diverse situated actions. Finally, there is extensive corrective communication to steer the course of jointly created public knowledge in the right direction (Hutchins, 1995a).

B. TAKING THE THEME FURTHER Social psychological exploration of the socially distributed nature of cognition can be advanced by incorporating additional features of distributed cognition. Cognition makes use of tools and other aspects of the individual’s environment, aside from people and groups. Moreover, to lean on people and groups, tools (e.g., language) are also needed to coordinate and synchronize social interaction. Precisely how this is achieved has not been addressed extensively within social psychology. Despite the considerable interest in socially shared cognition (e.g., Resnick, Levine & Teasley, 1991) and shared realities (Hardin & Higgins, 1996; Higgins, 1992), the ways they are established and the purposes they serve is not addressed, nor is there any consideration of the use of tools (e.g., language) by means of which individual behaviors are coordinated and synchronized. This, in fact, is a general issue for this emerging field. For instance, Caporael (1997) notes that: ‘‘The role of artifacts in distributed cognition and shared reality has not received much attention in social psychology’’ (p. 290). As she further notes, ‘‘Hardin and Higgins (1996) made virtually no mention of the role of artifacts in constituting socially shared reality’’ (p. 290). As a conceptual approach to these issues, consider that tools, in A. Clark’s (1997) terms, provide scaVolds for cognitive activity. For instance, to multiply two five-digit numbers, which most of us would not be able to perform mentally, we use paper and pencil as tools. Such tools provide scaVolds by downloading knowledge and releasing cognitive space for the performance of other practical tasks (see also Vygotsky, 1978). ScaVolds can take diVerent forms. The availability of practical tools such as hammers, scissors, and drills, provide mechanical scaVolds to achieve solutions, which would otherwise be diYcult if not impossible (e.g., by using a stone instead of a hammer). Language as a tool fulfills the same function. Take, for instance, a task-oriented group, such as the crew of a ship (cf. Hutchins, 1995a). The synchronized and coordinated communication between the diVerent crew members is intended to structure collective action such that a constantly monitored and publicly shared cognitive state is achieved. The socially distributed knowledge between crew members in its public coordination, as well as physical tools such as charts and compasses, in which knowledge is literally embodied, become scaVolds for the successful navigation of a ship.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

97

Thus, when cognitive activities are distributed across space, but also in any simple communication, language becomes an indispensable resource to structure a recipient’s cognitive properties (Semin, 2000a). This section considers issues that remain wide open for social psychological investigation, regarding the nature and cognitive properties of tools and of language. 1. Properties of Tools Tools provide a clear illustration of the distributed nature of cognition. Tools include not only physical tools (like scissors or hammers) but also what we might call computational artifacts (computers, calculators, maps, and charts), and communicative tools (language, symbolic notation for music or math). What is the significance of tools, and the ways they ease and enable cognition, in the context of socially situated cognition? The evolution of human society as it is known today is based on the invention of tools that extend our sensorimotor and cognitive abilities (Preston, 1998). It is impossible to think of current human existence without the aid of tools. Tools are products of complex cognitive eVorts to find adaptive solutions to problems. They emerge historically as standardized solutions to recurring problems, such as driving a nail into hard material, navigating a vessel over open water, or transmitting ideas or requests to other people. Tools . . . are feats of centuries of engineering, (and) are not only the products of collective experience and knowledge, they also (preserve and) represent this knowledge. These special tools contained the distilled knowledge about the relation between a task and the best fit between a task or a goal and human propensities (e.g., physical propensities such as moments, handling, and vision. . . . Tools have properties that have been engineered to optimize their use in a variety of contexts or practical domains. For instance, in the case of hammers, the tool has a shaft and a peen, a hard, solid head at a right angle to the handle; depending on their functions, hammers can also display other properties. One such property is a claw on the head for extracting nails, which typically appears on carpenter’s hammers. A tool’s properties are distinct from its aVordances, the variety of things that one can do with it or its uses. (Semin, 1998, p. 230–231, emphasis in the original)

Tools embody solutions to problems in that their properties emerge adaptively for the specific problem at hand. Tools allow socially situated cognition to take place by delegating processing demands to external aids, scaVolds, and resources. Tools also preserve the functional knowledge that shapes their structure. Thus, while a solution means finding a means to unravel a problem, the means by which this is achieved is also preserved in the tools that are created for the solution. Thus, not only the architecture of

98

¨ N R. SEMIN ELIOT R. SMITH AND GU

our minds and bodies but also external resources or tools are used by and therefore constrain human cognition. 2. Coupling Humans and Tools All tools (language, symbolic notation for music or math, scissors, hammers, and pliers, etc.) are designed for the specific problem for which they have been developed. They are practical aids to solve recurrent situated problems. By definition, all types of situated action requires a two-way interface between human embodiment and the nature of the problem, whatever it may be (e.g., cutting a piece of paper neatly, driving a nail into a piece of wood, declaring a contract null and void). Thus, scissors, hammers, pliers, and language have particular forms and properties that emerge for the type of situated problem they have been designed to resolve. They are also adapted to the corresponding parts of the human body. Scissors provide an illustration of this dual adaptation (Clark, 1997; Semin, 2000b). One part of a pair of scissors is optimally suited to the particular anatomical and motoric makeup of the human hand. The other part is perfectly designed for cutting a sheet of paper neatly. The unique quality of such tools is their twoway adaptation. Tools are dually adapted to the constraints of embodiment and to the distinctive features of the object or problem. The dual adaptation principle, although perhaps more diYcult to visualize than with hardware tools, also applies to tools such as language and symbolic systems. For instance, some argue that the reliance in early Greece on the use of ordinary language and an algebra that used letter symbols to represent unknown quantities stunted algebraic development in that civilization (Seanger, 1997). Adaptation of Arabic numerals changed this. ‘‘EVective mathematical notation allows a maximum amount of information to be unambiguously displayed in foveal and parafoveal vision’’ (Seanger, 1997, p. 132). Similarly, musical notation underwent a major change in the eleventh century after its early emergence around the eighth and nineth centuries. ‘‘Neurophysiologically, both the new musical and new textual formats presupposed a broad eye-voice span and the extensive use of parafoveal vision in the left visual field to perceive the forms of melody, words, and word groupings’’ (in Seanger, 1997, p. 141). The shift from scriptura continua (continuous writing without any spacing) to the introduction of diVerential spacing between the letters in a word and between words during the Renaissance was a crucial technological innovation that was responsible for the emergence of silent reading (Saenger, 1997). This is another instance of how specific tools (writing) are gradually adapted to human processing constraints to facilitate rapid and silent reading, a prototypically cognitive skill.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

99

Because tools are the products of dual adaptation, they carry information about the types of constraints that are introduced by our brain-body makeup. Thus, the shape of a scissors handle is an adaptation to the particular grip that is the most eYcient way of distributing pressure by the hand. The particular spacing between letters in a word and between words is informative of the facilitatory link between perceptual processing, reading, and text comprehension. As these examples illustrate, systematic examinations of tool properties can be informative about both psychological and task constraints that have contributed to the shape of the tools in the first place. Moreover, understanding the properties of tools is also highly informative about how the agent is coupled to the problem or task, because the choice of tool is indicative of the type of relationship or interdependence between agent and task. These relationships point to entire research agendas for social psychology. As one example of this type of analysis, research on the linguistic intergroup bias (Douglas & Sutton, 2003; Semin et al., 2003) demonstrates that people’s choice of linguistic forms (as a tool) reflects specific cognitive and motivational processes such as intergroup relations. 3. No Expertise Required for Preserved Knowledge As users of tools, we do not need to know anything about the knowledge that has gone into the production of the tool. All we need to know is the causal relationship between the tool and the goal we want to achieve. If we want to communicate with a long-lost friend, all we have to do is write a letter and deliver the duly stamped envelope in a mailbox, which will take it to its destination. We do not have to know the machinations of the postal system or the particular person(s) who will be handling our letter. Alternatively, an e-mail is enough. We do not have to know anything about the machinations of our computer or the Internet. Similarly, we do not have to reinvent a hammer or a pair of scissors, or language, syntax, and semantics. In other words, the complex cognitive and practical knowledge that has gone into the production of the tool is downloaded into the tool. The significance of this conclusion is that considerable cognitive load can be avoided or delegated to external devices that are manageable with minimal manipulative skills. At the same time, the very fact that such downloading takes place changes human mental processes. As Vygotsky has argued, the development of such tools changes the structure of social events. ‘‘It does this by determining the structure of a new instrumental act just as a technical tool alters the process of natural adaptation by determining the form of labor operations’’ (Vygotsky, 1978, p. 24).

100

¨ N R. SEMIN ELIOT R. SMITH AND GU

4. Methodological Corollary: Understand the Properties of Tools Since tools are used to aid cognition, their properties also constrain and influence cognition. This becomes a key methodological point from the SSC perspective. The primary tool from this viewpoint is, of course, language. Two aspects of language in particular have a great impact on situated cognition: properties of linguistic representations in themselves, and the general conventions of communication. The SSC perspective holds that cognition is distributed across inner mental processes, the tools by which cognition is implemented (e.g., language), and the environment as a scaVold (e.g., as a memory aid). Therefore properties of those tools may contribute to regularities in observed behavior—regularities that may be mistakenly attributed solely to inner processes. As a central example, the widespread use of textual stimuli for psychological research means that we are perpetually in danger of having our studies identify properties of text and mistakenly interpreting those as properties of mental representations and psychological processes. A typical illustration of this is found in the social psychological research on the causality implicit in interpersonal verbs (Brown & Fish, 1983). The general phenomenon is the following. Take two sentences, which vary only in terms of the verb—namely, ‘‘John helped David because he is kind’’ and ‘‘John likes David because he is nice.’’ The general finding is that the types of verb systematically disambiguate the ambiguous personal pronoun ‘‘he’’ to either the sentence subject (John in sentence 1) or to David (in sentence 2). A large body of social psychological literature (see Semin, 1998, 2001 for reviews) has treated this as a phenomenon that has to be explained in terms of cognitive processes. The simple fact that this phenomenon is a particular cognitive property of interpersonal verbs rather than a specific cognitive process is a much more parsimonious and situated explanation. Conceptually, this explanation implies treating language as a tool with specific properties that is used as a structuring resource for communication purposes, rather than treating language and language-driven processes as purely intraindividual phenomena (Semin, 2000b). In this situated perspective the choice of a particular verb over another becomes informative about how a speaker intends to structure the representation of an event and its reception by a listener. Another example showing that the properties of language can be mistaken for properties of psychological processes is found in a replication of Asch’s classic impression formation studies (1946). Semin (1989) used a dictionary of synonyms and antonyms to compute the semantic overlap between the diVerent stimulus sets (including warm, cold, polite, blunt) and the response scales used by participants in the studies. By developing an index of the strength of semantic associations between the composite stimulus sets and

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

101

the response items, Semin was able to show that the distinctive patterns obtained in actual Asch data could be replicated simply by using a dictionary and no subjects. There was a 90% overlap between the dictionary-derived index and the Asch data. This finding suggests that our studies are often examining regularities of the types of tools that we use, particularly language, rather than regularities related to psychological processes. To put it bluntly, as Clancey (1997a) did: If you give research participants textual stimuli, you should be prepared to find results that suggest they have texts inside their minds. 5. Language as a Tool Social psychologists might profitably undertake a more systematic examination of the properties of tools or artifacts and the role these tools (language, inter alia) play in the creation of intersubjective reality. Some of the issues raised include examining more specifically collectively shared constructs, such as collective identity or shared world views, that organize (and are organized by) lower-level activity (Caporael, 1997). The social nature of cognition is probably most evident when we use such external resources such as collectively shared constructs, but the point is equally valid when such social structures have been internalized (Vygotsky, 1986). These considerations are very akin to thinking about syntax and language use. Examined more deeply, any use of language has two aspects. One is the communication of meaning and the other is a structure that carries this meaning. Whereas the former is subjective, the latter is intersubjective—for example, in order to be able to convey meaning that is initially unshared or subjective, word meanings, syntactic rules, and the like must be shared. Language use then amounts to drawing on shared structure to convey a potentially novel and unique meaning. Language use not only transmits meaning, but in addition reproduces and reinforces the structure (syntax and semantics). The same structure can carry a whole host of communicative contents for a great variety of actors. Although the structure is determinate in terms of its properties (specified by semantic and syntactic rules), its aVordances and potential uses are indeterminate (Semin, 1998). The analogy here is between language and speech on the one hand and tool and tool use on the other (Semin, 1996). The tool metaphor in itself is not new (see Vygotski, 1978; Wertsch, 1998), nor is the distinction between language and speech (deSaussure, Mead, Riceour, inter alia). What is new in our proposal is that each and every human communication has two interrelated fundamental features, namely (1) the reproduction of a structure, without which (2) meaning could not be conveyed (Giddens, 1976). Where the tool and tool use analogy falls

102

¨ N R. SEMIN ELIOT R. SMITH AND GU

short (in particular in the writings of people from the sociocultural school, e.g., Wertsch, 1998) is that while literal tools have a real existence independent of their use, the tools of communication (language) do not have an existence independent of communication. They are reproduced in communication. In terms of social behavior, the fact that I utter a sentence in English contributes to the reproduction of English as a language. This, as Giddens notes (1979), is the unintended consequence of uttering that sentence. Language is the prime example of a social resource that structures, shapes, and regulates our internal processing. For instance, children ‘‘talk’’ to themselves silently or aloud in ways that improve their self-regulatory abilities and task performance (Clark, 1997). From time to time we all subvocally rehearse verbal self-instructions as we perform tasks that are not yet well learned. ‘‘Public speech, inner rehearsal, and the use of written and online texts are all potent tools that reconfigure the shape of computational space. Again and again we trade culturally achieved representation against individual computation. Again and again we use words to focus, clarify, transform, oZoad, and control our own thinkings. Thus understood, language is not the mere imperfect mirror of our intuitive knowledge. Rather, it is part and parcel of the mechanism of reason itself’’ (Clark, 1997, p. 207). Language is fundamental to social influence. 6. Reconceptualization of Role of Culture The SSC approach also oVers a principled understanding of the nature of culture and its relations to cognition, an increasingly important area of social psychology (Schwarz, 2000). Culture is not just a body of knowledge that is represented in the environment and learned by an individual. Rather, culture is a body of practices. ‘‘The idea that knowledge is a possession of an individual person is as limited as the idea that culture is going to the opera. Culture is pervasive; we are participating in a culture and shaping it by everything we do . . . Knowledge is pervasive in all our capabilities to participate in our society; it is not merely beliefs and theories describing what we do’’ (Clancey, 1997b, p. 271). This perspective puts culture in its true role as one of the central constraints on situated and adaptive action— and hence gives it a central place in socially situated cognition. An important defining feature of culture is given by tools and their use. Of course, tools, broadly speaking, are the distinctive feature of humans beings as a species. Importantly, tools carry cultural information and shape the historical and cultural range of things that are possible in socially situated cognition and action. As Tomasello and Rakoczy (2003) note: ‘‘. . . such things like languages, symbolic mathematics, and complex social institutions are not individual inventions arising out of humans’ extraordinary individual

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

103

brainpower, but rather they are collective cultural products created by many diVerent individuals and groups of individuals over historical time’’ (p. 121). It is therefore not surprising that cognition is historically, socially, and culturally contextualized in its situated expression. This view is shared in a number of converging schools of thought (e.g., Cole, 1996; Tomasello, 2000; Wertsch, 1998), which owe their intellectual heritage to Vygotsky (1962/ 1986). To the extent that all action takes place by means of tools that mediate between people, all action is culturally and historically situated. Reversing this line of thought suggests that a proper analysis of tools and their use can provide insights into culture (e.g., Semin & Rubini, 1990; Semin, Go¨rts, Nandram, & Semin-Goossens, 2002). The SSC approach also leads to entirely diVerent ways of conceptualizing and investigating the age-old problem of the relationship between language and thought and consequently on the influence of cultural diVerences in lexical and grammatical categories. The classic example is the following question. If two cultures linguistically code the color spectrum diVerently, do they then perceive and represent colors incommensurably? The classic take on this problem has been an individual-centered one (e.g., Hardin & Banaji, 1993; HoVman, Lau, & Johnson, 1986; Hunt & Agnoli, 1991). In contrast, a SSC approach to this problem treats lexical and grammatical categories as resources (tools) in a social context. When language is considered as a tool for action, it becomes a structuring resource for communication purposes. This changes the nature of the problem regarding the language-cognition interface dramatically, into one concerning how speakers’ language use influences listeners’ representation of some reality. Language and its use are always intended to influence thought. But the classic individual-centered representational conceptualization of the problem uses a specific methodological angle in which participants are always put into the listeners’ role (and not speakers). Methodologically, both the speaker and listener perspectives need to be taken into account if we are to understand the relationship between language and cognition by highlighting the eVects of motivated communication as a structuring resource. As a methodological corollary, one should seek cultural diVerences in situated practices, not in mental contents. The implication of this view of culture is that cultural diVerences are in the first place diVerences in repertoires of socially situated action, rather than just diVerences in inner representations. Such an emphasis is clear in the recent work by Hong et al. (2000) on their ‘‘dynamic constructivist’’ approach to culture, which argues that thinking and acting as a member of a given culture is a state rather than a stable trait. And the state is readily aVected by the situation. One of the authors of that paper who is Spanish but currently resides in the United States, ‘‘often surrounds herself with Spanish music, food, and paintings to

104

¨ N R. SEMIN ELIOT R. SMITH AND GU

keep alive her Spanish ways of thinking and feeling’’ (2000, p. 718). Michael Tomasello expresses a similar conclusion: ‘‘In cultural learning young children learn to use the tools, artifacts, symbols, and other cognitive amplifiers of their culture’’ (2000, p. 357). In other words, the focus of research on cultural diVerences should broaden from diVerences in mental contents to diVerences in cultural practices as well as the tools (including symbolic tools such as language) aVorded by diverse cultures. Like all tools, culture represents a set of two-way adaptations that both reflect and influence cognition.

V. Summary This chapter has described four basic themes of the socially situated cognition approach, which we have argued are equally basic to social psychology’s unique conceptual focus as a field. To a large extent, social psychology recognizes that cognition—including the formation of attitudes, stereotypes, person impressions, and other inner representations—is essentially for action. Our field has recognized the importance of embodiment, the role of bodily states (such as emotion and motivation) in cognition and the reciprocal influences between bodily movements and judgments (as when putting on a facial expression resembling a smile causes people to judge that cartoons are especially funny). The situated nature of cognition is also a major theme of social psychology; Asch, Sherif, Milgram, and other researchers intuitively grasped the power of social situations to control action, generating behavior that would not have been predicted from an abstracted (nonsituated) inventory of their participants’ beliefs, values, and attitudes. Finally, the distributed nature of cognition, and the importance of shared reality when people interact in dyads, groups, or larger cultural communities, is also a significant area of ongoing social psychological investigation. However, despite its general compatibility with social psychological theories and perspectives, the SSC perspective calls for some shifts of conceptual focus and of research methodology for our field. Social psychology has not yet fully appreciated the implications of the shift from computation to biology as a metaphor and framework for understanding cognition (Barsalou, 1999a, p. 77; Caporael, 1997; Fiske, 1992). Psychologists implicitly adopt a computational perspective when they try to explain social behavior (or any other kind) solely on the basis of the agent’s inner representations, processes, beliefs, and goals, through the conceptual lens of information processing. As an alternative, the biological perspective carries the insight that cognition is for action, and that embodiment and the situated nature of adaptive action are crucial constraints. We have noted ways that the four themes of action–orientation,

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

105

embodiment, situatedness, and distributed cognition–suggest further areas for conceptual and empirical exploration, areas that are intrinsically social psychological in terms of the definition of our field but that we have barely begun to explore. These explorations are important, for as we have implied, the field of social psychology can almost be defined as the study of socially situated cognition. Or, in other words, the socially situated cognition perspective selects social psychology as the central standpoint from which the entire body of the cognitive sciences makes conceptual sense. Most if not all of human cognition is social, for we are intrinsically interdependent agents, who rely on other people, use tools such as language, and make use of cognitive abilities shaped through evolution in social contexts (Caporael, 1997). As Clancey (1997a, p. 366) stated, the ‘‘overarching content of thought is not scientific models [descriptions, symbolic representations of states of the world], but coordination of an identity’’ in a social context. Although social psychologists have already investigated much of this territory, we have done so without an overall map. The socially situated cognition perspective can put these seemingly disparate areas of study into conceptual relationship to each other. The approach will also help researchers in adjoining areas—cognitive scientists, cognitive and developmental psychologists, and the like—understand why social psychology is important, even central to the analysis of all human behavior.

Acknowledgments Supported by grants from the National Institute of Mental Health, No. K02 MH01178; from the National Science Foundation, No. 0091807-BCS; and from the Free University Amsterdam and NWO (Dutch Science Foundation), No. 575-31-008. We thank to Chris Agnew, Ximena Arriaga, Robert Krauss, Diane Mackie, and Laura Sweeney for extensive comments on earlier versions of this paper.

References Agnew, C. R., Van Lange, P. A. M., Rusbult, C. E., & Langston, C. A. (1998). Cognitive interdependence: Commitment and the mental representation of close relationships. Journal of Personality and Social Psychology, 74, 939–954. Agre, P. E. (1997). Computation and human experience. New York: Cambridge University Press. Agre, P. E., & Chapman, D. (1990). What are plans for? In P. Maes (Ed.), New architectures for autonomous agents: Task-level decomposition and emergent functionality (pp. 17–34). Cambridge, MA: MIT Press.

106

¨ N R. SEMIN ELIOT R. SMITH AND GU

Alexander, C. N., & Knight, G. W. (1971). Situated identities and social psychological experimentation. Sociometry, 34, 65–82. Allport, G. W. (1954). The historical background of modern social psychology. In G. Lindzey (Ed.) The handbook of social psychology (Vol. 1, pp. 3–56). Cambridge, MA: AddisonWesley. Almassy, N., Edelman, G. M., & Sporns, O. (1998). Behavioral constraints in the development of neuronal properties: A cortical model embedded in a real-world device. Cerebral Cortex, 8, 346–361. Ambady, N., Bernieri, F. J., & Richeson, J. A. (2000). Toward a histology of social behavior: Judgmental accuracy from thin slices of the behavioral stream. In M. P. Zanna (Ed.) Advances in experimental social psychology (Vol. 32, pp. 201–271). San Diego, CA: Academic Press. Asch, S. E. (1946). Forming impressions of personality. Journal of Abnormal and Social Psychology, 41, 258–290. Asch, S. E. (1955). Studies of independence and conformity: A minority of one against a unanimous majority. Psychology Monographs, 70, 1–70. Austin, J. L. (1962). How to do things with words. Oxford: Oxford University Press. Baldwin, M. W. (1992). Relational schemas and the processing of social information. Psychological Bulletin, 112, 461–484. Ballard, D. H., Hayhoe, M. M., Pook, P. K., & Rao, R. P. N. (1997). Deictic codes for the embodiment of cognition. Behavioral and Brain Sciences, 20, 723–767. Bargh, J. A., Chen, M., & Burrows, L. (1996). Automaticity of social behavior: Direct effects of trait construct and stereotype activation on action. Journal of Personality and Social Psychology, 71, 230–244. Barsalou, L. W. (1999a). Language comprehension: Archival memory or preparation for situated action? Discourse Processes, 28, 61–80. Barsalou, L. W. (1999b). Perceptual symbol systems. Behavioral and Brain Sciences, 22, 577–660. Beer, R. D. (1995). Computational and dynamic languages for autonomous agents. In R. F. Port & T. van Gelder (Eds.), Mind as motion: Explorations in the dynamics of cognition (pp. 121–147). Cambridge, MA: The MIT Press. Bekkering, H., & Neggers, S. F. W. (2002). Visual search is modulated by action intentions. Psychological Science, 13, 370–374. Bem, D. J. (1967). Self-perception: An alternative interpretation of cognitive dissonance phenomena. Psychological Review, 24, 183–200. Billig, M., & Tajfel, H. (1963). Social categorization and similarity in intergroup behavior. European Journal of Social Psychology, 3, 27–52. Blair, I. V. (2002). The malleability of automatic stereotypes and prejudice. Personality and Social Psychology Review, 6, 242–261. Blair, I. V., Judd, C. M., Sadler, M. S., & Jenkins, C. (2002). The role of Afrocentric features in person perception: Judging by features and categories. Journal of Personality and Social Psychology, 83, 5–25. Bless, H. (2000). The interplay of affect and cognition: The mediating role of general knowledge structures. In J. P. Forgas (Ed.), Feeling and thinking: The role of affect in social cognition (pp. 201–222). New York: Cambridge University Press. Bless, H., & Schwarz, N. (1999). Sufficient and necessary conditions in dual-process models: The case of mood and information processing. In S. Chaiken & Y. Trope (Eds.), Dual-process theories in social psychology (pp. 423–440). New York: Guilford. Blumer, H. (1969). Symbolic interactionism: Perspective and method. Englewood-Cliffs, NJ: Prentice-Hall.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

107

Bodenhausen, G. V. (1990). Stereotypes as judgmental heuristics: Evidence of circadian variations in discrimination. Psychological Science, 1, 319–322. Boroditsky, L., & Ramscar, M. (2002). The roles of body and mind in abstract thought. Psychological Science, 13, 185–189. Bourhis, R. Y., & Giles, H. (1977). The language of intergroup distinctiveness. In H. Giles (Ed.), Language, ethnicity and intergroup relations (pp. 119–135). London: Academic Press. Bourhis, R. Y., Giles, H., Leyens, J. P., & Tajfel, H. (1979). Psycholinguistic distinctiveness: Language divergence in Belgium. In H. Giles & R. St. Clair (Eds.), Language and social psychology (pp. 158–185). Oxford: Blackwell. Braitenberg, V. (1984). Vehicles. Cambridge, MA: MIT Press. Brewer, M. B. (1991). The social self: On being the same and different at the same time. Personality and Social Psychology Bulletin, 17, 475–482. Brewer, M. B., & Harasty, A. S. (1996). Seeing groups as entities: The role of perceiver motivation. In R. M. Sorrentino & E. T. Higgins (Eds.) Handbook of motivation and cognition: The interpersonal context (Vol. 3, pp. 347–370). New York: Guilford Press. Brooks, R. (1999). Cambrian intelligence. Cambridge, MA: MIT Press. Brooks, R. A. (1986/1999). A robust layered control system for a mobile robot. In R. Brooks (Ed.), Cambrian intelligence (pp. 3–26). Cambridge, MA: MIT Press. (Original publication, 1986). Brooks, R. A. (1991). Intelligence without reason. Cambridge, MA: MIT AI Lab Memo 1293. Brown, R., & Fish, D. (1983). The psychological causality implicit in language. Cognition, 14, 237–273. Buller, D. B., & Anne, R. K. (1988). The effects of vocalics and nonverbal sensitivity on compliance: A speech accommodation theory explanation. Human Communication Research, 14, 301–332. Cacioppo, J. T., Priester, J. R., & Berntson, G. G. (1993). Rudimentary determinants of attitudes: II. Arm flexion and extension have differential effects on attitudes. Journal of Personality and Social Psychology, 65, 5–17. Caporael, L. (1997). The evolution of truly social cognition: The core configurations model. Personality and Social Psychology Review, 1, 276–298. Carlsmith, J. M., Collins, B. E., & Helmreich, R. L. (1966). Studies in forced compliance: I. The effect of pressure for compliance on attitude change produced by face-to-face roleplaying and anonymous essay writing. Journal of Personality and Social Psychology, 4, 1–13. Carlston, D. E. (1994). Associated systems theory: A systematic approach to cognitive representations of persons. In R. S. Wyer & T. K. Srull (Eds.) Advances in social cognition (Vol. 7, pp. 1–78). Hillsdale, NJ: Lawrence Erlbaum Associates. Chaiken, S., Liberman, A., & Eagly, A. H. (1989). Heuristic and systematic information processing within and beyond the persuasion context. In J. S. Uleman & J. A. Bargh (Eds.), Unintended thought (pp. 212–252). New York: Guilford. Chapman, D. (1991). Vision, instruction, and action. Cambridge, MA: MIT Press. Chartrand, T. L., & Bargh, J. A. (1999). The chameleon effect: The perception-behavior link and social interaction. Journal of Personality and Social Psychology, 76, 893–910. Clancey, W. J. (1995). A Boy Scout, Toto, and a bird. In L. Steels & R. Brooks (Eds.), The artificial life route to artificial intelligence: Building embodied, situated agents (pp. 227–236). Hillsdale, NJ: Lawrence Erlbaum Associates. Clancey, W. J. (1997a). Situated cognition: On human knowledge and computer representations. New York: Cambridge University Press. Clancey, W. J. (1997b). The conceptual nature of knowledge, systems, and activity. In P. J. Feltovich, K. M. Ford & R. R. Hoffman (Eds.), Expertise in context (pp. 247–291). MIT Press.

108

¨ N R. SEMIN ELIOT R. SMITH AND GU

Clancey, W. J. (1999). Conceptual coordination: How the mind orders experience in time. Hillsdale, NJ: Lawrence Erlbaum Associates. Clark, A. (1999a). Minds, brains, and tools. Unpublished manuscript. Department of Philosophy, Washington University, St. Louis. Clark, A. (1999b). Where brain, body, and world collide. Cognitive Systems Research, 1, 5–17. Clark, A. (1997). Being there. Cambridge, MA: MIT Press. Clark, A., & Chalmers, D. (1995). The extended mind. Unpublished, Department of Philosophy, Washington University, St. Louis. Clark, H. H. (1992). Arenas of language use. Chicago: University of Chicago Press. Clark, H. H. (1996). Using language. New York: Cambridge University Press. Clark, H. H., & Shober, M. F. (1992). Asking questions and influencing answers. In J. M. Tamur (Ed.), Questions about questions (pp. 15–48). New York: Russell Sage. Clark, H. H., Schreuder, R., & Buttrick, S. (1983). Common ground and the understanding of demonstrative reference. Journal of Verbal Learning and Verbal Behavior, 22, 245–258. Cole, M. (1996). Cultural psychology: A once and future discipline. Cambridge, MA: Harvard. Cooper, J., & Fazio, R. H. (1984). A new look at dissonance theory. In L. Berkowitz (Ed.) Advances in experimental social psychology (Vol. 17, pp. 229–266). New York: Academic Press. Cosmides, L. (1989). The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Wason Selection Task Cognition, 31, 187–276. Coupland, N. (1984). Accommodation at work: Some phonological data and their implications. International Journal of the Sociology of Language, 46, 49–70. Coupland, J., & Giles, H. (1991). Language: Contexts and consequences. Pacific Grove, CA: Brooks/Cole. Crocker, J. (1999). Social stigma and self-esteem: Situations construction of self-worth. Journal of Experimental Social Psychology, 35, 89–107. Crocker, J., Voelkl, K., Testa, M., & Major, B. (1991). Social stigma: The affective consequences of attributional ambiguity. Journal of Personality and Social Psychology, 60, 218–228. Damasio, A. (1994). Descartes’ error. New York: Grossett Putnam. Devine, P. G., Hamilton, D. L., & Ostrom, T. M. (1994). Social cognition: Impact on social psychology. Orlando, FL: Academic Press. Dewey, J. (1929/1958). Experience and Knowledge. New York: Dover. Douglas, K. M., & Sutton, R. M. (2003). Effects of communication goals and expectancies on language abstraction. Journal of Personality and Social Psychology, 84, 682–696. Dunning, D., & Cohen, G. L. (1992). Egocentric definitions of traits and abilities in social judgment. Journal of Personality and Social Psychology, 63, 341–355. Edelman, G. (1987). Neural Darwinism. New York: Basic Books. Fazio, R. H., Effrein, E. A., & Falender, V. J. (1981). Self-perceptions following social interaction. Journal of Personality and Social Psychology, 41, 232–241. Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). On the automatic activation of attitudes. Journal of Personality and Social Psychology, 50, 229–238. Festinger, L. (1950). Informal social communication. Psychological Review, 57, 271–282. Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press. Festinger, L., & Carlsmith, J. M. (1959). Cognitive consequences of forced compliance. Journal of Abnormal and Social Psychology, 58, 203–210. Fiske, A. P. (1991). Structures of social life. New York: Free Press. Fiske, A. P., & Haslam, N. (1996). Social cognition is thinking about relationships. Current Directions in Psychological Science, 5, 143–148.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

109

Fiske, A. P., Haslam, N., & Fiske, S. T. (1991). Confusing one person with another: What errors reveal about the elementary forms of social relations. Journal of Personality and Social Psychology, 60, 656–674. Fiske, S. T. (1992). Thinking is for doing: Portraits of social cognition from daguerreotype to laserphoto. Journal of Personality and Social Psychology, 63, 877–889. Fiske, S. T. (1993). Controlling other people: The impact of power on stereotyping. American Psychologist, 48, 621–628. Foerster, J., & Strack, F. (1996). Influence of overt head movements on memory for valenced words: A case of conceptual-motor compatibility. Journal of Personality and Social Psychology, 71, 421–430. Franklin, S. P. (1995). Artifical minds. Cambridge, MA: The MIT Press. Frick-Horbury, D. (2002). The effects of hand gestures on verbal recall as a function of highand low-verbal-skill levels. Journal of General Psychology, 129, 137–147. Frick-Horbury, D., & Guttentag, R. E. (1998). The effects of restricting hand gesture production on lexical retrieval and free recall. American Journal of Psychology, 111, 43–62. Friedman, R. S., & Forster, J. (2000). The effects of approach and avoidance motor actions on the elements of creative insight. Journal of Personality and Social Psychology, 79, 477–492. Friedman, R. S., & Forster, J. (2002). The influence of approach and avoidance motor actions on creative cognition. Journal of Experimental Psychology, 38, 41–55. Fussell, S. R., & Krauss, R. M. (1989a). The effects of intended audience on message production and comprehension: Reference in a common ground framework. Journal of Experimental Social Psychology, 25, 203–219. Fussell, S. R., & Krauss, R. M. (1989b). Understanding friends and strangers: The effects of audience design on message comprehension. European Journal of Social Psychology, 19, 509–526. Fussell, S. R., & Krauss, R. M. (1991). Understanding friends and strangers: The effects of audience design on message comprehension. European Journal of Social Psychology, 21, 509–525. Gibson, J. J. (1966). The senses considered as perceptual systems. Boston: Houghton Mifflin. Gibson, J. J. (1979). The ecological approach to visual perception. Boston: Houghton Mifflin. Giddens, A. (1976). New rules of sociological method. London: Hutchinson. Giddens, A. (1979). Central problems in social theory. London: Macmillan Press. Gigone, D., & Hastie, R. (1993). The common knowledge effect: Information sharing and group judgment. Journal of Personality and Social Psychology, 65, 959–974. Gigone, D., & Hastie, R. (1997). The impact of information on small group choice. Journal of Personality and Social Psychology, 72, 132–140. Giles, H., & Wiemann, J. M. (1987). Language, social comparison, and power. In C. R. Berger & S. H. Chaffee (Eds.), Handbook of communication sciences (pp. 350–384). Newbury Park, CA: Sage. Giles, H., Mulac, A., Bradac, J. J., & Johnson, P. (1987). Speech accommodation theory: The next decade and beyond. Communication Yearbook, 10, 13–48. Glenberg, A. M. (1997). What memory is for. Behavioral and Brain Sciences, 20, 1–55. Glenberg, A. M., & Robertson, D. A. (1999). Indexical understanding of instructions. Discourse Processes, 28, 1–26. Goffman, I. (1959). The presentation of self in everyday life. Garden City, NY: Doubleday. Goffman, I. (1961). Encounters: Two studies in the sociology of interaction. Indianapolis: BobbsMerill. Goldstone, R. L., & Barsalou, L. W. (1998). Reuniting perception and conception. Cognition, 65, 231–262.

110

¨ N R. SEMIN ELIOT R. SMITH AND GU

Gollwitzer, P. M. (1990). Action phases and mind-sets. In E. T. Higgins & R. M. Sorrentino (Eds.) Handbook of motivation and cognition: Foundations of social behavior (Vol. 2, pp. 53–92). New York: Guilford Press. Gollwitzer, P. M. (1996). The volitional benefits of planning. In P. M. Gollwitzer & J. A. Bargh (Eds.), The psychology of action: Linking cognition and motivation to behavior (pp. 287–312). New York, NY: Guilford Press. Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54, 493–503. Gollwitzer, P. M. & Bargh, J. A. (Eds.) (1996). The psychology of action: Linking cognition and motivation to behavior. New York: Guilford Press. Greeno, J. G. (1998). The situativity of knowing, learning, and research. American Psychologist, 53, 5–26. Grice, H. P. (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.) Syntax and semantics (Vol. 3, pp. 44–58). New York: Academic Press. Grice, H. P. (1978). Further notes on logic and conversation. In P. Cole & J. L. Morgan (Eds.), Syntax and semantics: 9, Pragmatics (pp. 113–128). New York: Academic Press. Hardin, C. D., & Higgins, E. T. (1996). Shared reality: How social verification makes the subjective objective. In R. M. Sorrentino & E. T. Higgins (Eds.) Handbook of motivation and cognition (Vol. 3, pp. 28–84). New York: Guilford Press. Hardin, C., & Banaji, M. R. (1993). The influence of language on thought. Social Cognition, 11, 277–308. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley. Heider, F., & Simmel, M. (1944). An experimental study of apparent behavior. American Journal of Psychology, 57, 243–249. Hendriks-Jansen, H. (1996). Catching ourselves in the act: Situated activity, interactive emergence, evolution, and human thought. Cambridge, MA: MIT Press. Higgins, E. T., & Semin, G. R. (2001). Communication and social psychology. In N. J. Smelser & B. Baltes (Eds.), International encyclopedia of the social & behavioral sciences (pp. 2296–2299). Oxford: Pergamon. Higgins, E. T. (1981). The ‘‘communication game’’: Implications for social cognition and persuasion In E. T. Higgins, C. P. Herman & M. P. Zanna (Eds.), Social cognition: The Ontario symposium (pp. 343–392). Hillsdale, NJ: Lawrence Erlbaum. Higgins, E. T. (1992). Achieving ‘‘shared reality’’ in the communication game: A social action that creates meaning Journal of Language and Social Psychology, 11, 107–125. Higgins, E. T., & McCann, C. D. (1984). Social encoding and subsequent attitudes, impressions, and memory: ‘‘Context-driven’’ and motivational aspects of processing Journal of Personality and Social Psychology, 47, 26–39. Higgins, E. T., & Rholes, W. S. (1978). Saying is believing: Effects of message modification on memory and liking for the person described. Journal of Experimental Social Psychology, 14, 363–378. Higgins, E. T., & Sorrentino, R. M. (Eds.) (1990). Handbook of motivation and cognition: Foundations of social behavior (Vol. 2). New York, NY: The Guilford Press. Higgins, E. T., McCann, C. D., & Fondacaro, R. (1982). The ‘‘communication game’’: Goaldirected encoding and cognitive consequences Social Cognition, 1, 21–37. Hoffman, C., Lau, I. J., & Johnson, D. R. (1986). The linguistic relativity of person cognition: An English-Chinese comparison. Journal of Personality and Social Psychology, 51, 1097–1105. Holmes, J. (2000). Social relationships: The nature and function of relational schemas. European Journal of Social Psychology, 30, 447–497.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

111

Hong, Y., Morris, M. W., Chiu, C., & Benet-Martinez, V. (2000). Multicultural minds. American Psychologist, 55, 709–720. Hunt, E., & Agnoli, F. (1991). The Whorfian hypothesis: A cognitive psychology perspective. Psychological Review, 98, 377–389. Hutchins, E. (1995a). Cognition in the wild. Cambridge, MA: MIT Press. Hutchins, E. (1995b). How a cockpit remembers its speeds. Cognitive Science, 19, 265–288. Isen, A. M. (1987). Positive affect, cognitive processes, and social behavior. In L. Berkowitz (Ed.) Advances in experimental social psychology (Vol. 20, pp. 203–253). New York: Academic Press. James, W. (1890). Psychology. New York: Holt. Johnson-Laird, P. N. (1983). Mental models. Cambridge, MA: Harvard University Press. Jones, E. E. (1985). Major developments in social psychology during the last five decades. In G. Lindzey & E. Aronson (Eds.) Handbook of social psychology (Vol. I, pp. 47–107). New York: Random House. Jones, E. E., Rhodewalt, F., Berglas, S., & Skelton, J. A. (1981). Effects on strategic selfpresentation on subsequent self-esteem. Journal of Personality and Social Psychology, 41, 407–421. Kenny, D. A., & Albright, L. (1987). Accuracy in interpersonal perception: A social relations analysis. Psychological Bulletin, 102, 390–402. Kerr, N. L., Niedermeier, K. E., & Kaplan, M. F. (2000). On the virtues of assuming minimal differences in information processing in individuals and groups. Group Processes and Intergroup Relations, 3, 203–217. Kirsh, D. (1995). The intelligent use of space. Artificial Intelligence, 73, 31–68. Kirshner, D., & Whitson, J. A. (1997). Situated cognition: Social, semiotic, and psychological perspectives. Mahwah, NJ: Lawrence Erlbaum. Kosslyn, S. M. (1994). Image and brain: The resolution of the imagery debate. Cambridge, MA: The MIT press. Krauss, R. M., & Fussell, S. R. (1988). Other-relatedness in language processing: Discussion and comments. Journal of Language and Social Psychology, 7, 263–279. Krauss, R. M., & Fussell, S. R. (1996). Social psychological models of interpersonal communication. In E. T. Higgins & A. W. Kruglanski (Eds.), Social psychology: Handbook of basic principles (pp. 655–701). New York: The Guilford Press. Krauss, R. M., Freyburg, R., & Morsella, E. (2002). Inferring speaker’s physical attributes from their voices. Journal of Experimental Social Psychology, 38, 618–625. Kruglanski, A. W. (1996). Motivated social cognition: Principles of the interface. In E. T. Higgins & A. W. Kruglanski (Eds.), Social psychology: Handbook of basic principles (pp. 493–520). New York, NY: The Guilford Press. Lakoff, G. (1987). Women, fire, and dangerous things: What categories reveal about the mind. Chicago: University of Chicago Press. Lakoff, G., & Johnson, M. (1999). Philosophy in the flesh. New York: Basic Books. Latour, B. (1986). Visualization and cognition: Thinking with eyes and hands. Knowledge and society: Studies in the sociology of culture past and present, 8, 1–40. Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral participation. New York: Cambridge University Press. Lerner, J. S., & Keltner, D. (2000). Beyond valence: Toward a model of emotion-specific influences on judgement and choice. Cognition and Emotion, 14, 473–493. Levin, H., & Lin, T. (1988). An accommodating witness. Language and Communications, 8, 195–198.

112

¨ N R. SEMIN ELIOT R. SMITH AND GU

Levine, J. M., Resnick, L. B., & Higgins, E. T. (1993). Social foundations of cognition. Annual Review of Psychology, 44, 585–612. Linville, P. W., Fisher, G. W., & Salovey, P. (1989). Perceived distributions of the characteristics of ingroup and outgroup members. Journal of Personality and Social Psychology, 57, 165–188. Livingston, R. W., & Brewer, M. B. (2002). What are we really priming? Cue-based versus category-based processing of facial stimuli Journal of Personality and Social Psychology, 82, 5–18. Maass, A. (1999). Linguistic intergroup bias: stereotype perpetuation through language. In M. P. Zanna (Ed.) Advances in Experimental Social Psychology (Vol. 31, pp. 79–122). San Diego, CA: Academic Press. Markus, H., & Kunda, Z. (1986). Stability and malleability of the self-concept. Journal of Personality and Social Psychology, 51, 858–866. Marr, D. (1982). Vision: A computational investigation into the human representation and processing of visual information. San Francisco: Freeman. Marsh, R. L., Hicks, J. L., & Bink, M. L. (1998). Activation of completed, uncompleted, and partially completed intentions. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 350–361. Mataric, M. (1991). Navigating with a rat brain: A neurobiologically-inspired model for robot spatial representation. In J.-A. Meyer & S. W. Wilson (Eds.), From animals to animats (pp. 169–175). Cambridge, MA: MIT Press. Maturana, H., & Varela, F. (1987). The tree of knowledge: The biological roots of human understanding. Boston: New Science Library. McCann, C. D., Higgins, E. T., & Fondacaro, R. A. (1991). Primacy and recency in communication and self-persuasion: How successive audiences and multiple encodings influence subsequent evaluative judgments. Social Cognition, 9, 47–66. McCulloch, W. S. (1988). Embodiments of mind. Cambridge, MA: MIT Press. McGuire, W. J., & McGuire, C. V. (1986). Journal of Personality and Social Psychology. McGuire, W. J., & McGuire, C. V. (1980). Salience of handedness in the spontaneous selfconcept. Perceptual and Motor Skills, 50, 3–7. McGuire, W. J., & McGuire, C. V. (1981). The spontaneous self-concept as affected by personal distinctiveness. In M. D. Lynch, A. Norem-Hebeisen & K. Gergen (Eds.), The self-concept. New York: Ballinger. McGuire, W. J., McGuire, C. V., & Cheever, J. (1986). The self in society: Effects of social context on the sense of self. British Journal of Social Psychology, 25, 259–270. McGuire, W. J., McGuire, C. V., Child, P., & Fujioka, T. (1978). Salience of ethnicity in the spontaneous self-concept as a function of one’s ethnic distinctiveness in the social environment. Journal of Personality and Social Psychology, 36, 511–520. McGuire, W. J., & Padawer-Singer, A. (1976). Trait salience in the spontaneous self-concept. Journal of Personality and Social Psychology, 33, 743–754. Mead, G. H. (1934). Mind, self, and society. Chicago: University of Chicago Press. Moscovici, S. (1984). Social representations. In R. Farr & S. Moscovici (Eds.), Social Representations (pp. 3–38). Cambridge: Cambridge University Press. Mullen, B., Brown, R., & Smith, C. (1992). Ingroup bias as a function of salience, relevance, and status: An integration. European Journal of Social Psychology, 22, 103–122. Murray, S. L., Holmes, J. G., Dolderman, D., & Griffin, D. W. (2000). What the motivated mind sees: Comparing friends’ perspectives to married partners’ views of each other. Journal of Experimental Social Psychology, 36, 600–620.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

113

Murray, S. L., Holmes, J. G., & Griffin, D. W. (1996). The benefits of positive illusions: Idealization and the construction of satisfaction in close relationships. Journal of Personality and Social Psychology, 70, 79–98. Neuberg, S. L., & Fiske, S. T. (1987). Motivational influences on impression formation: Outcome dependency, accuracy-driven attention, and individuating processes. Journal of Personality and Social Psychology, 55, 431–444. Neumann, R., & Strack, F. (2000). Experiential and nonexperiential routes of motor influence on affect and evaluation. In H. Bless & J. P. Forgas (Eds.), The message within: Role of subjective experience in social cognition and behavior (pp. 52–68). Philadelphia: Psychology Press. Newell, A., & Simon, H. (1972). Human problem solving. Englewood Cliffs, NJ: Prentice-Hall. Newtson, D., & Enquist, G. (1976). The perceptual organization of ongoing behavior. Journal of Experimental Social Psychology, 12, 436–450. Niedenthal, P. M., Halberstadt, J. B., & Innes-Ker, A. H. (1999). Emotional response categorization. Psychological Review, 106, 337–361. Nisbett, R. E., & Ross, L. (1980). Human inference: Strategies and shortcomings of social judgment. Englewood Cliffs, NJ: Prentice-Hall. Norenzayan, A., & Schwarz, N. (1999). Telling what they want to know: Participants tailor causal attributions to researchers’ interests. European Journal of Social Psychology, 29, 1011–1020. Norman, D. A. (1993). Cognition in the head and in the world: An introduction to the special issue on situated action. Cognitive Science, 17, 1–6. Opotow, S. (1990). Moral exclusion and injustice: An introduction. Journal of Social Issues, 46(1), 1–20. Orne, M. T. (1962). On the social psychology of the psychological experiment: With particular reference to demand characteristics and their implications. American Psychologist, 17, 776–783. Ostrom, T. M. (1984). The sovereignty of social cognition. In R. S. Wyer & T. K. Srull (Eds.) Handbook of social cognition (Vol. 1, pp. 1–38). Hillsdale, NJ: Lawrence Erlbaum. Peeters, G. (1991). Evaluative inference in social cognition: The roles of direct versus indirect evaluation and positive-negative asymmetry. European Journal of Social Psychology, 21, 131–146. Pfeifer, R., & Scheier, C. (1997). Sensory-motor coordination: The metaphor and beyond. Robotics and Autonomous Systems, 20, 157–178. Pfeifer, R., & Scheier, C. (1999). Understanding intelligence. Cambridge, MA: MIT Press. Picard, R. (1997). Affective computing. Cambridge, MA: MIT Press. Pichert, J. W., & Anderson, R. C. (1977). Taking different perspectives on a story. Journal of Educational Psychology, 69, 309–315. Preston, B. (1998). Cognition and tool use. Mind and Language, 13, 513–547. Prinz, W. (1990). A common coding approach to perception and action. In O. Neumann & W. Prinz (Eds.), Relationships between perception and action (pp. 167–201). Berlin, Germany: Springer-Verlag. Perdue, C. W., Dovidio, J. F., Gurtman, M. B., & Tyler, R. B. (1990). Us and them: Social categorization and the process of intergroup bias. Journal of Personality and Social Psychology, 59, 475–486. Putnam, W., & Street, R. (1984). The conception and perception of noncontent speech performance: Implications for speech accommodation theory. International Journal of the Sociology of Language, 46, 97–114.

114

¨ N R. SEMIN ELIOT R. SMITH AND GU

Quinn, D. M., & Crocker, J. (1999). When ideology hurts: Effects of belief in the Protestant ethic and feeling overweight on the psychological well-being of women. Journal of Personality and Social Psychology, 77, 402–414. Resnick, L. B., Levine, J. M., & Teasley, S. D. (1991). Perspectives on socially shared cognition. Washington, DC: APA Books. Rhodewalt, F., & Agustsdottir, S. (1986). Effects on self-presentation on the phenomenal self. Journal of Personality and Social Psychology, 50, 47–55. Rizzolatti, G., & Arbib, M. A. (1998). Language within our grasp. Trends in Neurosciences, 21, 188–194. Rommetweit, R. (1974). On message structure: A framework for the study of language and communication. New York: Wiley. Rosenthal, R., & Rosnow, R. L. (1969). Artifact in behavioral research. New York: Academic Press. Ruscher, J. B., & Duval, L. L. (1998). Multiple communicators with unique target information transmit less stereotypical impressions. Journal of Personality and Social Psychology, 74, 329–344. Sacks, H., Schegloff, E. A., & Jefferson, G. (1974). A simplest systematics for the organization of turn-taking for conversation. Language, 50, 696–735. Schank, R. C., & Abelson, R. P. (1977). Scripts, plans, goals, and understanding. Hillsdale, NJ: Lawrence Erlbaum. Schlenker, B. R., Dlugolecki, D. W., & Doherty, K. (1994). The impact of self-appraisals and behavior—the power of public commitment. Personality and Social Psychology Bulletin, 20, 20–33. Schu¨tz, A. (1962). Collected papers. The Hague, Netherlands: Nijhof. Schwarz, N. (2000). Social judgment and attitudes: Warmer, more social, and less conscious. European Journal of Social Psychology, 30, 149–176. Schwarz, N. (1994). Judgment in a social context: Biases, shortcomings, and the logic of conversation. Advances in Experimental Social Psychology, 26, 123–162. Schwarz, N. (1996). Cognition and communication: Judgmental biases, research methods, and the logic of conversation. Hillsdale, NJ: Erlbaum. Seanger, P. (1997). Space between words: The origins of silent reading. Stanford, CA: Stanford University Press. Searle, J. R. (1969). Speech acts. Cambridge, MA: Cambridge University Press. Semin, G. R. (1989). Impressions of personality revisited: The contribution of linguistic factors to attribute inferences. European Journal of Social Psychology, 19, 85–101. Semin, G. R. (1995). Interfacing language and social cognition. Journal of Language and Social Psychology, 14, 182–194. Semin, G. R. (1996). The relevance of language for social psychology. In C. McGarty & A. Haslam (Eds.), The message of social psychology: Perspectives on mind and society (pp. 291–304). Oxford: Blackwell. Semin, G. R. (1998). Cognition, language, and communication. In S. R. Fussell & R. J. Kreuz (Eds.), Social and cognitive psychological approaches to interpersonal communication (pp. 229–257). Hillsdale, NJ: Lawrence Erlbaum. Semin, G. R. (2000a). Language as a cognitive and behavioral structuring resource: Questionanswer exchanges. In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology (pp. 75–104). Chichester: Wiley. Semin, G. R. (2000b). Agenda 2000: Communication: Language as an implementational device for cognition. European Journal of Social Psychology, 30, 595–612.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

115

Semin, G. R. (2001). Language and social cognition. In A. Tesser & N. Schwarz (Eds.) Handbook of social psychology (Vol. 1, pp. 159–180). Intraindividual Processes. Oxford: Blackwell. Semin, G. R., de Montes, G. L., & Valencia, J. F. (2003). Communication constraints on the Linguistic Intergroup Bias. Journal of Experimental Social Psychology, 39, 142–148. Semin, G. R., Go¨rts, C., Nandram, S., & Semin-Goossens, A. (2002). Cultural perspectives on the linguistic representation of emotion and emotion events. Cognition and Emotion, 16, 11–28. Semin, G. R., & Rubini, M. (1990). Unfolding the category of person by verbal abuse. European Journal of Social Psychology, 20, 463–474. Sherif, M. (1936). The psychology of social norms. New York: Harper. Shober, M. F. (1998). Different kinds of conversational perspective taking. In S. R. Fussell & R. J. Kreuz (Eds.), Social and cognitive approaches to interpersonal communication (pp. 145–174). Hillsdale NJ: Laurence Erlbaum. Simpson, J. A., & Kenrick, D. T. (Eds.). (1997). Evolutionary social psychology . Mahwah, NJ: Lawrence Erlbaum. Sinclair, L., & Kunda, Z. (1999). Reactions to a black professional: Motivated inhibition and activation of conflicting stereotypes. Journal of Personality and Social Psychology, 77, 885–904. Singer, J. A., & Salovey, P. (1988). Mood and memory: Evaluating the network theory of affect. Clinical Psychology Review, 8, 211–251. Skowronski, J. J., & Carlston, D. E. (1989). Negativity and extremity biases in impression formation: A review of explanations. Psychological Bulletin, 105, 131–142. Sloman, A. (1997). What sort of architecture is required for a human-like agent? In M. Wooldridge & A. Rao (Eds.), Foundations of rational agency. Kluwer. Smith, E. R., & DeCoster, J. (2000). Dual process models in social and cognitive psychology: Conceptual integration and links to underlying memory systems. Personality and Social Psychology Review, 4, 108–131. Smith, L. B., Thelen, E., Titzer, R., & McLin, D. (1999). Knowing in the context of acting: The task dynamics of the A-not-B error. Psychological Review, 106, 235–260. Snyder, M., & Cantor, N. (1998). Understanding personality and social behavior: A functionalist strategy. In D. T. Gilbert, S. T. Fiske & G. Lindzey (Eds.), Handbook of social psychology (4th ed., Vol. 1, pp. 635–679). Boston, MA: McGraw-Hill. Sorrentino, R. M., & Higgins, E. T. (Eds.) (1986). Handbook of motivation and cognition: Foundations of social behavior. New York: Guilford Press. Stasser, G. (1992). Information salience and the discovery of hidden profiles by decision-making groups: A ‘‘thought experiment.’’ Organizational Behavior and Human Decision Processes, 52, 156–181. Stasser, G. (1999). The uncertain role of unshared information in collective choice. In L. L. Thompson & J. M. Levine (Eds.), Shared cognition in organizations: The management of knowledge (pp. 49–69). Mahwah, NJ: Lawrence Erlbaum. Stasser, G., & Stewart, D. (1992). Discovery of hidden profiles by decision-making groups: Solving a problem versus making a judgment. Journal of Personality and Social Psychology, 63, 426–434. Steels, L., & Brooks, R. (1995). The artificial life route to artificial intelligence: Building embodied, situated agents. Hillsdale, NJ: Lawrence Erlbaum. Stevens, L. E., & Fiske, S. T. (1995). Motivation and cognition in social life: A social survival perspective. Social Cognition, 13, 189–214.

116

¨ N R. SEMIN ELIOT R. SMITH AND GU

Strauss, A. (1959). Mirrors and masks: The search for identity. Glencoe, Il: Free Press of Glencoe. Tajfel, H., Billing, M. G., Bundy, R. P., & Flament, C. (1971). Social categorization and intergroup behavior. European Journal of Social Psychology, 1, 149–178. Thelen, E., & Smith, L. (1994). A dynamic systems approach to the development of cognition and action. Cambridge, MA: MIT Press. Thompson, L., & Fine, G. A. (1999). Socially shared cognition, affect, and behavior: A review and integration. Personality and Social Psychology Review, 3, 278–302. Tice, D. M. (1992). Self-concept change and self-presentation: The looking-glass self is also a magnifying glass. Journal of Personality and Social Psychology, 60, 218–228. Tiedens, L. Z., & Linton, S. (2001). Judgment under emotional certainty and uncertainty: The effects of specific emotions on information processing. Journal of Personality and Social Psychology, 81, 973–988. Tindale, R. S., & Kameda, T. (2000). ‘‘Social sharedness’’ as a unifying theme for information processing in groups Group Processes and Intergroup Relations, 3, 123–140. Tindale, R. S., & Sheffey, S. (2002). Shared information, cognitive load, and group memory. Group Processes and Intergroup Relations, 5, 5–18. Tomasello, M. (2000). Culture and cognitive development. Current Directions in Psychological Science, 9, 37–40. Tomasello, M., & Rakoczy, H. (2003). What makes human cognition unique? From individual to shared to collective intentionality. Human Development, 18, 121–147. Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987). Rediscovering the social group: a self-categorization theory. Oxford: Blackwell. Varela, F., Thompson, E., & Rosch, E. (1991). The embodied mind: Cognitive science and human experience. Cambridge, MA: MIT Press. Vera, A., & Simon, H. A. (1993). Situated action: A symbolic interpretation. Cognitive Science, 17, 7–48. Vygotski, L. S. (1956/1987). The collected works of L. S. Vygotski. Vol. 1 Problems of general psychology. New York & London: Plenum Press. Vygotsky, L. (1962/1986). Language and thought. Cambridge, MA: MIT Press. (Original publication, 1962). Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Cambridge, MA: Harvard University Press. Wegner, D. M. (1995). A computer network model of human transactive memory. Social Cognition, 13, 319–339. Wegner, D. M., & Wheatley, T. (1999). Apparent mental causation: Sources of the experience of will. American Psychologist, 54, 480–492. Wells, G. L., & Petty, R. E. (1980). The effects of overt head movements on persuasion: Compatibility and incompatibility of responses. Basic and Applied Social Psychology, 1, 219–230. Wentura, D., Rothermund, K., & Bak, P. (2000). Automatic vigilance: The attention-grabbing power of approach- and avoidance-related social information. Journal of Personality and Social Psychology, 78, 1024–1037. Wertsch, J. V. (1998). Mind as action. New York & Oxford: Oxford University Press. Wigboldus, D., Semin, G. R., & Spears, R. (2000). How do we communicate stereotypes? Linguistic bases and inferential consequences. Journal of Personality and Social Psychology, 78, 5–18. Wilson, M. (2002). Six views of embodied cognition. Psychonomic Bulletin and Review, 9, 625–636.

SOCIALLY SITUATED COGNITION: COGNITION IN ITS SOCIAL CONTEXT

117

Wood, W., & Eagly, A. H. (2002). A cross-cultural analysis of the behavior of women and men: Implications for the origins of sex differences. Psychological Bulletin, 128, 699–727. Yeh, W., & Barsalou, L. W. (2000). The situated nature of concepts. Unpublished, Department of Psychology, Emory University. Zajonc, R. B. (1960). The process of cognitive tuning and communication. Journal of Abnormal Social Psychology, 61, 159–167. Zajonc, R. B., & Adelmann, P. K. (1987). Cognition and communication: A story of missed opportunities. Social Science Information, 26, 3–30. ¨ ber das leehalten Behalten von erledigten und unerleighten Handlingen. Zeigarnik, B. (1927). U Psycholische Forschung, 9, 1–85.

This Page Intentionally Left Blank

SOCIAL AXIOMS: A MODEL FOR SOCIAL BELIEFS IN MULTICULTURAL PERSPECTIVE

Kwok Leung Michael Harris Bond

A five-factor structure of general beliefs or social axioms was established in five cultures by Leung et al. (2002). This chapter describes a global research program designed to evaluate the universality and meaning of this structure. Student data from 40 cultures and adult data from 13 cultures were collected, with both types of data providing strong support for the generality of this five-factor structure. Specifically, social cynicism indicates the extent to which actors expect positive versus negative outcomes from their engagements with life, especially with more powerful others. Social complexity indicates an actor’s judgments about the variability of individual behavior and the number of influences involved in determining social outcomes, both issues reflecting a complex theory of social causation. Reward for application indicates how strongly a person believes that challenges and diYculties will succumb to persistent inputs, such as relevant knowledge, exertion of eVort, or careful planning. Religiosity indicates an assessment about the positive, personal, and social consequences of religious practice, along with the belief in the existence of a supreme being. Fate control indicates the degree to which important outcomes in life are believed to be fated, but are predictable and alterable. The validity and usefulness of this structure are established by correlating citizen profiles across these dimensions for each cultural group with societal characteristics. Furthermore, within-culture and cross-cultural studies are reviewed to support the meanings of these axioms. Theoretical implications of our findings and directions for future research are discussed.

119 ADVANCES IN EXPERIMENTAL SOCIAL PSYCHOLOGY, VOL. 36

Copyright 2004, Elsevier Inc. All rights reserved. 0065-2601/04 $35.00

120

KWOK LEUNG AND MICHAEL HARRIS BOND

Man, biologically denied the ordering mechanisms with which the other animals are endowed, is compelled to impose his own order upon experience. —Peter L. Berger, The Sacred Canopy: Elements of a Sociological Theory of Religion

Each of us confronts a material, a social, and a spiritual universe that must be structured so that we can negotiate our way through the maze of life. This structuring involves prioritizing our goals for the various types of situation we confront and then using our received and achieved knowledge to realize our goals. Through a lifetime of transactions with our world, we develop a blueprint for action or inaction at this interface. The valuational component of this interface has been well studied by social and personality psychologists (e.g., Rokeach, 1973). This work has been leavened culturally by social scientists who have attempted to incorporate culturally distinctive values into their measurements (e.g., Kluckhohn & Strodtbeck, 1961), establish the metric equivalence of their value constructs for persons in many cultural groups (e.g., Schwartz, 1992), and then administer their measures to equivalent samples from a variety of cultures (e.g., Bond, 1988). The result of these labors has been a mapping of the value universe, enabling comparisons to be made in the value profile of representative persons socialized into diVerent cultural traditions (e.g., Echter et al., 1998). The best-known research program on values around the globe is probably the work of Schwartz (1992), who used theoretical considerations and the measures they suggested to discover a pan-cultural typology of values. Given that the value constructs used were comprehensive and meaningful for all these persons, cross-cultural studies could then be conducted to link these value constructs to social behavior. DiVerences in the behavioral profile of persons from these diVerent cultural groups could then sometimes be unpackaged (Bond, 1998) by using the diVerent strengths of these value constructs across cultural groups to explain diVerences in behavioral intensions and actual behaviors across these cultural groups (e.g., Bond, Leung, & Schwartz, 1992; Sagiv & Schwartz, 1995). Or sometimes not. Ip and Bond (1995) found that diVerences between Hong Kong Chinese and Americans in their use of self-descriptive categories, such as traits or social roles, were unrelated to any of Schwartz’s (1992) domains of value, either within or across cultural groups. Values did not unpackage category use in this case. Whenever we fail to unpackage diVerences in outcomes across cultures by using a psychological construct, a variety of possibilities suggest themselves. For instance, the behavioral outcome in question may be normatively constrained in one or more of

SOCIAL AXIOMS

121

the cultural groups and not be controlled by an internal, psychological mediator, like values (Smith & Bond, 2004). Another possibility, which is more relevant to this chapter, is that other types of psychological construct may be driving the behavioral outcome within the cultural groups, such as expectancies (Leung, 1987; Leung, Bond, & Schwartz, 1995). Only recently have cross-cultural psychologists begun calling for and testing hypotheses about cultural mediation of psychological processes (e.g., Brockner, 2003), and our development of theory and mediating mechanisms is nascent. This chapter describes a research program designed to identify one such psychological construct, general beliefs or social axioms. This program was designed to establish pan-cultural dimensions of what people hold to be true by developing a Social Axioms Survey that integrates the psychological literature on beliefs, which is primarily Euro-American in origin, and inputs from two other cultural traditions, Chinese and Venezuelan. A five-dimensional structure of social axioms was identified in Hong Kong and Venezuela, and subsequently replicated in the United States, Germany, and Japan (Leung et al., 2002). The universality of this structure of social axioms was then evaluated by a round-the-world study, the results of which are presented in this chapter. Two strategies are adopted to evaluate the validity and usefulness of this structure. First, we develop citizen profiles across these dimensions for each national group, and these profiles and dimension scores are then linked to societal characteristics to discover how individual ‘‘consciousness’’ tracks institutional processes and is sustained by these ‘‘plausibility structures’’ (Berger & Luckmann, 1966). In other words, the correlations between the social axiom scores of citizens and country-level indexes constitute a nomological network on which the meaning and validity of these dimensions can be established. Second, whatever these citizen profiles, persons within any given nation will endorse them to varying degrees and use them accordingly to anticipate, guide, and rationalize their behavior. Social axioms function like other individual diVerences constructs, with their own nomological networks linking them to other constructs such as values, and combining with these other psychological constructs to generate behavior. This pattern of linkage among the internal constructs and with behavior may vary from culture to culture. Additionally, external, social factors contribute to improving behavioral predictions, diYcult as these situational forces may be to conceptualize and measure (e.g., Seeman, 1997; Wicker, 1992). Establishing these facts, however, requires within-culture and cross-cultural studies that treat beliefs or other mediators as individual diVerences factors, investigating their impact in concert with other internal, and theoretically relevant, external, social factors. This chapter also presents some emerging data from this perspective on social axioms in the hope of stimulating extensions of this

122

KWOK LEUNG AND MICHAEL HARRIS BOND

work into other behavioral domains. Unlike the first strategy, results from this approach are pitched at the individual level (Leung & Bond, 1989).

I. Beliefs in Social Psychological Research Beliefs are key concepts in major disciplines in social sciences, including psychology, sociology, anthropology, and political science. Bar-Tal (2000, ch. 2) has presented a historical review of the study of shared beliefs in social psychology, tracing its intellectual roots to Wundt and Durkheim. For our purposes, the social psychological studies of beliefs may be roughly grouped into four categories based on their intellectual traditions. A. SHARED BELIEFS This perspective argues that beliefs are necessarily social in nature, and hence are widely shared within a social group, such as a community and a nation. The best-known approach in this tradition is social representations theory initiated by Moscovici some 50 years ago, itself inspired by the work of Durkheim on collective representation. The central idea of this approach is that an understanding of reality is made possible by social representations, which are shared among people in a group. According to Moscovici (2001), a social representation ‘‘chooses and combines our shared concepts, links together accepted assertions, decides which aspects from our categories are examples for classifying people and things’’ ( p. 18). Social representations allow people to understand their social world, enable them to interact and communicate with each other eVectively, and guide their actions. Social representations exist not only in people’s minds but are also represented in products of human endeavors, such as books, architecture, and movies. In Moscovici’s formulation, social representations are dynamic in nature and can change as a function of interpersonal interaction and communication. People develop social representations for a variety of objects, events, situations, and domains, and these representations may vary in complexity and abstraction. A simple example of how social representations aVect our perception is provided by Duveen (2001), concerning the relative location of Prague and Budapest in relation to Vienna. Most people are likely to believe that both Prague and Budapest are to the east of Vienna, because both cities belong to the former Eastern Bloc. In reality, Prague is west of Vienna, and is closer to Western than to Eastern Europe. Wagner and Kronberger (2001) have provided a more complex example in their investigation of the social representations of genetic engineering of foodstuV in Europe. They have

SOCIAL AXIOMS

123

identified some key beliefs in these representations, such as the belief that genes introduced into living beings make them bigger and monstrous. The emergence of these specific beliefs was then traced back to widespread beliefs about how bacteria cause diseases and how misguided science can generate ‘‘Frankensteinian’’ monsters. In short, people’s understanding and perception of unfamiliar objects are often influenced by social representations of objects related to these unfamiliar objects. A major criticism of social representations theory is its vagueness, especially with regard to the construction of its concepts (Jahoda, 1988). Markus and Plaut (2001) noted jokingly that, ‘‘Social representations was the type of theory that if you forgot your lecture notes on the day you intended to cover this material, you were in trouble’’ (p. 183). In a recent book, Bar-Tal (2000) has described a special kind of social representation, labeled as societal beliefs, whose primary function is to provide the basis for a group’s identity. The framework proposed by Bar-Tal is well defined, with clearly specified concepts and hypotheses. Societal beliefs are defined as, ‘‘enduring beliefs shared by society members, with contents that are perceived by society members as characterizing their society’’ (p. 39). Societal beliefs are organized around themes, and Bar-Tal has described several types of societal belief: patriotism, security, siege, and delegitimization. If security is used as an example, beliefs under this theme are concerned with assessments of the extent of perceived insecurity, types of perceived threat, means for attaining security, and the conditions deemed necessary for security. Bar-Tal further proposed the notion of an ethos, which is constituted by a particular configuration of societal beliefs and provides the characterization of a society. For instance, based on the work of McClosky and Zaller (1984), Bar-Tal described the American ethos as dominated by beliefs based on capitalism and democracy. Thus, shared beliefs reflect how people construct their social world to seek meaning and understanding of social realities. This construction is shared, thereby allowing people to interact and communicate with each other, providing them with a common social identity, and guiding them toward a consensual course of action. In this research tradition, social representations are typically studied in a particular context, and the beliefs involved are context specific.

B. BELIEFS AND LAY THEORIES In navigating their everyday social worlds, people need some assumptions of how their social worlds function. These understandings, often expressed as beliefs, are known as implicit or lay theories. The root of this work is often traced to Heider’s (1958) naive analysis of actions, in which people are

124

KWOK LEUNG AND MICHAEL HARRIS BOND

viewed as lay scientists who develop, test, and revise hypotheses in their daily life. Another pioneer in this area is Kelly (1963), who introduced the notion of personal construct, which people use to perceive and interpret events and to chart a course of action. Furnham (1988) has provided an overview of lay theories in various domains, including medicine, psychiatry, psychology, economics, education, statistics, and law. Although researchers have identified structures underlying the lay beliefs in domains that interest them, no attempt has been made to search for and develop a fundamental, contextfree structure of lay beliefs. Subsequent to Furnham’s (1988) review, lay theories have continued to flourish in the social psychological literature. However, current research on lay theories, as in many fields of psychology, has assumed a more cognitive outlook. For example, in the area of group perception the notion of perceived entitativity has become an influential topic and channeled much of the research in this area. The central notion is that people have a tendency to view aggregates of individuals as social entities bonded together in some way. The degree of perceived entitativity of a group of individuals has been shown to exert a significant impact on the perceiver’s judgment and behavior toward these individuals (Brewer & Harasty, 1996; Hamilton & Sherman, 1996). Lickel et al. (2000) have found that perceived entitativity varied across diVerent types of group, with intimacy groups (e.g., friends) showing the highest level of perceived entitativity, followed by task groups (e.g., work teams), social categories (e.g., women), and loose associations (e.g., classical music lovers). Dasgupta, Banaji, and Abelson (1999) found that people ascribed more hostile intentions and traits to groups that were perceived to consist of members with similar physical features. These findings suggest that physical features provide one basis of perceived entitativity. A second example comes from the cross-cultural research on attribution. Morris and colleagues (e.g., Morris, Menon, & Ames, 2001; Morris & Peng, 1994) proposed the notion of implicit theories of agency for understanding social perception in general and cross-cultural diVerences in attribution in particular. Implicit theories of agency refer to ‘‘conceptions of kinds of actors, notions of what kind of entities act intentionally and autonomously’’ (Morris et al., p. 169). This theorizing then is used to account for consistent cultural diVerences in attribution between Chinese and Americans, namely, that Chinese are more likely to attribute an action to social causes; Americans, to dispositional causes. This pattern of cultural diVerences is consistent with the view that Chinese are likely to subscribe to the view of agentic collectives (e.g., families and groups), whereas Americans are likely to subscribe to the view of agentic individuals. For instance, Menon, Morris,

SOCIAL AXIOMS

125

Chiu, and Hong (1999) found that Americans were more likely to endorse statements such as ‘‘individuals possess free will’’ and ‘‘follow their internal direction,’’ where Singaporeans were more likely to endorse similar statements about organizations. In summary, lay beliefs have been studied in many domains across the past three decades, and many exciting findings have been documented. However, no attempt has been made to search for a basic structure of lay beliefs that is domain general.

C. PROCESS MODELS OF BELIEFS An independent but related development that parallels the study of lay theories is concerned with the processes underlying the formation and change of beliefs. Research on process models of beliefs in social psychology is often associated with the long tradition of research on attitudes, which are viewed as consisting of cognitive, aVective, and behavioral components. Given that beliefs are the key components of attitudes, process models of beliefs have emerged alongside process models of attitudes. Work by Festinger (1957) on cognitive dissonance pioneered the research on process models of beliefs. Cognitive dissonance is generated when people are confronted with two contradictory cognitions, resulting in cognitive, aVective, and behavioral changes. Subsequently, perhaps the most widely known research program on social psychological processes underlying beliefs is that exploring the self-fulfilling prophecy, which provides compelling evidence for the capability of beliefs to create social reality (Snyder, 1984). A more recent process approach to beliefs is the theory of lay epistemics proposed by Kruglanski (1989), which describes and explains the formation and modification of human knowledge through hypothesis generation and validation. This model, like other process models, focuses on the dynamics involved in perceptual and cognitive processes, so that the content of beliefs and cognitions is accorded only peripheral attention. Another well-known category of process models of beliefs is concerned with the organization of knowledge structures, such as schemas and scripts (e.g., Abelson, 1981; Rumelhart, 1984). Abelson (1988) added to this line of work by highlighting the importance of strongly held beliefs. One example of such beliefs is provided by Roseman (1994), who examined beliefs about disarmament. In general, the beliefs studied in this tradition tend to be textured and domain specific, and the discovery of a general, context-free structure to beliefs has not been a goal of this line of work.

126

KWOK LEUNG AND MICHAEL HARRIS BOND

D. BELIEFS AS INDIVIDUAL DIFFERENCES Beliefs have been widely used as individual diVerences variables to explain and predict social behavior. In this tradition, belief scales are developed and their validity and usefulness is demonstrated by significant relationships with a variety of variables in an expected fashion. The concept of locus of control, which refers to a person’s belief about whether events that happen to oneself are within or outside of one’s control (Rotter, 1966), may be the best-known example of this line of enquiry. Locus of control has been shown to relate to a wide spectrum of behaviors (e.g., Lau & Leung, 1992; Spector, 1982). The work of Wrightsman (1992) on beliefs about human nature provides another example of this approach. He identified six factors of such beliefs: trustworthiness, altruism, strength of will and rationality, independence, complexity, and variability. Individual beliefs about human nature were shown to relate to many interpersonal behaviors. For instance, counselors who believed that people possess strength of will and rationality were preferred as counselors and better liked by their peers (Wrightsman, Richard, & Noble, 1966). One point to note is that while belief items are found in many scales in the individual diVerences literature, they are often mixed together with items that tap values or behaviors. This conflation creates theoretical ambiguity and imprecision in model development. Scales that are based entirely on beliefs are rare, and locus of control and beliefs about human nature are among the very few scales that are primarily based on beliefs.

E. SECTION SUMMARY The previous review of the major research on beliefs makes it clear that belief is a key construct in social psychology and that significant discoveries have been made about the content of beliefs in diVerent domains and their underlying psychological processes. Although the research on belief is vibrant and ubiquitous, one obvious observation is that findings in diverse areas are typically tied to a particular context and are extremely rich in detail. Confronting the work about beliefs in social representations, lay theories, and belief scales can create confusion because of its diversity and range. In biology, the range of flora and fauna is perhaps even more diverse, but a taxonomy primarily guided by evolutionary theory has organized the vast multitude of living organisms into coherent and ordered groups. Such a taxonomy is lacking for beliefs, and the field has not developed a theoretical scheme for organizing beliefs into a coherent structure. The reason for the lack of interest in developing such a scheme, in our opinion, is twofold. First,

SOCIAL AXIOMS

127

the level of abstraction of a great deal of belief research is pitched at the concrete end of the continuum. For instance, most research in shared belief is tied to a particular context, and the beliefs identified are typically concrete and specific to that context. Second, when researchers work on more abstract, or context-free beliefs, as is common with lay theories and belief scales, they often focus on a specific belief, such as locus of control or group entitativity. The simultaneous examination of a wide range of general beliefs has not yet been a research agenda in this literature. To address this omission, a major goal of our research program on social axioms is to identify a pan-cultural structure of broad, context-free beliefs and to examine how this structure is related to a wide range of social behavior. II. Social Axioms as a Basic Psychological Construct ‘‘There is a mask of theory over the whole face of nature.’’ —William Whewell, The Philosophy of the Inductive Sciences

A. THE IMPORTANCE OF GENERAL BELIEFS Humans confront James’ ‘‘booming, buzzing confusion’’ with a naive scientific persistence in extracting order from chaos. Our earliest experience with this process is the discovery of the sequences that inform pleasurable and painful experiences—appetitive and aversive conditioning. Subsequent developmental progression involves higher-order conditioning with physical and symbolic stimuli and stimulus complexes. Then we begin to extract domain-specific knowledge about the relations among the elements and events made salient through conditioning, so that we may deal more eVectively with the micro-worlds that confront us in the developmental niches provided by our culture (Super & Harkness, 1986). Useful constructs are deployed, assessed, and elaborated. Armed with this knowledge about how our world operates, we become better able to manage our lives more eVectively, extracting the rewards we have come to desire and avoiding the punishments we have learned to fear. This pervasive ordering process may be an innate, irresistible disposition, part of our human endowment, as Kant maintained. Voegelin (1956) believes that our search for order is driven by the basic terror of chaos: Any historical society is an order, a protective structure of meaning, erected in the face of chaos. Within this order the life of the group as well as the life of the individual makes sense. Deprived of such order, both group and individual are threatened with the most fundamental terror, the terror of chaos, that Emile Durkheim called anomie, literally a state of being ‘‘order-less.’’ (p. 9)

128

KWOK LEUNG AND MICHAEL HARRIS BOND

Within any given society, individuals will engage in this ordering process to varying degrees, depending in part on their need for cognition (Cacioppo, Petty, Feinstein, & Jarvis, 1996). In addition, there will be variations within a given society in the resulting constructions of reality maintained by its citizens. This variation in the content of the cognitive constructions of our worlds arises from the socialization influences operative in a citizen’s proximal social environments across the developmental sequence. We expect that the informational and attitudinal inputs and reinforcement contingencies provided by one’s parents will be especially important (Harkness & Super, 1996), along with genetic impairments that result in certain individuals’ belief systems being labeled as ‘‘delusional’’ (Oltmanns & Maher, 1988), and the particular events of an individual’s life course (Bandura, 1982). This accumulated knowledge about the world is most obviously derived from the explicit acquisition of information and training about the material world transmitted through schooling. Simultaneously and implicitly, however, we human beings are learning about how the social world functions, our position within that interpersonal and institutional maze, our capacity to shape outcomes for ourselves and for others, and the role of other social and spiritual forces external to ourselves in directing the course of personal, interpersonal, social, economic, and political events. Many of these beliefs are generalizations, extracted from direct experience across a variety of contexts or mediated by mentors, and deduced from indirect experience provided through the media, literature, and cultural myths and sustained by social discourse. These beliefs are likely to show cultural variations as diVerent cultures are associated with diVerent physical and social environments (Triandis, 1964). As previously reviewed, psychologists have studied these understandings using diVerent rubrics, such as lay theories (Furnham, 1988) and shared beliefs (Bar-Tal, 2000; Garfield, 1988).

B. SOCIAL AXIOMS AS CONTEXT-FREE GENERAL BELIEFS These synthesized understandings of how the world works are termed social axioms. Beliefs vary in specificity. Some beliefs are anchored in a context, defined by the actors involved and tied to a particular setting in a given time period. For instance, we may develop specific beliefs about a city we have visited (e.g., visiting Kyoto in April when the cherry trees blossom is relaxing and uplifting for the Japanese) or about the people we know (e.g., John is often forgetful when under a heavy workload). Specific beliefs are numerous, and each is only applicable to a narrow range of situations and actors. However, the specificity of these beliefs is often instrumental to the understanding of why a particular behavior occurs, as in the case of some

SOCIAL AXIOMS

129

attitudinal models, which rely on salient beliefs in a given context to predict specific behaviors (Fishbein & Ajzen, 1975). In contrast, some beliefs are general and may be viewed as generalized expectancies, a concept first proposed by Rotter (1966) to characterize the construct of locus of control. Because these general beliefs are pitched at a high level of abstraction, they are context-free and are related to a wide spectrum of social behaviors across diverse contexts, actors, targets, and time periods, as in the case of locus of control. We choose the label social axioms for these general beliefs in the sense that, like axioms in mathematics, these beliefs are basic premises that people endorse and rely onto guide their actions. These beliefs are axiomatic because they are often assumed to be true as a result of personal experiences and socialization. Beliefs have been defined in many diVerent ways, and the following two definitions are representative: ‘‘If a man perceives some relationship between two things or between something and a characteristic of it, he is said to hold a belief ’’ (Bem, 1970, p. 4). ‘‘A proposition to which a person attributes at least a minimal degree of confidence. A proposition, as a statement about an object(s) or relations between objects and/or attributes, can be of any content’’ (Bar-Tal, 1990, p. 14).

Based on these definitions, social axioms are formally defined as: ‘‘generalized beliefs about oneself, the social and physical environment, or the spiritual world, and are in the form of an assertion about the relationship between two entities or concepts’’ (Leung et al., 2002, p. 289).

This definition implies that a social axiom has the structure—A is related to B. A and B can be any entities, and the relationship may be causal or correlational. So, for example, the belief statement, ‘‘Hard work leads to reward,’’ asserts that a causal connection exists between labor and positive outcomes for the laborer. It is a general statement, because there are many forms of ‘‘hard work’’ just as there are many types of ‘‘reward.’’ It is not an attitude or a value, since the respondent is not assessing the desirability either of ‘‘hard work’’ or ‘‘reward.’’ Axioms claim truth-for-the-actor; they do not assess desired goals. Specifically, a value assumes the form: A is good/desirable/important. A is a value, and its importance is determined by the importance or desirability that people attach to it. Because the structure of a value is quite similar to the structure of a belief, some researchers actually regard a value as an evaluative belief. We do not deny that an evaluative dimension underlies most, if

130

KWOK LEUNG AND MICHAEL HARRIS BOND

not all beliefs, because an evaluative dimension underlies all semantic objects (Osgood, Suci, & Tannenbaum, 1957). However, we argue that beliefs are diVerent from values because, whereas the evaluative component of a value is general, it is specific for a belief. In other words, if the desirability pole of an evaluative belief becomes specific, it turns into a social axiom. For instance, ‘‘Wars are bad’’ and ‘‘power is desirable’’ are evaluative beliefs and would be classified as values rather than social axioms. In contrast, ‘‘Wars will lead to the destruction of civilization,’’ and ‘‘absolute power corrupts absolutely’’ are regarded as social axioms because these statements define a specific relationship between two entities. Social axioms also diVer from norms, often expressed as normative beliefs, which are prescriptive in nature, and take the form—A should do X, where A is a person and X is an act. ‘‘We should protect our environment’’ is a normative belief, but not a social axiom, because the statement prescribes a proper course of action rather than claim in a relationship between two entities. As noted, previous research on beliefs often involves a congeries of social axioms, values, and normative beliefs, and these diVerent constructs are not carefully distinguished (e.g., see the belief scales included in Robinson, Shaver, & Wrightsman, 1991). The locus of control scale (Rotter, 1966), perhaps the most famous belief scale, actually contains some normative statements. For example, the item ‘‘one should always be willing to admit mistakes’’ is a normative statement rather than a general belief. Normative statements, such as ‘‘Do you believe that parents should allow children to make most of their own decisions?’’ can also be found in the Nowicki-Strickland Internal-External local of control scale (Nowicki & Duke, 1983). As noted, belief scales composed of pure beliefs are rare. More examples of social axioms are listed in the description of our research program.

C. A FUNCTIONALIST VIEW OF SOCIAL AXIOMS One commonality that spans across the work on social representations, lay theories, and attitudes is a functionalist approach to beliefs, described and endorsed at the beginning of this section. This view proposes that beliefs and other attitudinal constructs serve at least four functions for human survival and functioning (Katz, 1960; Kruglanski, 1989). Following this argument, we propose that these axioms ‘‘facilitate the attainment of important goals (instrumental ), help people protect their self-worth (ego-defensive), serve as a manifestation of people’s values (value-expressive), and help people understand the world (knowledge)’’ (Leung et al., 2002, p. 288). Given this extensive range of functions, social axioms qualify to be considered as a

SOCIAL AXIOMS

131

fundamental psychological constructs. We expect that they may be linked to other broad psychological constructs like values, and predict more specific psychological constructs such as domain-specific eYcacies (Bandura, 2002) or beliefs about the causes and cures of psychological problems (Luk & Bond, 1992). Axioms help channel one’s behavior, as in expectancy-value theories (Feather, 1982), and provide mechanisms for explaining personal outcomes, interpersonal exchanges and environmental events, both human and physical. Such a functionalist argument also provides the conceptual basis of the cross-cultural project on values by Schwartz (1992). In essence, Schwartz argued that human beings are confronted with a set of universal problems for survival and eVective functioning. In his words, ‘‘Values represent, in the form of conscious goals, three universal requirements of human existence to which all individuals and societies must be responsive: needs of individuals as biological organisms, requisites of coordinated social interaction, and survival and welfare needs of groups’’ (p. 4). Thus, people from diverse cultural and ecological backgrounds should develop a common scheme of value types to guide their actions and choices in addressing these three pancultural needs. Consistent with this argument, his multinational study demonstrated the universality of 10 value types across highly diverse cultural groups, with diVerent socioeconomic-political backgrounds. However, specific cultural contexts shape the relative strength of each cultural group’s endorsement of these universal value types. Note that Schwartz’s (1992) work, like our work on social axioms presented here, but unlike Hofstede’s (1980) famous cultural dimensions, is also at the individual level because the individual is the unit of analysis and the theorizing is psychological rather than societal. Following the functionalist tradition in attitude research and Schwartz’s (1992) logic for a universal structure of values, we propose that social axioms, like values, are instrumental for individuals in coping with a set of universal problems of survival and eVective functioning, and hence the structure underlying these axioms should be identifiable in diVerent cultural groups with diverse backgrounds. In other words, the commonality of the basic problems that all human beings face should lead to the emergence of a pan-cultural structure of social axioms. Obviously, the fact that these social axioms are recognizable by people of diverse cultural origins and structured in a similar fashion across diVerent cultural groups does not mean that cultural contexts exercise no influence on their level of endorsement. On the contrary, we expect that the axioms themselves will vary in their endorsement levels as a function of the socialization experiences crystallizing around culture, gender, education level, religion, age, and social class.

132

KWOK LEUNG AND MICHAEL HARRIS BOND

III. Dimensions of Social Axioms . . . we hold also that some constructions serve us better than others in our eVorts to anticipate comprehensively what is actually going on. The big question is, of course, which ones and how do we know. —George A. Kelly, The Psychology of the Unknown. In D. Bannister (Ed.), New Perspectives in Personal Construct Theory.

A. THE INITIAL FIVE-CULTURE STUDY OF SOCIAL AXIOMS Prescriptive knowledge and operating beliefs are legion within a given cultural tradition, let alone across the world’s major cultural systems. However, many are merely alternative ways of expressing the same observation about how the world works. Simplifying these various axioms into manageable groupings first requires due diligence in culling the full range of general beliefs, so that any resulting typology may claim comprehensiveness. Leung et al. (2002) have provided the initial attempt at identifying a pan-cultural structure of social axioms across five cultural groups. Their starting point is the psychological literature on beliefs, which is primarily Euro-American in origin. The psychological literature on beliefs was extracted from three volumes of survey instruments, containing more than 300 scales (Miller, 1991; Robinson et al., 1991; Stewart, Hetherington, & Smith, 1984). Items were included if they were consistent with our definition of social axioms. Input from other cultural sources that are distinct from the Euro-American traditions is also needed to broaden the cultural coverage of the axioms, Leung et al. (2002, Study 1) gathered beliefs from a variety of sources in Hong Kong Chinese culture—proverbs, newspaper reports, cultural stories, and structured interviews with a broad range of citizens. An additional guarantor of comprehensiveness involved undertaking the same process in an ostensibly diVerent cultural tradition, so that additional and possibly distinctive axioms relevant and salient to culture carriers of that tradition could be added to those from our Cantonese cohort. To that end, a similar process of belief identification was undertaken by Sharon Reimel de Carrasquel and her colleagues in Venezuela, a primarily Catholic country in South America. Such open-ended, multichannel searching led to the identification of more than 2000 items in Hong Kong and 1000 items in Venezuela. The resulting beliefs were assembled and obvious repetitions were deleted from the group. Domain-specific beliefs were likewise dropped or rewritten to render them context-free or less idiomatic or metaphorical. To check the comprehensiveness of the beliefs collected, statements about similar topics and issues were grouped into four broad categories:

SOCIAL AXIOMS

133

Psychological attributes—axioms concerning characteristics or tendencies of persons Orientation toward the social world—axioms about the social characteristics of groups, organizations, and societies Social interaction—axioms about how people interact with each other Environment—axioms about aspects of the environment that have implications for social behavior

Furthermore, based on our subjective judgments, the items were grouped into 33 subcategories across these four broad areas. This classification was not meant to be definitive but as a simple heuristic to help evaluate the comprehensiveness of the items. For several subcategories with insuYcient representation, additional items were generated so that these subcategories would have a chance to surface as a coherent group in subsequent analysis. A total of 182 axioms were finally identified after this exhaustive process, with each item phrased in simple language. A 5-point scale was used for the items, with labels ranging from ‘‘strongly believe,’’ ‘‘believe,’’ ‘‘no opinion,’’ to ‘‘disbelieve,’’ and ‘‘strongly disbelieve.’’ Three language versions were developed, in Chinese, English, and Spanish, with the English version as the standard. The survey was then administered to citizens and university students of both Hong Kong and Venezuela. Cluster analysis was first conducted in each of the two cultures to identify major groupings among the items and to facilitate interpretation of factors arising from the subsequent factor analysis (Gorsuch, 1983, p. 211). Three to nine factor solutions were examined in each culture, using both varimax and oblique rotations. The scree plots and the interpretability of factors were used as criteria to determine the optimal number of factors, determined to be five for both cultural groups. The orthogonal solution was adopted because it was closely similar to the oblique solution and easier to interpret. A factor analysis based on a combined sample of the two groups was conducted to arrive at an optimal solution for both cultural groups. To avoid the eVects of diVerences in the item means across the two cultural groups, which would aVect the factor analytic outcomes, we followed the procedure recommended by Becker (1996) for the meta-analysis of factor structures. The correlation matrix of each cultural group was transformed by the Fisher transformation and averaged to generate a combined matrix, which was then transformed back to a correlation matrix for factor analysis. This procedure weights each culture’s correlation matrix equally and assumes that each cultural group’s sample provides the best approximation of correlations in its population. As expected, a five-factor solution was

134

KWOK LEUNG AND MICHAEL HARRIS BOND

optimal, which is defined by 60 items. To check how closely the factor structure of each cultural group resembles the common structure, a Procrustes rotation was performed on these 60 items by rotating the factor structure of each culture toward the common factor structure (see Van de Vijver & Leung, 1997). Congruence coeYcients were calculated to evaluate the fit between the resultant factor structure and the common structure for each culture. For Hong Kong, the coeYcients ranged from .88 to .98, and for Venezuela, the range was from .90 to .97. These numbers do not suggest a perfect fit for some of the factors, but Leung et al. (2002) concluded that the common factor structure was reasonably good for the two cultural groups. Leung et al. labeled and described the five dimensions in the following terms. 1. Factor one is labeled social cynicism, because the items represent a negative view of human nature, especially as it is easily corrupted by power, a biased view against some groups of people, a mistrust of social institutions, and a disregard of ethical means for achieving an end. 2. The second factor is labeled social complexity because the items in this factor suggest that there are no rigid rules, but rather multiple ways of achieving a given outcome, and that apparent inconsistency in human behavior is common. 3. The third factor is labeled reward for application because the items represent a general belief that eVort, knowledge, careful planning, and the investment of other resources (Foa, 1971) will lead to positive results and help avoid negative outcomes. 4. The fourth factor is labeled spirituality (to be renamed religiosity below), as the items assert the existence of supernatural forces and the beneficial functions of religious belief. 5. The fifth factor is labeled fate control, as the items represent a belief that life events are predetermined and that there are some ways for people to influence these outcomes. Interestingly, lay people accept the logical contradiction between predetermination and their ability to alter predetermined events. In fact, practices for avoiding bad luck are commonplace in many cultures, and the contradiction involved in the simultaneous belief in predetermination and possibilities for altering one’s fate may be widespread in everyday life. Several items with relatively lower factor loadings are included in the final solution, because in the initial stage of the research program, it is better to err on the side of overinclusiveness. See Table I for the items defining these five dimensions. One step toward establishing these five dimensions of social axioms as universal is to replicate them in other cultural contexts. Leung et al. (2002, Study 2) evaluated the replicability of these five axiom dimensions in three more cultural groups: the United States, Japan, and Germany. The

TABLE I Five-Factor Solution Based on Leung et al. (2002)

Item

1 Social cynicism

Powerful people tend to exploit others

.59

Power and status make people arrogant

.54

Kind-hearted people are easily bullied

.50

Significant achievement requires one to show no concern for the means needed for that achievement

.48

Kind-hearted people usually suVer losses

.48

Old people are usually stubborn and biased

.48

Young people are impulsive and unreliable

.47

It is easier to succeed if one knows how to take shortcuts

.44

Females need a better appearance than males

.44

It is rare to see a happy ending in real life

.43

People will stop working hard after they secure a comfortable life

.42

People deeply in love are usually blind

.41

To care about societal aVairs only brings trouble for yourself

.40

Most people hope to be repaid after they help others

.37

2 Social complexity

3 Reward for application

4 Spirituality

5 Fate control

.31 .28

.30

(continues)

TABLE I (continued )

Item

1 Social cynicism

Harsh laws can make people obey

.36

Old people are a heavy burden on society

.35

The various social institutions in society are biased toward the rich

.34

Humility is dishonesty

.25

2 Social complexity

3 Reward for application

4 Spirituality

5 Fate control

.32

.31

One’s behaviors may be contrary to his or her true feelings

.58

People may have opposite behaviors on diVerent occasions

.57

One has to deal with matters according to the specific circumstances

.54

There is usually only one way to solve a problem

.49

Human behavior changes with the social context

.45

There are phenomena in the world that cannot be explained by science

.45

Current losses are not necessarily bad for one’s long-term future

.42

To deal with things in a flexible way leads to success

.41

To plan for possible mistakes will result in fewer obstacles

.39

.26 .35

To experience various lifestyles is a way to enjoy life

.35

Individual eVort makes little diVerence in the outcome

.26

.32

One’s appearance does not reflect one’s character

.14

.25

.60

One will succeed if he or she really tries Adversity can be overcome by eVort

.29

.60

Every problem has a solution

.51

Good deeds will be rewarded, and bad deeds will be punished

.51

Hardworking people will achieve more in the end

.51

One who does not know how to plan his or her future will eventually fail

.50

Knowledge is necessary for success

.50

The just will eventually defeat the wicked

.48

Competition brings about progress

.31

.28

.34 .32

A modest person can make a good impression on people

.30

Caution helps avoid mistakes Mutual tolerance can lead to satisfactory human relationships

.26

.45 .37

Social justice can be maintained if everyone cares about politics Failure is the beginning of success

.32

.46

.29

Belief in a religion helps one understand the meaning of life

.69

Belief in a religion makes people good citizens

.65

Religious faith contributes to good mental health

.63

There is a supreme being controlling the universe

.26

.63

(continues)

TABLE I (continued )

Item

1 Social cynicism

2 Social complexity

3 Reward for application

.33

.53 .25

Ghosts or spirits are people’s fantasy Religious beliefs lead to unscientific thinking

5 Fate control

.53

Religious people are more likely to maintain moral standards Religion makes people escape from reality

4 Spirituality

.30

.26

.47 .44

Individual characteristics, such as appearance and birthday, aVect one’s fate

.66

Good luck follows if one survives a disaster

.55

Fate determines one’s successes and failures

.55

There are certain ways to help us improve our luck and avoid unlucky things

.54

There are many ways for people to predict what will happen in the future

.43

All things in the universe have been determined

.42

A person’s talents are inborn

.42

Most disasters can be predicted

.24

Note. Some items are recoded, so that all primary loadings are positive regardless of the actual wording. The variances accounted for by these five factors are 8.89% (social cynicism), 7.94% (social complexity), 5.22% (reward for application), 4.09% (spirituality), and 3.28% (fate control).

SOCIAL AXIOMS

139

United States and Germany, as generally individualistic cultures (Hofstede, 1980), provide a good contrast to the two generally collectivistic groups from which the social axioms were derived. These two countries, being primarily Protestant, also extend the coverage of religions, because Venezuela is primarily Catholic, and Hong Kong has a Buddhist tradition. Finally, like Hong Kong, Japan is generally collectivistic and Buddhist, but Japan is much higher in uncertainty avoidance and masculinity than Hong Kong (Hofstede, 1980). If the same five dimensions of social axioms also emerge in these three cultural groups, any claim of their universality is strengthened. Based on the first study, a short form of the Social Axioms Survey containing the 60 items presented in Table I was created and administered to university students in these three cultural groups. Confirmatory factor analysis was conducted for each culture to determine whether the prior structure provided a good fit to the data. For the U.S. sample, the results pointed to an adequate fit, with a goodness of fit index of .93. For the Japanese sample, the fit was marginal, with a goodness of fit index of .88. For the German data, the fit was very good, with a goodness of fit index of .95. Procrustes rotation also was used to check the fit of the factor structure of each culture to the common structure obtained before. Because of missing values in the German data, only Japanese and the U.S. data were included. The five-factor model was obtained in each culture by rotating the factor structure toward the common structure identified before. Congruence coeYcients were then calculated between the corresponding factors to evaluate their similarity, and most results pointed to an acceptable level of congruence. However, for Japan, spirituality (religiosity) and fate control showed low congruence (both at .68), whereas for the United States, social complexity (.70) and fate control (.52) showed low congruence. The dimension of fate control seemed problematic for both groups, whereas other dimensions were reasonably congruent. Leung et al. (2002) concluded that the five factors are distinct and identifiable in these two new cultural groups, but the demand for factor congruence should be relaxed to the demand for factor similarity. For practical purposes, as long as factors show a good level of similarity across cultures, they should be viewed as universal. This controversial point is discussed in a subsequent section.

IV. A Global Study of Social Axioms This wide and universal theatre Presents more woeful pageants than the scene Wherein we play in. —Shakespeare, As You Like It

140

KWOK LEUNG AND MICHAEL HARRIS BOND

The structure of social axioms identified by Leung et al. (2002) was originally derived from two cultures and subsequently replicated in three more cultures. A fundamental question remains unanswered: Does the origin of this structure from Hong Kong and Venezuela result in a cultural bias and limit its generality? In other words, can these five, tentatively designated dimensions of individual belief be identified in a broader sample of our world’s cultural groups, or are there fewer or more dimensions common to our patterns of social cognition? Regardless of the number of belief dimensions, what is their composition, and can each of these five dimensions be defined by the same items? Such structural equivalence (Van de Vijver & Leung, 1997) is a precondition for eVecting cross-cultural psychological comparisons, but it is an elusive goal when individuals from a large number of diVerent cultural groups are being considered. Bosland (1985), for example, was unable to extract psychologically equivalent measures across cultures from the Hofstede’s (1980) value survey, so that comparisons could not be made across persons, genders, or nations in the individual-level endorsement of psychological values. To answer this question of universality, we have orchestrated a global project on social axioms, involving the collaboration of more than 50 colleagues from around the world and data collection from 40 national/cultural groups. The goal was to identify an optimal structure of axioms in a culturally balanced fashion across a diverse range of cultures and determine whether this structure resembles the structure originally identified in Hong Kong and Venezuela.

A. METHODS AND PARTICIPANTS The 60-item Social Axioms Survey developed by Leung et al. (2002) was used in the round-the-world survey. We used the short form because the other items in the original item pool did not generate additional factors in the previous study, and it is unlikely that they would lead to new factors in the global study. However, for some cultural groups, some items identified in Germany in the previous study were added, but subsequent analysis showed that these new items added nothing to the results, and they have been excluded from the analyses reported herein. Data from university students were collected in 40 national/cultural groups, and adult data were collected in 13 national/cultural groups. The adult data were based on convenient samples, and no claims are made about their representativeness. The details of the participating cultural groups are given in Table II.

141

SOCIAL AXIOMS TABLE II Sample Information

Country

Sample size (student)

Sample size (adult) 152

Spanish

284

106

Dutch, French

Argentina Belgium

Questionnaire language

Brazil

200

Portuguese

Canada

146

English

China

160

Chinese

Czech Republic

100

Czech

Estonia

124

Estonian

Finland

100

Finnish, Swedish

France

120

French

Georgia

118

Georgian

Germany

272

86

Greece

136

680

Greek

Hong Kong

162

162

Chinese

Hungary

258

68

India

710

110

Indonesia

178

German

Hungarian English, Bengali Bahasa Indonesia

Iran

84

Israel

150

Italy

138

Italian

Japan

180

Japanese

Korea

222

Korean

Latvia

142

Latvian

Lebanon

110

English

Malaysia

324

Bahasa Malaysia, English

Netherlands

252

Dutch

New Zealand

200

Nigeria (Yoruba)

94

Persian 96

Hebrew

English 62

English

Norway

104

Norwegian

Pakistan

142

Urdu

Peru

122

Spanish

Philippines

172

English

Portugal

304

Portuguese

Romania

128

Romanian (continues)

142

KWOK LEUNG AND MICHAEL HARRIS BOND TABLE II (continued ) Country

Sample size (student)

Russia

116

Singapore

138

Spain

104

Taiwan

246

Thailand Turkey United Kingdom United States (Caucasian)

Sample size (adult)

Russian English

62

Spanish Chinese

90

English

198

Turkish

80 682

Venezuela Total

76

Questionnaire language

7590

English 616

English

62

Spanish

2338

Note. Both student and adult samples are gender-balanced.

B. DATA ANALYSIS AND RESULTS The first step in data analysis was to discard cases with considerable missing data. A total of 18 respondents from the student data set were dropped because they had more than 40% of missing items, but no cases were dropped from the adult data set. Gender-balanced samples were created so that the results would not be biased by any gender diVerences in the endorsement of the items. Specifically, cases from the gender with more respondents were randomly discarded for each cultural group to maintain a 50% –50% gender ratio. The final sample consisted of 7590 students from 40 cultural groups and 2338 adults from 13 cultural groups. The sample size varied across diVerent cultural groups, and a minimum was set at 80 male and female respondents for the student data, and national/cultural groups that did not meet this minimum were discarded. For the adult data, because the number of cultural groups was smaller, a lower minimum was set at 62 respondents so as to include more groups. In the final data sets, the sample size varied above these minimums to some extent across cultural groups. The best procedure to identify a factor structure of social axioms in a culturally balanced fashion is to follow the procedure for meta-analysis of factor analysis used by Leung et al. (2002) as described previously. This procedure yields a structure that provides an optimal fit across all the cultural groups included. Specifically, two sets of factor analyses were conducted: one on the student data and the other one on the adult data. For each set, a 60  60 correlation matrix was calculated for each national/

SOCIAL AXIOMS

143

cultural group, and the resulting correlations were transformed into a matrix of Fisher z scores. Missing values in the data for each group were handled by the use of pairwise correlations. These many matrices were then averaged across the cultural groups involved, and the resulting Fisher z scores of the pooled matrix transformed back into correlations. The final matrix was then subjected to a factor analysis. This procedure treats each of the cultural groups equally, weighting less populous nations the same as the more populous, thereby equating their cultural voice in the resulting pan-cultural factor analysis. No nation’s structure is privileged by becoming the standard template against which other groups are compared, as is the case in many cross-cultural studies of personality (see e.g., McCrae, 2002). By virtue of this equality solution, however, emic (culture-specific) relationships are ‘‘averaged out’’ and the resulting matrix shows strong correlations for pairs of variables likely to be associated across most, if not all, of the constituent cultures (see Bond, 1988, for a variant of this approach). Note that this analysis is conducted at the individual level, and that the results of a Hofstede-type culture-level analysis are reported elsewhere (Bond, Leung, Au, Tong, Reimel de Carrasquel et al., in press). A crucial decision in exploratory factor analysis is the number of factors to extract. Although our previous work with the five foundation cultural groups suggested five factors, we explored by extracting and rotating to orthogonal structure a number of possibilities (3 to 9 factor solutions). Our concerns were to identify a solution that would provide packages of conceptually understandable items that were loaded on factors strongly and discretely. Such solutions promise greater likelihood that the resulting factor groupings will be applicable and meaningful in all the cultural/national groups. A five-factor solution proved best for both student and adult data by these considerations, as well as by the scree plots (Fig. 1). For identifying the salient items in a factor, we decided to apply stringent criteria to increase the probability that the items constituting a factor would show adequate generality across the 40 national/cultural groups contributing to the pan-cultural solution. We used a somewhat high criterion of .35 for a minimum loading of items to their factors and the absence of sizable secondary loadings (>.30). For the student data, the criteria based on the size of primary and secondary loadings resulted in 39 items. The factor loadings based on a factor analysis of these 39 items are given in Table III. For the adult data, the factor loadings derived from a factor analysis of these 39 items are given in Table IV. Oblique rotations were also tried, and the results are similar to those of orthogonal rotations. For the sake of easy interpretation, the results presented here are based on orthogonal rotation. Two observations are obvious from the results presented in Tables III and IV. The factor structure is remarkably similar across students and adults, and

144

KWOK LEUNG AND MICHAEL HARRIS BOND

Fig. 1.

Scree plots for student and adult samples.

these two structures are highly similar to the structure from the earlier, fivenation study. To confirm these impressions, a Procrustes rotation was conducted to evaluate the factor similarity between the student and the adult structures, and between each of these two structures and the structure of Leung et al. (2002) reported before. All 60 items were included so that the present structures can be compared with the structure of Leung et al. (2002).

SOCIAL AXIOMS

145

For the student and adult structures, the congruence coeYcients across these 60 items ranged from .92 to .98, with a mean of .96, suggesting a very high level of congruence (Chan, Ho, Leung, Chan, & Yung, 1999). Equally impressive were the congruence coeYcients between the student structure and the Leung et al. (2002) structure (.86 to .97, with a mean of .93), and between the adult structure and the Leung et al. (2002) structure (.79 to .96, with a mean of .90). In other words, the current structures derived in a culturally balanced fashion across a very diverse set of cultures are highly similar to the structure based on data from Hong Kong and Venezuela. Fate control showed the lowest congruence for both groups, which confirmed its lower cross-cultural congruence reported in the earlier study (Leung et al., 2002). Nevertheless, given such striking convergence, the same labels will be used to characterize the factors, except for our change to religiosity, because most items in this factor concern the social functions of religion and the consequences of religious adherence. Because the structure of the student data was derived from a large sample size, it is estimated reliably and is adopted as the standard structure in the rest of the discussion. Eleven items define social cynicism. The average itemtotal correlations across these 11 items for the 40 cultural/national groups was .33, with a high of .46 (Iranians) and a low of .15 (Russians). Note that in calculating an item-total correlation, the item involved is excluded in computing the total score. Only one item yielded a negative item-total correlation of .02 in one of the national/cultural groups. With regard to the average item-total correlations for these 11 items, the high was .42 (‘‘Kind-hearted people usually suVer losses’’) and the low was .26 (‘‘Young people are impulsive and unreliable’’). Seven items define religiosity. The average item-total correlation across these seven items for the 40 cultural/national groups was .44, with a high of .68 (Turks) and a low of .13 (Thais). Only two items yielded a single, negative item-total correlation each of .003 and .09. With regard to the average item-total correlations for these seven items, the high was .57 (‘‘Belief in a religion helps one understand the meaning of life’’) and the low was .35 (‘‘Religious beliefs lead to unscientific thinking’’). Nine items define reward for application. The average item-total correlation across these nine items for the 40 cultural/national groups was .31, with a high of .49 (Iranians) and a low of .18 (Georgians). No item showed a negative item-total correlation. With regard to the average item-total correlations for these nine items, the high was .42 (‘‘One will succeed if he/she really tries’’) and the low was .23 (‘‘Caution helps avoid mistakes’’). Six items define fate control. The average item-total correlation across these six items for the 40 cultural/national groups was .29, with a high of .41 (Romanians) and a low of .12 (Pakistanis). Only two items yielded a single,

TABLE III Factor Solution for the Student Sample of 40 Cultural Groups

Item

1 Social cynicism

Powerful people tend to exploit others

.60

Power and status make people arrogant

.59

Kind-hearted people usually suVer losses

.57

Kind-hearted people are easily bullied

.53

People will stop working hard after they secure a comfortable life

.45

Old people are usually stubborn and biased

.45

The various social institutions in society are biased toward the rich

.44

It is rare to see a happy ending in real life

.44

To care about societal aVairs only brings trouble for yourself

.42

People deeply in love are usually blind

.39

Young people are impulsive and unreliable

.38

People may have opposite behaviors on diVerent occasions

2 Social complexity

3 Reward for application

.60

Human behavior changes with the social context

.54

One’s behaviors may be contrary to his or her true feelings

.54

One has to deal with matters according to the specific circumstances

.48

Current losses are not necessarily bad for one’s long-term future

.40

There is usually only one way to solve a problem

.39

One will succeed if he or she really tries

.63

Hardworking people will achieve more in the end

.59

4 Religiosity

5 Fate control

Adversity can be overcome by eVort

.56

Every problem has a solution

.50

Knowledge is necessary for success

.49

One who does not know how to plan his or her future will eventually fail

.45

Competition brings about progress

.42

Failure is the beginning of success

.40

Caution helps avoid mistakes

.36

Belief in a religion helps one understand the meaning of life

.75

Religious faith contributes to good mental health

.72

There is a supreme being controlling the universe

.62

Belief in a religion makes people good citizens

.61

Religion makes people escape from reality

.59

Religious beliefs lead to unscientific thinking

.54

Religious people are more likely to maintain moral standards

.51

Individual characteristics, such as appearance and birthday, aVect one’s fate

.60

There are many ways for people to predict what will happen in the future

.60

There are certain ways to help us improve our luck and avoid unlucky things

.52

Most disasters can be predicted

.51

Fate determines one’s successes and failures

.48

Good luck follows if one survives a disaster

.48

Note. Some items are recoded, so that all primary loadings are positive regardless of the actual wording. All secondary loadings are lower than .3. The variances accounted for by these five factors are 7.49% (social cynicism), 4.73% (social complexity), 6.48% (reward for application), 7.25% (religiosity), and 5.06% (fate control).

TABLE IV Factor Solution of the Adult Sample Based on the Items Identified in the Student Solution

Item

1 Social cynicism

Kind-hearted people usually suVer losses

.62

Power and status make people arrogant

.62

Powerful people tend to exploit others

.61

Kind-hearted people are easily bullied

.60

People will stop working hard after they secure a comfortable life

.47

It is rare to see a happy ending in real life

.46

2 Social complexity

3 Reward for application

4 Religiosity

5 Fate control

To care about societal aVairs only brings trouble for yourself

.42

The various social institutions in society are biased toward the rich

.39

People deeply in love are usually blind

.32

Young people are impulsive and unreliable

.31

.39

Old people are usually stubborn and biased

.28

.26

People may have opposite behaviors on diVerent occasions

.59

One’s behaviors may be contrary to his or her true feelings

.54

Human behavior changes with the social context

.50

Current losses are not necessarily bad for one’s long-term future

.50

There is usually only one way to solve a problem

.38

One has to deal with matters according to the specific circumstances

.38

.29

.62

Adversity can be overcome with eVort One will succeed if he or she really tries

.61

One who does not know how to plan his or her future will eventually fail

.55

Every problem has a solution

.55

Hardworking people will achieve more in the end

.49

Knowledge is necessary for success

.47

Competition brings about progress

.47

Caution helps avoid mistakes

.29

Failure is the beginning of success

.25

.27

Belief in a religion helps one understand the meaning of life

.75

Religious faith contributes to good mental health

.71

Belief in a religion makes people good citizens

.67

There is a supreme being controlling the universe

.62

Religion makes people escape from reality

.58

Religious people are more likely to maintain moral standards

.56

Religious beliefs lead to unscientific thinking

.26

.54

There are many ways for people to predict what will happen in the future

.62

Individual characteristics, such as appearance and birthday, aVect one’s fate

.54

Good luck follows if one survives a disaster

.50 .50

There are certain ways to help us improve our luck and avoid unlucky things Fate determines one’s successes and failures Most disasters can be predicted

.34

.44 .43

Note. Some items are recoded, so that the primary loadings are positive regardless of the actual wording. Only loadings larger than .25 are presented. The variances accounted for by these five factors are 7.66% (social cynicism), 4.74% (social complexity), 6.32% (reward for application), 7.84% (religiosity), and 5.53% (fate control).

150

KWOK LEUNG AND MICHAEL HARRIS BOND

negative item-total correlation each of .003 and .04. With regard to the average item-total correlations for these six items, the high was .34 (‘‘Individual characteristics, such as one’s appearance and birthday, aVect one’s fate’’) and the low was .26 (‘‘Most disasters can be predicted’’). Six items define social complexity. The average item-total correlation for these six items across the 40 cultural/national groups was .23, with a high of .37 (Singaporeans) and a low of .04 (Iranians). Five items yielded eight negative item-total correlations, but only two are worse than .10. With regard to the average item-total correlations for these six items, the high was .30 (‘‘People may have opposite behaviors on diVerent occasions’’) and the low was .16 (‘‘There is usually only one way to solve a problem’’). See Table V for means of these axiom dimensions of the 40 cultural groups.

C. FACTOR SIMILARITY VERSUS FACTOR CONGRUENCE Before delving into the meaning of these dimensions of social axioms, a methodological issue concerning the notion of factor similarity must be addressed. Any claim of universality of a factor structure requires its congruence (being identical in a statistical sense) across diVerent cultural groups. The factor structure we have identified by means of the metaanalytic procedure is optimal for these 40 cultural groups considered as a whole. This factor structure is very stable because of a high degree of similarity between the structures based on student and adult data and between the current structures and the structure reported by Leung et al. (2002). The meta-analytic procedure guarantees a reasonably good fit between the factor structure and the data from most of the constituent cultures, but it is possible that for a small set of cultures, a modified structure would provide a better fit to the data. Perhaps to some statistical purists, we have not identified a factor structure that is ‘‘congruent’’ across the 40 cultural groups studied. Technically speaking, this criticism is valid, but we believe that in conducting global studies involving many cultural groups, it is impractical to set perfect congruence as the standard in the search for universal structures. Cross-cultural research involves many steps that are likely to generate random and systematic errors, such as translation and procedures followed in administering questionnaires (Van de Vijver & Leung, 1997), which significantly jeopardize any real chance to identify a congruence structure, even if it does exist. Recently, for example, Heine, Lehman, Peng, and Greenholtz (2002) found that people tend to evaluate themselves relative to similar others, and this social comparison tendency may constitute a source of error

151

SOCIAL AXIOMS TABLE V Citizen Axiom Scores of the Student Sample Social cynicism

Social complexity

Reward for application

Religiosity

Fate control

American (Caucasian)

2.65

4.10

3.66

3.18

2.46

Belgian

2.97

4.03

3.36

2.58

2.58

Brazilian

2.81

3.98

3.54

3.39

2.49

British

2.75

4.11

3.46

2.81

2.35

Canadian

2.63

4.20

3.74

3.10

2.43

Chinese

3.03

4.08

3.74

2.92

2.90

Czech

2.77

4.10

3.29

3.10

2.62

Dutchman

2.62

4.18

3.18

2.73

2.56

Estonian

3.16

4.11

3.81

2.70

2.81

Filipino

2.84

4.09

4.03

3.52

2.60

Finn

2.76

4.08

3.59

3.07

2.54

French

3.05

4.08

3.56

2.60

2.62

Georgian

3.37

3.88

3.69

3.65

3.00

German

3.32

4.33

3.76

2.93

2.77

Greek

3.32

4.02

3.73

3.13

2.37

Hong Kong Chinese

3.13

4.08

3.70

3.44

2.69

Hungarian

2.96

4.13

3.40

2.99

2.67

Indian

3.04

3.92

4.19

3.37

2.97

Indonesian

2.72

3.96

4.14

4.22

2.91

Iranian

2.89

3.79

4.12

4.15

2.85

Israeli

2.76

4.16

3.60

2.60

2.53

Italian

2.74

4.01

3.28

2.72

2.29

Japanese

3.16

4.04

3.50

2.65

2.59

Korean

3.16

3.98

3.85

3.10

2.98

Latvian

3.05

4.02

3.58

3.10

2.77

Lebanese

3.05

4.11

3.77

3.10

2.47

Malaysian

2.88

3.93

4.29

4.30

2.96

New Zealander

2.77

4.14

3.59

2.83

2.34

Nigerian

2.98

3.89

4.04

3.67

3.08

Norwegian

2.66

4.37

3.53

2.55

2.01

Pakistani

3.29

3.77

4.15

4.40

3.15

Peruvian

3.29

3.67

3.88

3.21

2.48

Portuguese

2.87

3.90

3.61

3.09

2.43

Citizen

(continues)

152

KWOK LEUNG AND MICHAEL HARRIS BOND TABLE V (continued ) Social cynicism

Social complexity

Reward for application

Religiosity

Fate control

Romanian

3.23

3.72

3.74

3.29

2.55

Russian

3.09

3.86

3.82

3.12

2.97

Singaporean

2.93

4.14

3.78

3.24

2.52

Spanish

2.89

4.14

3.48

2.40

2.27

Taiwanese

3.30

4.22

3.87

3.22

3.01

Thai

3.22

3.80

3.98

3.43

3.14

Turk

2.94

4.14

3.97

3.48

2.68

Citizen

in cross-cultural comparison. Finally, cultural diVerences may actually give rise to a slightly diVerent twist and slant to similar notions, thereby making perfect congruence highly unlikely in the real world. Our goal in this multicultural project on social axioms is to identify and itemize factors across a wide range of cultures, and consequently the items identified in the five-factor, pan-cultural factor structure are unlikely to be the best set of items that could be used to define these factors for some of the cultural groups. Obviously, locally derived items should show a higher level of reliability and validity for mapping these social axioms for these cultural groups (e.g., Cheung & Leung, 1998). However, we present some evidence in subsequent sections to show that our pan-cultural items are adequate markers for the identified axiom dimensions. This logic also has been adopted by some researchers engaged in global studies. In his search for universal value types, Schwartz (1992) also adopted a logic of similarity rather than a logic of congruence in assessing the usefulness of specific values as universal markers for the value types. Schwartz found that 87% of these value items were located in the correct value type in 70% or more of the samples. In other words, some mismatches occurred between the markers and the value types in some cultural groups. Our view is that this ‘‘good enough’’ logic in the operational definition of universal markers does not weaken Schwartz’s near-universal claim for his value types. In fact, Spini (2003) has recently examined the measurement invariance of the value structure proposed by Schwartz (1992) by means of confirmatory factor analysis across 21 countries. With an approach that is moderately strict from a statistical point of view, Spini concluded that most of the value types showed measurement invariance. However, perfect congruence has not been achieved, and Spini concluded that cross-cultural

SOCIAL AXIOMS

153

invariance in this case is like deciding whether the glass is half empty or full. In his words, ‘‘If the aim of research is to set forth a definitive theoretical model of the cultural diVerences on the basis of universal and shared value, the glass is probably half empty. . . . But if we take the view . . . that research gradually approaches reality, then the present research represents more than a half-full glass’’ ( pp. 20–21). Another example is derived from the research on the universality of the big-five model of personality orchestrated by McCrae, Costa, and colleagues (see McCrae, 2002). Five dimensions of personality traits are claimed to be fundamental and hence universal: agreeableness, extroversion, neuroticism, conscientiousness, and openness. The cross-cultural generality of this model using the NEO-PI-R is typically tested at the facet rather than the item level (e.g., McCrae, Zonderman, Costa, Bond, & Paunonen, 1996). In other words, facet means are obtained by averaging the items defining the facets, which then become the basic unit of factor analysis conducted to establish the cross-cultural congruence of the big-five model. To the best of our knowledge, there is no attempt to evaluate whether the items defining the facets are congruent across the diverse cultures in which the big-five model has been examined. Technically, it is extremely hard, if not impossible, to confirm a factor model at the item level. The chance of obtaining factor congruence across a large number of cultures with a 240-item instrument, the NEO-PI-R, is probably next to zero, regardless of whether confirmatory factor analysis or Procrustes rotation is used. Nonetheless, in our view, the five-factor model shows impressive similarity across the diverse range of cultures studied.

D. CONTENT OF THE PAN-CULTURAL DIMENSIONS We now consider these five identified dimensions of human belief separately and as a totality. This discussion is primarily based on the content of the items and their conceptual linkage to previous research imbuing these dimensions with meaning. Evidence from within and across cultures will be presented later to enrich understanding of these belief dimensions and the psychological space they attempt to map. 1. Social Cynicism Many of these items are concerned with the corrosive eVects of power or authority derived from wealth or age, orienting others toward self-absorption and indiVerence to their fellow citizens. There are, however, additional

154

KWOK LEUNG AND MICHAEL HARRIS BOND

features about the uselessness of showing goodwill toward others and the inevitable frustration of charity and public spiritedness. This belief syndrome suggests a lay version of Hobbesian theory—‘‘nature red in tooth and claw,’’ of social life as a jungle of willful, callous, indolent, and foolish others requiring vigilance and skepticism for one’s survival. Conceptually, this belief construct is reminiscent of Machiavellianism (Christie & Geis, 1970), La Piere’s (1959) Freudian ethic, and Wrightsman’s (1992) first and third assumptions about human nature: untrustworthiness and selfishness (versus altruism). These overlappings are limited, however, by two considerations. First, the previous scales measuring these concepts often include non-belief items tapping values or behavioral intentions and self-reports. Social cynicism as a construct of belief may well predict domains of value and aspects of behavior, hence accounting for the intercorrelations among items measuring these diVerent constructs. However, this conflation of constructs undercuts the development of a model for behavior where beliefs, values, and other constructs will be linked to behaviors of interest. Second, the focus of other ‘‘cynicism’’ scales (e.g., that in the MMPI) is on the human actor himself or herself, and generally excludes items of belief about social, material, or spiritual factors in determining the outcomes of human life. Whether the social cynicism measured in the Social Axioms Survey is a projection of the perceiver’s personal cynicism is a conceptual and empirical issue worthy of further study.

2. Religiosity The contents of this belief dimension focus on the positive functions of religious belief in six of its seven items. The seventh item asserts the existence of a supreme being, which is a central tenet in the world’s monotheistic religions. As the dimension highlights the psychological and social consequences of religious belief, however, we think it apt to rename this factor from its former name of spirituality to religiosity. Religious belief and practice are features of all cultural groups. Sociologists have argued that they provide a conservative social thrust and exercise integrative social eVects on communities (Berger, 1967), while psychologists have argued that they provide meaning and a sense of shared purpose for the followers of the religion (Solomon, Greenberg, & Pyszczinski, 1991). Just as many social scientists would personally disagree with these positions, so too do the citizens in any given nation, thereby yielding variation across this belief syndrome. Measures of religious belief (Argyle, 2000) tend to be culture-specific in the sense that they assess endorsements of belief in the dominant religion of

SOCIAL AXIOMS

155

that culture. In addition, they tend to conflate measures of value and behavior (religious practice) along with beliefs about religion generally or the particular religion of the culture. We expect that our measure of religiosity will correlate with all these constructs but prefer the construct separation that a pure and more general measure of religiosity provides. 3. Reward for Application The nine items defining this core etic (a universal that is culture-general) all reflect an optimism that the challenges of life can be resolved by human endeavor and the application of individual resources to their solution. There are parallels in the content of these items to those items assessing coping styles (Lazarus & Folkman, 1984), the Protestant Work Ethic (Furnham et al., 1993), just world beliefs (Lerner, 1980), and especially internality in Rotter’s (1966) measure of internal-external locus of control that has inspired so much research (e.g., Phares, 1965). However, the items from the Social Axioms Survey do not ask the respondents to assess their own level of internality, their own level of personal self-eYcacy (Bandura, 2001), or their own level of application (Yik & Bond, 1993), but rather their understanding about the interface between human agency, whoever its origin, and environmental responsiveness. In this respect, Singelis, Hubbard, Her, and An (2003) did not find a significant relationship between reward for application and internality as measured by Rotter’s (1966) instrument, although the correlation was in the expected direction. There are some possible reasons for this disconfirmation of expectation. Rotter’s scale is itself multifactorial, so that a single score derived from totaling all its items represents a conflation of constructs (see e.g., Bond & Tornatsky, 1973). Furthermore, some of the items in Rotter’s scale ask about ascriptions of internality to the self; others ask about ascriptions of internality in the lives of people in the world. It is likely that reward for application does not overlap substantially with internality as measured by Rotter, and probably their distinctiveness arises for conceptual reasons. Its meaning awaits the results of future studies done both within and across cultures. 4. Fate Control A close inspection of the items constituting this factor suggests two major themes: the predictability of important outcomes in one’s life and the ‘‘fatedness’’ of one’s outcomes. Several items are concerned with the extent that individuals believe in the predictability of important outcomes, and the actions that they as individuals can undertake to shape them. The word fate

156

KWOK LEUNG AND MICHAEL HARRIS BOND

appears in two of the six items, and fate in these contexts also refers to a pervasive force determining one’s life outcomes. There are numerous scales tapping fate-related concepts in our literature, such as Dake’s (1992) fatalism or Wrightsman’s (1992) external control and irrationality. In common with others, Pepitone and SaYotti (1997) defined their concept of fate as a term ‘‘invoked to explain those life events that are perceived to be predestined and wholly under the control of some external power’’ (p. 25). The key word in their definition is wholly, a restriction that would exclude the possibility of controlling fate itself. But controllability of fate is precisely what is involved in one of the items defining fate control, viz., ‘‘There are certain ways to help us improve our luck and avoid unlucky things.’’ So, in the lay epistemology of the world’s peoples, even fate has a degree of controllability, being subject to knowledgeable forms of human agency, and our scientific constructions of the fate concept may be missing an important component of beliefs about outcomes with respect to their controllability by special practices or rituals. For this reason, we prefer to honor our respondents’ pan-cultural association of fate and other outcomes with their controllability and propose a more influencible conception of fate in our interpretation of this factor of fate control. Fate controls outcomes but is itself controllable. 5. Social Complexity The interpersonal world is complicated for some persons; individual behavior may vary from time to time and context to context and not map onto the individual’s feelings, thoughts, or traits in obvious ways. Present outcomes may not predict future outcomes, and the individual must deal with events on a case-by-case basis. The logic underlying events is not transparent, and daily life requires complex, lay theories for some people so that they can eVectively negotiate daily exchanges. Other people do not believe that the above is an accurate description of the world, instead assessing their worlds as constituted by a simpler, more easily predicted set of events. This contrast in world views gives social complexity its dimensionality. In associating to this construct, one is reminded of Openness to Experience from the NEO-PI-R (Costa & McCrae, 1992), need for closure (Kruglanski, 1989), and cognitive flexibility (Applegate, Kline, & Delia, 1991; Martin & Anderson, 1998). We must again be alert to the varied composition of these scales, often involving items tapping values, attitudes, behavioral intentions and self-reports, in addition to beliefs. Perhaps the closest parallel to social complexity then is Wrightsman’s (1992) complexity versus simplicity measure of assumptions persons make about the world. All these measures have inspired a host of studies in the social

SOCIAL AXIOMS

157

psychology of attribution, especially around the concept of attributional complexity (Fletcher, Danilovics, Fernandez, Peterson, & Reeder, 1986).

V. Evidence for the Meaning and Usefulness of the Social Axiom Dimensions Two diVerent strategies exist to evaluate the meaning and usefulness of these five axiom dimensions. The first strategy is rarely used in social psychology, because most social psychological research is monocultural. In the present research, however, we have collected comparable university student data from 40 cultural groups, and each group may be represented by a ‘‘citizen’’ axiom profile, that is, the average axiom score across male and female university students for the five axiom dimensions as identified in the above, pan-cultural analysis. So, by using university students as surrogates for representative samples from these 40 countries, we can create ‘‘citizen’’ scores for social cynicism, religiosity, reward for application, fate control, and social complexity. Using culture as a unit of analysis, we can correlate these citizen scores with a wide range of country-level indexes, building in essence a nation-level nomological network for interpreting the meaning of these axiom dimensions. This strategy targets each dimension of axioms separately, correlating citizen scores with societal features of a nation. This is a well-established approach in the study of modernity where variations across a dimension of values, like post-modernism (Inglehart, 1997), are linked to a measure of societal modernization, such as GNP/capita or a broader measure like Georgas, van de Vijver, and Berry’s (2004) factor of aZuence, a composite measure tapping level of economic development. Often the selection of societal correlates of psychological functioning is driven by theoretical considerations about how social systems and their institutional components promote the development of compatible psychological supports, like Durkheim’s (1893/1984) work on industrialization and alienation, Fukuyama’s (1995) on capitalism and trust, or Inglehart’s (1997) on modernization and post-modern values. We expect, for example, that some of our axiom dimensions form one of the psychological underpinnings of democracy (Sullivan & Transue, 1999). This speculation may be assessed by correlating the axiom dimensions with measures of a nation’s level of freedom or of democracy. Two issues associated with this approach need to be addressed. First, it is well known that individual-level and culture-level analysis bear no logical relationship with each other (e.g., Leung, 1989). The average citizen scores are based on an individual-level factor analysis, in which an individual is the

158

KWOK LEUNG AND MICHAEL HARRIS BOND

unit of analysis. However, the average citizen scores are at the culture level, and when we correlate average citizen scores with societal indexes, the analysis is obviously at the culture level. This approach may be viewed as an example of mixing levels, but we argue that this procedure is legitimate. Although the citizen scores are based on an individual-level factor structure, they represent meaningful constructs at the culture level. For example, in some cultures many citizens have a high score in social cynicism, whereas in some other cultures, most people are low in social cynicism. Intuitively, it seems plausible that these two groups of cultures are likely to diVer in some aspects at the societal level, which can only be revealed by correlating the average citizen scores with relevant societal indexes. Second, we should note that university students are hardly representative of their country’s population. Nonetheless, we propose to use their average scores as an approximation of the true citizen scores on the core etics of axioms for their country, assuming that the relative position of countries’ citizens, perhaps even their degree of diVerence from one another, will not diVer much from these scores. To evaluate this possibility, we have correlated the student and adult means for each of the axioms across the 11 cultural groups for which we have both types of data. The average correlation is supportive of our argument, with a mean of .83, a high of .89 (fate control), and a low of .68 (social complexity). The second strategy to evaluate the meaning of the axiom dimensions is common in social psychology, which involves the study of their antecedents and consequences across individuals. The evidence gathered so far in this line of work is reviewed in Section VII.

A. SOCIETAL VARIABLES ASSOCIATED WITH CITIZEN AXIOMS The institutional programs set up by society are subjectively real as attitudes, motives and life projects. —Peter Berger, The Sacred Canopy: Elements of a Sociological/Theory of Religion

Bandura (2002) has argued that, ‘‘Social structures are created by human activity to organize, guide and regulate human aVairs in given domains by authorized rules and sanctions’’ (p. 278). Social axioms are an individual cognitive form of organization, guidance, and regulation that would facilitate adaptation to cultural environments characterized by certain reinforcement conditions. Which of the five core etics, however, is responsive to which societal features? We have grouped the country-level indexes into two broad categories: socioeconomic-political indexes, some of which may

SOCIAL AXIOMS

159

not have any individual counterparts, as well as psychological indexes derived from aggregating psychological data across individuals. Before we present the results, we should note with Yang (1988) that there will be many psychological adaptations to the forces of modernization, although only those that bear some functional relationship to the institutions and social processes attendant on modernization will change (for a recent evaluation of this position, see Chang, Wong, & Koh, 2003). A major aspect of all cultural environments is the means of production used to generate economic output in that social system. More developed contemporary economic systems are characterized by their diversity, interdependency, regulation by laws and contract, focus on service and knowledge provision, scientism, and professionalization of the workforce (Landes, 1999). Socializing a workforce to function eVectively in such a complex environment should have a broad impact on the psychological functioning that characterizes its citizenry, its skills, its emotional experience (Wong & Bond, 2003), its values (Inglehart, 1997), and its beliefs. Yang (1988) has warned that modernization involves more than economic change, so we expect that additional, societal factors will shape citizen beliefs—religious traditions, especially their assumptions about the nature of man and their authority structures; philosophical teachings, especially those related to epistemology; ecological features, like disaster proneness and resource-richness; historical legacies, like colonization, immigration flows, domestic political violence, and interaction patterns of ethnic groupings within their borders. Measures of these various societal factors are challenging to develop, and when they are, we may discover that they themselves are correlated with measures of economic development. One strategy for disentangling these interrelated groupings is to control for aZuence in examining the correlations between the axiom dimensions and country-level indexes. We use a nation’s per capita GDP as a measure of its aZuence in calculating the partial correlations, which is a common strategy in this line of research (e.g., Van de Vliert, 2003). We should also point out that we rely on correlations to analyze the data because we are still in an exploratory stage, and complex analyses such as regression analysis may blur some interesting relationships. In addition, we have not grouped the societal indexes into categories because a factor analysis of these indexes has produced a pattern that is hard to interpret. In essence, our analytic strategy is to present the basic relationships, which are intended to lay the foundation for future work that will be more sophisticated and theory guided. Finally, we have included a wide range of indexes, but we only present the indexes that yield significant correlations. Given that many correlations were computed, a small number may be spurious, and we have dropped a few indexes because their results are hard to interpret.

160

KWOK LEUNG AND MICHAEL HARRIS BOND

B. CORRELATIONS WITH SOCIETAL VARIABLES The correlations of the citizen scores with the socioeconomic-political indexes are given in Table VI (see the Appendix for the sources of these indexes). As expected, aZuence as measured by per capita GDP correlated significantly with all five citizen scores. It may seem paradoxical that the same predictor works for all five axioms when the axiom dimensions from which citizen scores are derived are statistically independent of one another. That independence, however, is based on the factor analysis of the individual belief data from 40 cultural groups. That factor analysis enabled us to calculate factor scores for the individuals in all countries that were then averaged to produce the citizen scores. In fact, these five axioms based on citizen scores show some intercorrelations. Not surprisingly then, they sometimes show significant correlations with a given variable, as in the present case of aZuence. The fact that economic development as measured by national aZuence is related to the social axioms of citizens across all five dimensions suggests a ‘‘profile of beliefs for modern economies’’ characterized as low in social cynicism, religiosity, reward for application, and fate control, but high in social complexity. The negative correlation between reward for application and aZuence seems counterintuitive, because reward for application may be expected to promote economic development. However, this finding suggests that in an aZuent environment, people actually posit a weaker link between eVort and return. Perhaps measures that have been introduced in aZuent societies to equalize the impact of wealth, such as social welfare programs for the poor and high tax for the wealthy, are partly responsible for a weakened belief in reward for application in these societies. In less aZuent societies, however, people typically must fend for themselves, and more work is likely to lead to more return, giving rise to a stronger belief in reward for application. As discussed previously, aZuence may also show significant eVects on a wide range of societal phenomena and hence confound the relationships between social axioms and country-level indexes. To control for these eVects, partial correlations are reported, with per capita GDP being controlled for. In other words, significant relationships identified in this analysis are independent of a nation’s aZuence. A growing body of multicultural studies of individual psychological variables exists, ranging from McCrae’s (2002) presentation of big-five scores in 40 countries to Levine and Norenzayan’s (1999) analysis of pace of life in 31 countries. These psychological data are collated and analyzed to yield a nomological network of social beliefs in a latticework of other citizen characteristics. The correlations of the citizen scores with these psychological and behavioral indexes are given in Table VII. Again, per capita GDP was controlled for in these correlations.

TABLE VI Correlations Between Social Axioms and Socioeconomic-Political Indicators

Variable

Source

N

GDP per capita 2000 (PPP US$)

Human Development Report 2002, UN; The World Fact Book 2002

40

Average daytime temperature

National Geographic Atlas of the World 1990, as cited in Van de Vliert, Schwartz, Huismans, Hofstede, and Daan (1999)

Life expectancy at birth

Social cynicism .39

Social complexity .62*

Reward for application

Religiosity

Fate control

.62*

.62*

.60*

37

.45*

.41

Human Development Report 2001, UN

39

.39

.37

Population growth rate 2000–05

Statistical Division, UN

36

.55*

.49*

Number of persons per room 2001

Statistical Division, UN

23

.56*

.50

Urbanism 2000

Statistical Division, UN

39

Percent of GDP on education

Human Development Report 2001, UN

39

Percent of GDP on health

Human Development Report 2001, UN

32

Environmental sustainability index 2002

World Economic Forum (2002)

36

Human development index 1999

Human Development Report 2001, UN

39

.47*

.48*

.43*

Human rights

Humana (1992)

35

.56*

.43

.44*

.42*

.40 .33 .41

.56*

.67*

.36 .37

(continues)

TABLE VI (continued ) Variable Political rights and civil liberties 1992/93–2001/02

Source Freedom House (2002)

N

Social cynicism

Social complexity

39

Reward for application

Religiosity

.48*

.49*

.50*

.53*

Fate control

.42

Women’s status

Population Crisis Committee (1988)

35

Voter turnout at latest elections

Human Development Report 2000, UN

32

Working hours per week

International Labour Organization (2002)

28

Heart disease death rate

World Health Statistics Annual 1995–98

22

.52

Suicide rate

World Health Statistics Annual 1992–95

27

.54*

Alcohol consumption 1996

Human Development Report 2001, UN

36

.38

.45 .51*

.49*

.49*

.38

Note. All scales are scored so that a higher score indicates a higher level of the concept represented by the label. The correlations involving per capita GDP are simple correlations, and the rest are partial correlations with per capital GDP controlled for. All correlations are significant at the .05 level, and those with an asterisk are significant at the .01 level.

TABLE VII Correlations Between Social Axioms and Psychological Indicators at the Societal Level

Variable

Source

N

Social cynicism

Social complexity

Reward for application

Religiosity

Life satisfaction

World Value Survey 1990–93, as cited in Diener & Suh (1999)

21

Job satisfaction

International Survey Research (1995), as cited in Van de Vliert & Janssen (2002)

21

Satisfaction toward one’s company

International Survey Research (1995), as cited in Van de Vliert & Janssen (2002)

21

Positive aVect

World Value Survey 1990–93, as cited in Diener & Suh (1999)

24

.64*

Negative aVect

World Value Survey 1990–93, as cited in Diener & Suh (1999)

24

.45

Hedonic balance—Positive aVect minus negative aVect

World Value Survey 1990–93, as cited in Diener & Suh (1999)

24

.50

.47

.73*

.53

Pace of life

Levine & Norenzayan (1999)

19

Extraversion

McCrae (2002)

25

Agreeableness

McCrae (2002)

25

Conscientiousness

McCrae (2002)

25

Work ethic—Enjoyment of working hard

Lynn (1991)

22

Achievement via conformity

Lynn (1991)

22

Fate control

.69*

.55

.51

.60*

.50 .52

.49

.59*

.51 .54 .62* (continues)

TABLE VII (continued ) Variable

Source

N

Social Cynicism

Social Complexity

Reward for Application

Religiosity

Fate Control

Sources of guidance—Vertical (superiors)

Smith, Peterson, and Schwartz (2002)

32

.49*

Sources of guidance—Beliefs that are widespread in my nation

Smith, Peterson, and Schwartz (2002)

32

.42

Sources of guidance—Specialists

Smith, Peterson, and Schwartz (2002)

32

.45

View on leadership— Charismatic/value based

Den Hartog, House, Hanges, & Ruiz-Quintanilla (1999)

28

View on leadership—Humane

Den Hartog et al. (1999)

28

View on leadership—Teamoriented

Den Hartog et al. (1999)

28

Mate preference—Emphasizing mutual attraction; deemphasizing financial prospect, social status, and ambition

Shackelford & Schmitt (2002)

23

.74*

.54*

Mate preference—Emphasizing education and intelligence; deemphasizing desire for home and children

Shackelford & Schmitt (2002)

23

.53

.64*

Mate preference—Emphasizing sociability and pleasing disposition; deemphasizing religious background

Shackelford & Schmitt (2002)

23

.53

.65*

.46 .47

.46

.52* .72*

Tolerance of divorce

World Value Survey 1990–93, as cited in Diener, Gohm, Suh, and Oishi (2000)

25

.43

Percentage of thinking scientific advances will help mankind

Inglehart, Basan˜ez, and Moreno (1998)

23

.44

Importance of religion in life—% of very important

Inglehart et al. (1998)

24

.69*

Frequency of praying to God outside of religious services— % ‘‘often’’ or ‘‘sometimes’’

Inglehart et al. (1998)

20

.62*

Percentage of adult population that attends church at least once a week

World Value Survey 1990–93/ 1995–97, as cited in Swanbrow (1997)

26

Percentage of showing interest in politics

Inglehart et al. (1998)

23

In-group disagreement

Smith, Dugan, Peterson, and Leung (1998)

18

Other-referenced performance motive—compared with others’ performance

Lynn (1991), as cited in Van de Vliert & Janssen (2002)

22

.44

.53

.43

.55*

.43

.50 .60*

.68*

.67*

.49

Note. All scales are scored so that a higher score indicates a higher level of the concept represented by the label. All the correlations are partial correlations with per capital GDP controlled for. All correlations are significant at the .05 level, and those with an asterisk are significant at the .01 level.

166

KWOK LEUNG AND MICHAEL HARRIS BOND

1. Social Cynicism Higher social cynicism is related to less frequent church attendance, lower life satisfaction, lower satisfaction toward one’s company, lower hedonic balance (more negative than positive aVect), and a faster pace of life. Perhaps the reason for people with high social cynicism to show a fast pace of life is that they take a businesslike, transactional approach in conducting their lives. Social cynicism is related to a lower level of conscientiousness, a factor in the big-five personality model that is concerned with competence, order, dutifulness and discipline, and the will to achieve. Social cynicism is also related to lower achievement via conformity, rejection of the view that leadership is based on charisma and values, lower endorsement of teamoriented leadership, and more disagreement within the in-group. People high in social cynicism seem unmotivated to get along with others and unlikely to endorse the view that leaders should inculcate a set of values in their followers and guide them with a vision and a sense of meaning in their goal attainment activities. These correlations suggest that people high in social cynicism are likely to be lower in church attendance, unhappy, lower in achievement via conformity, more likely to run into interpersonal problems, have less faith in charismatic leaders, be less conscientious, and show a faster pace of life. This profile fits nicely with our earlier description of social cynicism as a combination of a negative view of human nature, a biased view against some social groups, a mistrust of social institutions, and a disregard of ethical means for achieving an end. People high in social cynicism seem unable to cope with their social world eVectively, resulting in considerable negative psychological outcomes.

2. Social Complexity A higher level of social complexity is related to a higher proportion of GDP spent on health, a higher voter turnout rate, more interest in politics, and a lower performance motive that takes into account the performance of other people. Our earlier description of social complexity suggests a combination of belief in the lack of rigid rules, the existence of multiple solutions to a problem, and the inconsistency in human behavior. These significant correlations do not lend themselves to an obvious interpretation based on this description. However, given that social complexity is correlated with economic development, these correlations may suggest a general pattern of modernity. We may speculate that the higher investment in health may reflect a complex view of health problems. The endorsement of multiple ways of doing things may reduce the need for considering others’ performance in monitoring one’s performance. Finally, a complex world

SOCIAL AXIOMS

167

view may predispose one to engagement in complex issues, such as politics. Obviously, these interpretations are highly speculative and tentative. 3. Religiosity Religiosity is related to a higher daytime temperature, a lower life expectancy, a higher population growth rate, a larger average number of people in a room, a lower percentage of GDP spent on health and education, less favorable human development, less human rights, fewer political rights and civil liberties, a lower status of woman, longer work hours, and less alcohol consumption. With regard to psychological variables, religiosity is related to higher positive and negative aVect, higher hedonic balance, a slower pace of life, higher agreeableness, a factor of the five-factor model of personality that refers to friendliness and kindness. Religiosity is also related to stronger endorsement of humane leadership, stronger endorsement of the view that scientific advances will help mankind, a higher importance of religion in one’s life, more frequent oVering of prayers to God, more frequent church attendance, and a stronger performance motive that takes into account the performance of others. Even though wealth has been controlled for in these correlations, there is a clear pattern that religiosity is associated with living in a diYcult socioeconomic environment. Perhaps when life is diYcult, people are more likely to discover the benefits of religious practice. In addition, the results clearly support our interpretation of religiosity in that people high in religiosity indeed engage in more religious activities and regard religion as more important in their life. In addition, they also experience more aVect in their daily life, report a slower pace of life, consume less alcohol, see themselves as more agreeable, and endorse a humane view of leadership, A few correlations are hard to interpret, including the link between religiosity and temperature, a performance motive based on others’ performance, and confidence in science. Perhaps high daytime temperature is related to discomfort in daily life (e.g., Van de Vliert, 2003), which may give rise to higher religiosity. Religiosity may also lead to a higher other-focus, resulting in a performance motive that is sensitive to the performance of others. Finally, religiosity is related to diYcult socioeconomic conditions, which may lead to placing more hope in science to improve human conditions. In any events, these interpretations are highly speculative and require future verification and refinement. 4. Reward for Application Reward for application is related to a higher daytime temperature, a lower life expectancy, a higher population growth rate, a larger average number of people in a room, a lower percentage of GDP spent on health, less favorable

168

KWOK LEUNG AND MICHAEL HARRIS BOND

human development, less human rights observance, fewer political rights and civil liberties, a lower status of woman, longer work hours, and lower alcohol consumption. With regard to psychological variables, reward for application is related to higher reliance on vertical sources (superiors) and widespread belief in one’s cultural traditions, and lower reliance on specialists to guide one’s action at work. Reward for application is also related to lower emphasis on mutual attraction, education and intelligence, as well as sociability and pleasing disposition, in mate preference, lower tolerance for divorce, stronger endorsement of humane leadership, a higher level of agreeableness, a stronger performance motive that takes into account others’ performance, and a stronger belief that scientific thinking will help mankind. It is interesting to note that, like religiosity, diYcult socioeconomic conditions are associated with a stronger belief in reward for application. In fact, the relationships with regard to this type of variable are highly similar to those for religiosity. Perhaps the belief in the benefits of hard work, knowledge, and careful planning is another strategy to cope with socioeconomic hardship. Reward for application is also related to lower tolerance for divorce and less emphasis on the role of mutual attraction, intelligence, and pleasing disposition in the selection of a spouse. Perhaps strong endorsement of this dimension leads to the view that eVort to establish a viable family can compensate for meager romantic feelings in intimate relationships. The emphasis on self-discipline by people high in reward for application may explain their reluctance to rely on specialists and their low alcohol consumption. In addition, people high in reward for application are obviously sensitive to how others are going to provide rewards for them. It is logical that they look to superiors and social conventions for guidance at the workplace and monitor their performance in light of the performance of other people. This interpersonal sensitivity is also consistent with their higher agreeableness and stronger endorsement of humane leadership. 5. Fate Control Fate control is related to a lower life expectancy, lower urbanism, a lower percentage of GDP spent on health, lower environmental sustainability, lower human development, less human rights observance, a lower status of woman, a lower voter turnout rate, a higher death rate due to heart disease, and a higher suicide rate. With regard to psychological variables, fate control is related to lower job satisfaction and lower satisfaction toward one’s company, a faster pace of life, and lower extraversion, a factor of the five-factor personality model that refers to outgoingness and sociability. Fate control is also related to lower

SOCIAL AXIOMS

169

work ethic (enjoyment of working hard), lower endorsement of teamoriented and charismatic leadership, and a lower emphasis on mutual attraction, education and intelligence in mate preference. Fate control is also related to more interest in politics, and a performance motive that takes into account others’ performance. Like religiosity and reward for application, diYcult socioeconomic conditions are associated with a strong belief in fate control. Perhaps the belief that life events are predetermined and that there are ways to influence these outcomes represents a way to cope with these diYculties. Consistent with the notion that people high in fate control do not take active control of events that happen to them and adopt a passive style in dealing with these events, high fate control is related to a higher suicide rate, a higher death rate due to heart disease, a lower voter turnout rate, a lower life satisfaction and satisfaction toward one’s company, a less active social life as reflected in low extraversion, lower endorsement of team-oriented and charismatic leadership, lower emphasis on education and intelligence in mate selection, and less positive work ethics. The positive correlations between fate control and interest in politics, other-focus in performance motive, and pace of life may seem puzzling at the first glance, but given that fate control does involve a belief in altering one’s fate, these correlations may reflect their attempt to positively shape their life outcomes. What remains unknown, of course, is when people high in fate control are passive and withdrawing, and when they attempt to exercise control over life events, an important topic for future research. Given that the patterns of correlations for religiosity, reward for application, and fate control are similar for quite a few variables, the question may be raised whether they really present distinct constructs. It seems that for ‘‘diYcult’’ environments, which are relatively less wealthy, resourceful, and stable (Triandis, 1973), reward for application, religiosity, and fate control are high, which may be regarded as coping responses in these diYcult environments. Furthermore, the psychological correlates of these three axioms suggest that they are related to a higher level of tolerance and interpersonal sensitivity. However, we argue for the distinctiveness of these three dimensions for two main reasons. First, they are derived from an orthogonal rotation of individual-level responses, and their correlates should be relatively distinct only at the individual level. As explained previously, the results presented here are at the national/cultural level, which explains why these citizen variables are related at a higher than expected level. For instance, the correlation between citizen religiosity and citizen reward for application is the highest, which, after controlling for aZuence, is .64. In a subsequent section, we will review evidence that supports their distinctiveness at the

170

KWOK LEUNG AND MICHAEL HARRIS BOND

individual level. Second, these three dimensions do show some unique correlations that clearly reflect their distinctiveness. For instance, religiosity is related to observance of religious activities, whereas reward for application to the reliance of superiors for guidance at work and intolerance of divorce, and fate control to suicide and death due to heart diseases.

C. SECTION SUMMARY Each of these axiom dimensions shows generally interpretable, meaningful relationships with a subset of the societal indexes, lending support to our interpretation of their meaning. We should point out that we were somewhat constrained in our analyses by the societal indexes currently available. New indexes may have to be tailormade for our particular purposes for future work in this area, as Simonton (1976), for example, has been attempting. The present work on social axioms along with other recent multicultural work on psychological constructs (e.g., by Allik and McCrae (2004) on the big-five model of personality), will stimulate their development.

VI. Clustering Nations in Terms of Their Citizens’ Beliefs The citizen scores on these five social axioms may be used in a diVerent way to illuminate the impact of societal conditions on beliefs. A cluster analysis of belief scores across the five dimensions will group citizens on the basis of their similarity to one another. This pattern of grouping will provide a ‘‘psychograph’’ of ‘‘axiom mates,’’ stimulating conjectures about the possible basis for the groupings. A hierarchical cluster analysis of the citizen’s z scores on the five belief dimensions across 40 nations was conducted, based on the method of average linkage. The pattern obtained in this analysis is given in Figure 2. Perhaps the monumental work of Hofstede (1980) has cast a long and rigid shadow on our lay theory about how cultures should group together. Most people’s first reaction to the dendrogram is likely to be a mixture of unfamiliarity and bewilderment. Only a few obvious linkages are apparent in this psychograph of axiom similarities: the Muslim group of Iranians, Malaysians, Indonesians, and Pakistanis, although the linkage with the Pakistanis takes a number of steps to achieve; the Brazilians and the Portuguese are linguistic neighbors; the Czechs and the Hungarians are geographical neighbors; the Canadians and the Americans are both linguistic and geographical

SOCIAL AXIOMS

171

Fig. 2. Hierarchical cluster analysis of citizens’ social axioms on five dimensions from 40 cultural groups.

172

KWOK LEUNG AND MICHAEL HARRIS BOND

neighbors; there is a Chinese grouping, but it, like the Muslim grouping, takes a number of steps to achieve, and shares its cluster with a number of disparate cultural groups. More striking are pairings and groupings that defy simple interpretations based on previous cultural maps: Peruvians with Romanians, Canadians and Americans with Finns, French with Japanese, and Israelis with New Zealanders and British; particularly striking is the clustering of Indians, Nigerians, Koreans, Russians, and Thais. The surprise at these sorts of groupings is relative to implicit models of how cultural systems and their psychological sequelae ought to organize themselves—geographically, linguistically, religiously, or in terms of diasporas. In fact, a primary objective of our research project is to search for results that are distinct from previous results. It is our hope that the cultural groupings presented will compel the field to explore new bases for organizing cultures and their associated dynamics.

VII. Roles of Social Axioms for Individual Behavior Thoughts are but dreams till their eVects be tried. —Shakespeare, The Rape of Lucrece

This section reviews the antecedents and consequences of social axioms at the individual level, at which most social psychological research operates. Traditional personality and social psychological research is conducted using participants from a single cultural group. Individuals, rather than nations, cultural groups, or the average citizen, become the focus of attention and analysis. Using the individual as the unit of analysis is a familiar format for psychologists and enables us to produce theories and findings about individuals rather than about other social units. Occasionally, such within-culture studies are extended to other cultural groups to assess the presumed generalizability of such discoveries. Often these processes are found to be ‘‘universal,’’ but sometimes not (Smith & Bond, 1998). Specific to our context, we hypothesize that a person’s endorsement levels for the five social axioms represent generalizations arising from a confrontation between his or her eco-social aVordances and his or her deployable skill repertoire. They are experienced-based judgments about how his or her world works with respect to these five axes of diVerence. A person’s position along these dimensions informs us about his or her assessment of the reality constraints within which he or she operates in the world. Once in place, these

SOCIAL AXIOMS

173

levels of general beliefs along the five dimensions will channel individual behavioral choices.

A. WITHIN-NATION STUDIES OF SOCIAL AXIOMS Identifying general beliefs and motivations as separate constructs in predicting behavior implies that they are nonoverlapping constructs, both useful in the prediction of other psychological outcomes. In a number of Hong Kong–based studies, social axioms have been shown to be only moderately related to values, suggesting a complementary role for these constructs, as would be hypothesized by expectancy-value theory (Feather, 1982). In a recent study by Keung and Bond (2002), the degree of overlap between two dimensions of value derived from Schwartz’s (1992) value survey and the five social axioms was small—the highest correlation was .36 between other-concern and reward for application. Consistent with the finding of their relative independence and importance for the modeling of outcomes, social axioms were shown to add predictive power beyond that provided by values in predicting egalitarianism of political attitudes. Negative religiosity predicted the broad political orientation of freedom from regulation (e.g., euthanasia, abortion, and control of fire arms) over and above its predictability provided by the two broad dimensions of value. This specific finding echoes the culture-level relationship between religiosity and a low level of domestic political rights and civil liberties. We hypothesize that beliefs add predictive power to value orientations in predicting behaviors because of their channeling eVects. A person’s beliefs indicate one’s assessment of external constraints and aVordances, thereby changing the likelihood that goals may be achieved through engaging in particular tactics or strategies. To evaluate this possibility, Bond, Leung, Au, Tong, and Chemonges-Nielsen (in press) examined a person’s vocational preferences, preferred modes of coping, and use of diVerent conflict resolution styles among college students in Hong Kong. To evaluate the unique eVects of axioms on the dependent variables, Schwartz’s (1992) values were entered in the first step of a regression analysis, and the axioms were entered in the second step. Thus, any significant relationships between the axioms and the dependent variables are independent of values. Reward for application showed a significant relationship with conventional vocational preferences (e.g., loan manager) over and above the eVects related to values. Perhaps the return on eVort is perceived as higher for this type of vocation. With regard to conflict resolution styles, reward for application predicted the conflict resolution style of accommodation after values had exercised their eVects. Religiosity was positively related to both conflict

174

KWOK LEUNG AND MICHAEL HARRIS BOND

resolution styles of accommodation and competition. The relationship with the accommodative style is consistent with the correlation between religiosity and agreeableness at both the individual and culture levels (discussed in the next paragraph). The significant relationship with the competitive style is puzzling and awaits future replication. Perhaps people high in religiosity view conflict as involving issues that are either right or wrong, and a competitive style may be seen as eVective in bringing out the moral truth of a dispute. Social cynicism was related to a lower tendency to use the collaborative and compromising styles, both of which require mutual cooperation. These relationships confirm the negative view of people by individuals high in social cynicism. Finally, social complexity predicted the conflict resolution style of compromise and of collaboration. In other words, people are more likely to compromise or collaborate if they accept the complex nature of human behaviors. For coping styles, social complexity predicted the coping style of problem solving, which may suggest that a complex view of human behaviors encourages rational approaches to personal diYculties. Fate control was significantly related to distancing, a tendency to be passive and avoid thinking about diYculties, and to the wishful thinking coping style, which involves fantasizing and daydreaming. These findings are consistent with correlations obtained at the culture level, showing that fate control is associated with withdrawing reactions to diYculties in life, such as suicide. Social axioms do relate in predictable fashion to dimensions of the fivefactor model of personality. Leung and Bond (1998) reported that the beta weights for Hong Kong Chinese in predicting social cynicism were .18 agreeableness .21 conscientiousness; for religiosity, .26 agreeableness .21 openness to experience; for reward for application, .23 conscientiousness; for fate control, .25 neuroticism; for social complexity, .40 openness to experience. A welcome convergent and discriminant validity is evident in the patterning of these relationships. However, it must be borne in mind that the specific items constituting the NEO-PI-R measure of the big five factors include some belief items, not only about the self but also about the world. Relations between the Social Axioms Survey and the NEO-PI-R may thus arise in part because of construct overlap. We have also examined the axiom dimensions at the culture level. One observation is that social axioms show more significant relationships with the big five factors at the individual than at the culture level. Second, two of the four significant correlations at the culture level are replicated at the individual level: the negative correlation between social cynicism and conscientiousness and the positive correlation between religiosity and agreeableness. Of note, many big five items are behavioral self-reports combined with motivational self-assessments and self-reported beliefs to give a

SOCIAL AXIOMS

175

comprehensive measure of personality dispositions. We can separate these constructs, as done by Keung and Bond (2002), into motivational and belief components, and then examine their joint eVects on outcome measures, such as behavioral self-reports. Leung (in preparation) has conducted two surveys in Hong Kong that examined the correlations between social cynicism and job attitudes. In the first study, 605 Chinese employees in Hong Kong were randomly interviewed by telephone. Social cynicism showed significant negative correlations with job satisfaction, job commitment, and perceived organizational justice. In the second study, based on telephone interviews of 427 Chinese employees in Hong Kong, social cynicism was again found to show a significant negative correlation with job satisfaction and perceived organizational justice, as well as organizational commitment, evaluation of superiors, and life satisfaction. As reported previously, a negative correlation was also found between social cynicism and life satisfaction and job satisfaction at the culture level. The studies reviewed above were conducted with Hong Kong Chinese. In the following, a few studies conducted in other cultural contexts are discussed. Singelis, Hubbard et al. (2003) have examined the relationships between social axioms and several individual diVerence variables and behaviors among U.S. college students. The individual diVerence variables include locus of control, interpersonal trust, social desirability, and cognitive flexibility, which refers to a flexible and creative style of problem solving, and paranormal beliefs, which include beliefs in a supreme being and other paranormal phenomena. Singelis, Hubbard, et al. found that social cynicism correlated positively with external locus of control and supernatural beliefs (e.g., black cats can bring bad luck) and negatively with social desirability, interpersonal trust, and cognitive flexibility. Religiosity showed a highly positive correlation with traditional Christian beliefs (e.g., the belief in God, heaven and hell, and soul). Reward for application correlated positively with social desirability, cognitive flexibility, and traditional Christian beliefs. Fate control correlated positively with external locus of control and negatively with cognitive flexibility. Fate control also showed a negative correlation with traditional Christian beliefs, but positive correlations with spiritual beliefs, supernatural beliefs, and belief in precognition. Finally, social complexity correlated positively with cognitive flexibility and negatively with interpersonal trust. These correlations provide good support for the convergent and discriminant validity associated with the five axiom dimensions. Convergent validity is demonstrated by significant correlations between social cynicism and interpersonal trust (negative); religiosity and traditional religious beliefs; reward for application and social desirability; fate control with external

176

KWOK LEUNG AND MICHAEL HARRIS BOND

locus of control, and social complexity with cognitive flexibility. Discriminant validity is demonstrated by lower correlations with conceptually unrelated variables. With regard to social behaviors, Singelis, Hubbard et al. (2003) reported that religiosity was related to seeking advice from a spiritual adviser, praying, reading scriptures, and church attendance. These findings corroborate the relationship between religiosity and religious activities at the culture level previously discussed. Reward for application was related to trying harder the next time when unsuccessful, and working hard to maintain good interpersonal relationships. These findings seem consistent with the correlations between reward for application and a lower tolerance for divorce and a lower emphasis on mutual attraction in mate preference at the national level. Endorsement of fate control predicted having a lucky number and reading one’s horoscope, which corroborates the earlier finding that fate control is related to supernatural and spiritual beliefs as well as belief about precognition. Social complexity was related to feeling comfortable with talking to strangers and speaking one’s mind even if it may hurt others’ feelings. This finding seems to suggest that a pluralistic approach to issues by people high in social complexity may lower their attentiveness to how others may react to their behavior, which is consistent with the negative correlation between social complexity and an other-focus performance motive at the culture level. Van Bavel, Noels, and Williams (2002) have correlated the five axiom dimensions with a variety of self-reported behaviors among a group of undergraduates in Canada. Social cynicism is related to fewer organizations volunteered for; social complexity and fate control are related to checking the horoscope; religiosity is related to attending religious services and engaging in prayer or mediation. The correlation between social complexity and checking the horoscope seems puzzling, but the uncertainty associated with a complex world view may result in the reliance on diverse strategies to seek guidance for one’s actions. Rupf and Boehnke (2002) examined the relationship in Germany between social axioms and hierarchic self-interest, which refers to the hierarchical and self-serving nature of interpersonal relations, and right-wing behavior (e.g., listening to right-wing rock music). Results showed that right-wing behavior was related positively to social cynicism and negatively to religiosity. Hierarchic self-interest was related positively to social cynicism and reward for application, and negatively to religiosity. The relationships involving social cynicism are consistent with the negative world view of people high in social cynicism. The findings with regard to religiosity are consistent with the finding that religiosity is associated with agreeableness at the

SOCIAL AXIOMS

177

cultural level. It is unclear why reward for application is related to hierarchic self-interest. People endorsing this belief may expect that everyone should take care of himself/herself, giving rise to this correlation. Ward and Ramakrishnan (2003, Study 1) studied the relationship between social axioms and kiasu, the fear of losing out, among university students in Singapore. Kiasu behaviors reflect selfishness, rudeness, greed, calculation, and competitiveness. When social desirability is controlled for, reward for application showed a significant positive relationship with kiasu. Perhaps the emphasis on eVort and planning in reward for application is associated with behaviors that reflect a fear of losing out. This finding is consistent with the earlier finding that reward for application is related to hierarchic self-interest, and that reward for application is related to an other-focus performance motive at the culture level. Safdar, Lewis, Greenglass, and Daneshpour (2003) examined the relationships between social axioms and coping strategies regarding the September 11 terrorist attacks against the United States among university students from three diVerent religious groups (Muslim, Jewish, and Christian) from three nations: Canada, the United States, and the United Kingdom. Proactive coping, which refers to active strategies such as goal setting with selfregulatory goal attainment, as well as avoidance coping, characterized by delaying and lack of eVort in problem-solving, were included. Results showed that proactive coping was related positively to reward for application and to social complexity. Avoidance coping was also positively related to social complexity, which seems consistent with the earlier finding about the tendency for people high in social complexity to read the horoscope. Perhaps the indeterminacy associated with social complexity gives rise to the use of indirect coping strategies. Furthermore, these researchers reported that participants who were practicing their religions were more likely to engage in both active and avoidant coping behaviors. Active religious participation by the three religious groups (Muslims, Christians, and Jews) also was associated with high levels of reward for application, religiosity, and social complexity. Whereas the above studies explored the correlates of social axioms, Kurman and Ronen-Eilen (2004) conducted a study that examined the functional value of social axioms. Two immigrant groups in Israel were surveyed—one from the former Soviet Union and the other from Ethiopia. These two groups were asked to respond to the Social Axioms Survey and to estimate the axiom scores of average Israelis. An Israeli group was also asked to estimate the axiom scores of average Israelis, and another random sample of Israelis was asked to respond to the Social Axioms Survey to generate the profile for average Israelis. Results showed that Israelis provided more accurate estimates of the axiom scores of average Israelis than did the

178

KWOK LEUNG AND MICHAEL HARRIS BOND

immigrants. Furthermore, better knowledge of the average axiom scores in Israel was related to better social adaptation (psychological and interpersonal adjustment) and functional adaptation (living comfortably in Israel). The only exception concerns social cynicism, which suggests that knowledge about the cynicism level in Israel was related to worse adaptation. One explanation for this interesting finding is that the immigrant groups perceived average Israelis as more socially cynical than they actually were. Schwartz’s (1992) value survey was also included in the study, and the eVects of knowledge about values and axioms on adaptation were contrasted. Results showed that axioms were generally more predictive of adaptation than values. Finally, knowledge of the profile of axiom scores for average Israelis and adoption of this profile were contrasted in terms of their impact on adaptation of the immigrants. Results showed that axiom knowledge predicted social and functional adaptation, whereas axiom adoption only predicted social adaptation. These results provide strong support for our functionalist view of social axioms.

B. CROSS-CULTURAL INVESTIGATION OF INDIVIDUAL SOCIAL AXIOMS Within-culture studies may be extended multiculturally. This strategy enables researchers to (1) discover whether the psychological construct may be used to ‘‘unpackage’’ the cultural diVerences observed in the average levels of the outcome variable observed across cultural groups (Brockner, 2003), (2) establish whether the process linkages of psychological constructs to other psychological or behavioral constructs generalize across social systems, and (3) associate cultural variables with variations in the strength of the linkage between the constructs examined, as done by Diener and Diener (1995) in showing that the strength of the correlation between selfesteem and life satisfaction was positively correlated with the individualism of that national group. This type of cross-cultural work moves the discipline toward developing and testing models of psychological process and behavior that include cultural variations in their composition. To date, one study following this approach has been completed. Fu et al. (in press) argued that observation of one’s own outcomes, as well as those of others, will lead to assessments of how eVective various strategies of influence will be when used in one’s social environment. Furthermore, the eVectiveness of a given influence strategy will be derived from the variable social aVordances provided by each cultural context. Thus, detectable linkages should exist among cultural characteristics of nations and the eVectiveness

SOCIAL AXIOMS

179

of influence strategies as rated by their culture members. Social axioms reflect considered experience on how things work in one’s social system, and therefore they can be used to unpackage observed diVerences in the rated eVectiveness of various influence strategies. For example, Fu et al. (in pess) argued that a socially cynical belief set reflects a realistic assessment that one’s social environment is responsive to and compliant in response to displays of power. Consistent with this reasoning, they found that the higher the social cynicism of persons in 12 nations, the higher the rated eVectiveness of using assertive influence tactics within each of the nations involved. Furthermore, social cynicism could be used to unpackage the average diVerences in rating the usefulness of assertive tactics of influence across the 12 nations. Relationship-based strategies add the weight of others in the proximal social field to any influence attempt, and these strategies should likewise be more eVective in social environments characterized as power-oriented. Again, Fu et al. found that the higher the social cynicism of persons in their 12 nations, the higher the rated eVectiveness of using relationship-based influence tactics, both within and across the nations involved. People high in social cynicism may see people as manipulatable and responsive to gains and losses. Thus, using an assertive strategy, which emphasizes social pressure and coercion, is likely to be perceived as eVective. Conversely, a belief set high in reward for application reflects an experience-based assessment of one’s social environment as yielding to a personal investment of knowledge, planning, and competition. In such an environment, debate, reasoning, and persuasion would be relatively more eVective than when this belief about the world is low. Consistent with this prediction, Fu et al. (in press) found that the higher the belief in reward for application of persons in their 12 nations, the higher the rated eVectiveness of using relationship-based influence tactics, both within and across the nations involved. Thus, three culture-general relationships between axioms and broad strategies of interpersonal influence explain diVerences in their rated eVectiveness both within and across cultural groups. Hypotheses about the relationship between a particular dimension of social axiom and the social outcome in question are based on reasoning tailored to the outcome in question and its relation to a physical, social, or spiritual environment reflected by its belief profile. This profile of social axioms is both individual and cultural, derived from the personal learning history of the actor holding that belief and the cultural tradition in which he or she is embedded. Such a conceptual approach may be taken with beliefs to traditional areas of social psychology, such as attributions of responsibility and in-group favoritism, or emerging areas, like moral disengagement and interpersonal justice. For

180

KWOK LEUNG AND MICHAEL HARRIS BOND

instance, it is possible that people high in reward for application may emphasize eVort in accounting for successes and failures, and that social cynicism may moderate the eVects of justice perception.

C. SECTION SUMMARY The studies reviewed previously were conducted at the individual level, including participants from several cultural groups, and involved several individual-diVerences variables and social behaviors. The overall results warrant two conclusions: first, the findings are interpretable and meaningful based on our description of these five axiom dimensions, and considerable convergent and divergent validity has been demonstrated. Second, the findings are generally coherent across studies and corroborate some of the correlations obtained at the culture level. Overall, the findings at the individual level point to satisfactory reliability and validity of these five axioms dimensions.

VIII. Conclusions and Directions for Future Research ‘‘. . . ensconsing ourselves into seeming knowledge when we should submit ourselves to an unknown fear.’’ —Shakespeare, All’s Well That Ends Well

A. THEORETICAL SIGNIFICANCE OF SOCIAL AXIOMS The notion of beliefs is not new, and before specific issues are explored, it is important to ascertain the theoretical significance and the ‘‘added value’’ of social axioms. Our axiom framework is important for at least three reasons, and the first reason is obvious and has been elaborated. Crosscultural psychology has been dominated by value frameworks, and social axioms provide an alternative perspective for understanding and interpreting cultural similarities and diVerences. Axioms may provide explanations for cultural phenomena that are diYcult to explain by values, and propel cross-cultural research into new areas, some of which have been discussed in previous sections. On a deeper level, social axioms provide a very diVerent perspective on theorizing about cultural influences than do values. Whereas values are concerned with what is important and desirable, axioms are

SOCIAL AXIOMS

181

concerned with perceived contingencies in one’s social world, which may give rise to very diVerent theoretical frameworks. For instance, individualism-collectivism has been related to a wide range of social behaviors and a coherent framework has been established (e.g., Triandis, 1995). A similar attempt can be made with regard to the dimensions of social axioms, such as social cynicism. Research that focuses on how social cynicism shapes social behaviors and what leads to a socially cynical world view may eventually lead to a cognitively oriented framework that is very diVerent from valuebased frameworks, such as individualism-collectivism. Second, our results have pointed to some very interesting possibilities with regard to the role of these social axioms in shaping behaviors. For instance, the country-level analysis has uncovered some very intriguing findings, such as substantial correlations between social cynicism and satisfaction measures, between social complexity and concern for politics, between reward for application and the deemphasis on internal characteristics in mate preference, between religiosity and agreeableness, and between fate control and suicide. Why these relationships emerge may currently prompt speculation, but future work is likely to lead to theoretical breakthroughs in these important areas. Third, we have developed pure measures of beliefs in this project. Most socalled belief scales involve a mixture of constructs, thus conflating normative and motivating influences along with the directive power of ‘‘pure’’ beliefs. We do not yet know much about the empirical consequences of this conflation, but very diVerent theories clearly are associated with normative and evaluative processes, and a more cognitive approach to measuring ‘‘beliefs’’ may lead to new discoveries.

B. THE UNIVERSALITY OF THE FIVE FACTORS OF SOCIAL AXIOMS The five dimensions of social cynicism, religiosity, reward for application, fate control, and social complexity are universal, as far as we have been able to determine. They reflect the basic human issues of whether social life will bring positive outcomes, whether spiritual beliefs and religious practices are true and useful, whether individual enterprise yields benefits, whether fate predicts future events, and whether the course of interpersonal events follows simple or complicated rules. These axioms are pitched at a general level and may have been derived from synthesizing across domain-specific exchanges with the world. Given their grounding in transactions with the environment, however, we expect that social axioms will be powerful predictors of the manner in which an individual processes daily events and deals

182

KWOK LEUNG AND MICHAEL HARRIS BOND

with his or her material, interpersonal, social and spiritual worlds. Some evidence consistent with this view has already been reviewed. Each cultural system provides prolonged training in the life skills required for accommodating to its requirements. Those requirements derive from its historical legacy, institutional superstructure, and current eco-social and geopolitical challenges. Individuals within those social systems use cognitive representations of their worlds to varying degrees, extracting structure from their exchanges with their cultured world and channeling their future exchanges in light of their goals for living. Their endorsement of social axioms thus reflects in part their cultural location. The fact that many meaningful culture-level correlations were obtained between dimensions of social axioms and societal variables provides strong support for this view. Of interest, in a recent review of the cross-cultural research on the fivefactor model of personality, Triandis and Suh (2002) conclude that its universalist claim is compromised by the lack of data on illiterate populations. Literacy is not a cornerstone of our model, and hence this consideration may not aVect its generality, which can only be confirmed by future research that explores the existence of these five axiom dimensions among illiterate respondents.

C. A UNIVERSAL MODEL OF SOCIAL AXIOMS Models are developed either inductively or deductively. In our case, the five-factor model of social axioms and its subsequent confirmation with students and adults from a diverse range of cultures is inductively derived. Given the robustness of the structure and its meaningful associations with a wide range of variables across and within cultures, there are now suYcient grounds for accepting its conceptual basis and extending its range of convenience through further research. Beliefs about the world—people, their interactions, social institutions, and nonmaterial forces—is epistemology-in-action, an elaboration of the assessments individuals have come to make about the worlds they encounter with the agendas for survival and self-extension they bring to this lifelong encounter. Social axioms are the general cognitive distillates of people’s transactions with the world, reflecting the assessed operation of its forces and constraints. We next attempt to develop a framework that argues for the fundamental nature of these five dimensions of social axioms. Our framework is functionalist in orientation and begins with the assumption that universality must be linked to basic requirements for human survival and adaptation. Developmental psychologists have long known that attachment characterizes much of infantile behavior (e.g., Bowlby, 1969),

SOCIAL AXIOMS

183

but over time young children start to engage in exploratory behaviors and optimize between a balance of attachment and exploration (e.g., Ainsworth, Blehar, Waters, & Wall, 1978). In exploring and surviving the social and physical world, evolutionary psychology argues for the need to develop competence in two broad domains: social and problem solving (Keller, 1997). In the social domain, a crucial skill for survival and adaptation is the ability to deceive others and to detect deception, skills which have been documented extensively among primates and human beings (e.g., Humphrey, 1983; Whiten & Byrne, 1997). We propose that social cynicism emerges as a response to a fundamental requirement of survival and adaptation in a social world in which deception by others is frequent and gullibility dangerous. We extend this argument to include exploitation, oppression, and other negative elements of social life that threaten the well-being of people. In short, in navigating their social and physical world, people need to judge whether it is a benign or a dog-eat-dog world and react to and cope with it accordingly. In the problem-solving domain, survival obviously requires the acquisition of knowledge and skills, from feeding to reproduction, types of adaptation that are also well documented among primates and human beings. We propose that at least three fundamental issues must be settled for eVective problem solving. First, a fundamental question is whether problems encountered are generally solvable, and what constitutes an eVective coping strategy. An extreme judgment is that all problems are intractable, leading to the well-known state of learned helplessness, in which inaction is the principal mode of response regardless of whether problems are solvable or not (Seligman, 1975). A more diVerentiated judgment involves assessing the source of the causes for the events that aVect people, an assessment that is embodied in the construct of locus of control (Rotter, 1966). We argue that fate control represents a general judgment that diYculties encountered in our social world are controlled by fate, and that there are ways to predict and alter the decree of fate in our attempts to subdue these diYculties. People high in fate control are likely to attend to signs, signals, and omens, reacting in ways that they believe will help them steer clear of negative events and happenings. People low in fate control tend to discount fate and predetermination and will not divert their attention and resources to the coping strategy of enhancing luck and altering fated events. Second, a fundamental issue informing problem-solving activities is to determine the likelihood of the eVort in leading to some success—sort of a cost-benefit analysis as illustrated in the logic behind the expectancy model (Vroom, 1964). We argue that reward for application represents our general assessment of the cost-benefit features of the life tasks frequently encountered.

184

KWOK LEUNG AND MICHAEL HARRIS BOND

If rewards are judged to fall behind eVort, loss aversion will lead to lower investment of inputs. Third, in seeking a solution to a problem, one needs to know whether a contingent approach is required for most diYculties encountered in life or if a ‘‘one-size-fits-all’’ strategy is adequate. In our terminology, a complex world calls for contingent approaches that take into account situational and individual variations, the acceptance of inconsistency in many domains, and the assumption of multiple solutions to a problem. However, in a simple world, standardized solutions are generally eVective, inconsistency is an abnormality, and most problems have a single solution. It is noteworthy that in the creativity literature, the distinction between divergent and convergent thinking is well established (e.g., Brophy, 2001; Torrance, 1974). Divergent thinking involves the generation of a wide range of ideas and alternatives, whereas in convergent thinking, the focus is to search for the best idea or alternative. Obviously, the relative eVectiveness of these two styles depends on the specific task to be solved. We argue that social complexity involves a general judgment of whether the social world is pluralistic, where ‘‘many roads lead to Rome,’’ or monolithic, where truth is singular. This judgment has significant consequences for the problemsolving styles adopted to tackle problems and diYculties encountered. Finally, one major diVerence between human beings and other animals is the need for human beings to seek meaning in their existence (e.g., Williams, 1997). One important way to define the meaning of existence is through spiritual activities, including involvement in religious practices, and this argument has a long history in social sciences. Pepitone and associates (DeRidder, Hendriks, Zani, Pepitone, & SaYotti, 1999; Pepitone & SaYotti, 1997) provided a confirmation of the salience of religious beliefs in interpreting life events by participants from the United States, The Netherlands, Italy, and India. On a practical level, religions also help provide a stable social structure and psychological peacefulness and security. We argue that religiosity emerges both as a general response to the spiritual need of human beings (Berger, 1967) and as a solution to many social problems. In fact, the content of religiosity is more about worldly than spiritual concerns, because this dimension is primarily defined by the perceived functionality of religious activities. A schematic representation of this framework is given in Figure 3.

D. ARE THESE FIVE DIMENSIONS COMPREHENSIVE? The five-factor model of social axioms and its subsequent confirmation with students and adults from a diverse range of cultures is empirically based. Derived as they were, these five dimensions of general social axioms

SOCIAL AXIOMS

Fig. 3.

185

A universal model of social axioms.

are strong contenders to represent the core etic dimensions in the human belief system. At this point, we do not and cannot pronounce that the search for universal social axioms should end with our current eVort, as future endeavors may uncover additional dimensions. However, we do claim that our current dimensions have defined a basic framework for integrating any additional dimensions that may be uncovered. Our claim is based on the fact that we made every attempt to be comprehensive and culturally representative in our sampling of axioms that people may extract from their daily exchanges with the physical, personal, social, and spiritual worlds. Decisions about the number of factors to extract from a matrix of correlations is always a subjective exercise, but we have attempted to ‘‘hear the voice of the data’’; as the items organized themselves into various numbers of factors, we chose these five on the basis of the apparent similarity of items in their meaning and the discreteness of item loadings onto the factors. We also do not deny the existence of additional emic axioms. There are two senses in which intensive within-culture examination of beliefs, like Wrightsman’s (1992), may reveal cultural emics. First, particular general

186

KWOK LEUNG AND MICHAEL HARRIS BOND

beliefs, distinctive to that cultural tradition, may join the other pan-cultural definers of a factor to lend local fullness to one of the core etics. Second, local items may combine with some of the original 60 to produce additional, new factors. This conclusion would of course have to be taken as tentative until distinctive construct validity is established for any novel factor relative to the pan-cultural five. The same caveat applies to an entirely indigenous extraction of beliefs, not using the original 60 items: whatever factors are extracted from such work would have to demonstrate discriminant validity to be regarded as having revealed emicly distinctive factors (see Cheung et al.’s, 2001, work on the Chinese Personality Assessment Inventory for an example of this process). We encourage such explorations.

E. BUILDING CULTURALLY REFLECTIVE EMICS FROM PAN-CULTURAL ETICS Given the exhaustive and culturally open procedure adopted in developing the original survey, we propose that the items identified above represent the ‘‘core etics’’ of psychological axioms—the items and the five dimensions are universal. The content of these five dimensions in any particular cultural group should include these items but extend well beyond their core composition. Thus, for Hong Kong Chinese, the original 60-item belief survey yields additional definers for each of the five axiom dimensions (Bond, Leung, Au, Tong, & Chemonges-Nielsen, in press), and the addition of these emic (culture-specific) items increases their reliability. The usefulness of emic items has also been demonstrated by Singelis, Her et al. (2003) in a study involving diVerent ethnic groups within the United States. We expect similar enlargements to the range of convenience for these five axiom constructs in any and all cultural groups. Researchers from a given cultural tradition can use the items defining the core as a basis for adding new, culture-specific belief items to the 60-item survey. Such culture-focused explorations will allow for a richer, more comprehensive tapping of the five belief constructs as they manifest themselves in any particular cultural niche. In this way, these core etics may be used as a springboard into a fuller pool of culturally sensitive and resonant reflections of these universal constructs. A related issue is that only three of the five axiom dimensions show acceptable reliability in the majority of the cultural groups studied. The average Cronbach alphas across the 40 student and 13 adult data sets are as follows: social cynicism (student: .67; adult: .68), reward for application (student: .61; adult: .59), religiosity (student: .71; adult: .75), fate control (student: .53; adult: .55), and social complexity (student: .45; adult: .41). For within-culture research, the addition of emic items will improve the

SOCIAL AXIOMS

187

reliability of axiom dimensions that are low in reliability. For cross-cultural research, we defer to the current etic items to maintain cross-cultural comparability until we develop better alternatives. Perhaps the use of latent variables may provide one way to compensate for the lower reliability of some of the dimensions, because correction for unreliability is made when latent variables are being used.

F. DIRECTIONS FOR FUTURE RESEARCH This program of research has indicated at least five directions for future research. Given the recent emergence of this research, our discussion is concerned with general issues rather than specific research hypotheses. First, research should continue to explore the antecedents and consequences of these five social axioms within and across cultures. Our review has suggested a number of leads in this direction. For instance, it is interesting to examine how social cynicism plays out in the workplace. Previous research has shown that specific cynicism tied to an organization event such as downsizing can generate a wide range of negative attitudinal and behavioral reactions from employees (e.g., Andersson, 1996). It would be interesting to know how social cynicism, a general belief that is not event specific, is related to job attitudes and work behaviors. Other interesting areas of inquiry include the role of social axioms in the broad field of culture and social cognition (e.g., Kashima, 2001) and the relationship between social axioms and elementary forms of social behavior (Fiske, 1992). A related area for future research is to explore the relationships between social axioms and specific beliefs. We conceive of social axioms as higherorder constructs of belief under which subordinate belief domains are embedded. As an example of this logic, domain-specific beliefs about the causes and cures of psychological problems (Luk & Bond, 1992) may be related to more general axioms, particularly reward for application and social complexity. Neurotic clinical syndromes such as depression and paranoia, along with acting-out disorders, should likewise have distinctive belief profiles across the five dimensions. At the culture level, we have also uncovered meaningful and important societal correlates of the five axiom dimensions. Although most social psychological research is concerned with individual-level processes, a macro perspective on social psychological processes may have significant social relevance and implications for social policies and changes. The relationship between fate control and suicide rate and heart disease may serve as an example. What are the psychological processes underlying these two correlations? One possibility is to adopt what Leung and Bond (1989) term a strong

188

KWOK LEUNG AND MICHAEL HARRIS BOND

etic approach, which regards individual-level and culture-level processes as similar. Perhaps people high in fate control are more prone to suicidal thoughts and more susceptible to traps of unhealthy lifestyles that culminate in heart diseases. As a result, societies with more individuals who endorse fate control are higher in suicide rate and heart diseases. Alternately, we can assume that no relationship exists between individual-level and culture-level processes and explore whether independent causal mechanisms are needed to explain processes and relationships at these two diVerent levels. Obviously, the strong etic approach is simpler, and we have adopted it implicitly as our starting point in interpreting the findings reported in earlier sections, but we are open to the possibility that for some variables, diVerent processes may be involved at the two levels. In any event, answers to these issues are immensely important of improving the conditions of human existence. Second, two axiom dimensions, fate control and social complexity, show low internal reliabilities, a finding that is not ideal for within-culture research. Future research should identify new items that can improve the reliability of these two dimensions. Emic items can be developed for use in a particular cultural context and their cross-cultural generality assessed in subsequent cross-cultural research. A long-term goal is to develop better and more etic items to capture these axiom dimensions more reliably. Third, we have proposed a framework to account for the universality of the five axiom dimensions. In reviewing Schwartz’s (1992) universal theory of values, Lonner and Adamopolous (1997) noted a problem of indeterminacy in the theory when basic needs for human survival and functioning are linked to the 10 value types, because ‘‘the explanation of the emergence of structure is exogenous to the core of the theory’’ (p. 74). Other psychological models may provide an explanation of these value types, and our framework also suVers from the same problem of indeterminacy. However, theories that explain the origin of value and axiom structures are ‘‘metatheories,’’ and like evolution theory and the big bang theory, their validity will take decades to establish, perhaps constituting a long-term goal for the entire field. Fourth, considerable variation exists along dimensions of belief within a given cultural group. This variation is responsive to the range of personal possibilities oVered in complex cultural systems. Individuals must be socialized to undertake a range of roles and occupations, each with its attendant norms, expectations, values, and beliefs. We need to know more about the embeddedness of beliefs within social systems, as we already know about values, for example (e.g., Kohn, Slomczynski, & Schoenbach, 1986), and how these beliefs are socialized in the family, the schools, the playing fields, and the workplace. Finally, a primary motivation for our development of these social axioms has been to complement the frameworks based on value, such as those of

SOCIAL AXIOMS

189

Schwartz (1992), which have dominated cross-cultural psychology in the last decades. Now that the universality of the five social axioms is supported, future work should begin to examine how these two frameworks can be integrated for a better understanding of social behaviors and cultural similarities and diVerences. At a more concrete level, the expectancy model (Vroom, 1964) provides a well-known integration of valence and expectancy into a single framework. Given that values can be regarded as generalized valences, and social axioms as generalized expectancies, it is interesting to see if values and axioms can be integrated in a similar fashion as valences and expectancies. Another intriguing possibility is that the way these two types of construct ought to be integrated may depend on the specific social behavior involved. In our view, this type of research question is crucial to the development of cross-cultural psychology and will elevate cross-cultural theorizing to a new level. ‘‘There are truths on this side of the Pyrenees which are falsehoods on the other.’’ —Pascal, Pensees, 60 (294), G. Hofstede, trans.

Acknowledgments We thank Joel Brockner, Shalom Schwartz, Harry Triandis, Evert Van de Vliert, and the editor for their insightful comments on an earlier draft. Supported in part by a research grant provided by Research Grants Council of Hong Kong. This global project grew out of a project supported by a grant provided by the Research Grants Council of Hong Kong. We thank the following collaborators for collecting the data from the countries reported in this chapter: Markus Broer (Argentina); Filip Boen and Sophie M. Lambert (Belgium); Maria Cristina Ferreira (Brazil); Kimberly A. Noels and Jay van Bavel (Canada); Jianxin Zhang and Lina Chen (China); Iva Solcova and Iva Stetovska (Czech Republic); Toomas Niit and Kaisa-Kitri Niit (Estonia); Helena Hurme and Mia Bo¨ling (Finland); Vije´ Franchi (France); Guguli Magradze and Nino Javakhishvili (Georgia); Klaus Boehnke, Gu¨nter Bierbrauer, and Edgar Klinger (Germany); Penny Panagiotopoulou (Greece); Marta Fulop and Mihaly Berkics (Hungary); Nandita Chaudhary, Sujata Sriram, Anjali Ghosh, and Neharika Vohra (India); Dien Fakhri Iqbal (Indonesia); Saba Safdar (Iran and Canada); Jenny Kurman and Ram David Thein (Israel); Anna Laura Comunian (Italy); Fumio Murakami and Susumu Yamaguchi (Japan); Son Kyung Ae (Korea); Ivars Austers (Latvia); Charles Harb (Lebanon); Joseph O. T. Odusanya, Zainal A. Ahmed, and Rosnah Ismail (Malaysia); Fons van de Vijver and Xu Huang (Netherlands); Colleen Ward (New Zealand and Singapore); Andrew A. Mogaji (Nigeria); David Lackland Sam (Norway); Muhammad Jahan Zeb Khan (Pakistan); William E. Cabanillas (Peru); Ly SyCip (Philippines); Fe´lix Neto, Rosa Cabecinhas, and Paulo Xavier (Portugal); Margareta Dinca (Romania); Nadezhda Lebedeva, Andrianna Viskochil, and Oksana Ponomareva (Russia); Luis Oceja and Silvia Campo (Spain); KwangKuo Hwang (Taiwan); June Bernadette D’Souza (Thailand); Bilge Ataca (Turkey); Adrian

190

KWOK LEUNG AND MICHAEL HARRIS BOND

Furnham and J. Rees Lewis (United Kingdom); Theodore M. Singelis (United States); Sharon Reimel de Carrasquel (Venezuela).

Appendix References of the Variable Sources Cited in Tables VI and VII Central Intelligence Agency (2002). The world factbook 2002. Retrieved April 1, 2003, from http://www.odci.gov/cia/publications/factbook/index.html Den Hartog, D. N., House, R. J., Hanges, P. J., & Ruiz-Quintanilla, S. A. (1999). Cultural specific and cross-culturally generalizable implicit leadership theories: Are attributes of charismatic/transformational leadership universally endorsed? Leadership Quarterly, 10, 219–256. Diener, E., Gohm, C. L., Suh, E., & Oishi, S. (2000). Similarity of the relations between martial status and subjective well-being cross cultures. Journal of Cross-Cultural Psychology, 31, 419–436. Diener, E., & Suh, E. M. (1999). National diVerences in subjective well-being. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology (pp. 434–450). New York: Russell Sage Foundation. Freedom House (2002). Freedom in the world country ratings 1972–73 to 2001–2002 [data file]. Available from Freedom House Web site, http://www.freedomhouse.org/research/freeworld/FHSCORES.xls Humana, C. (1992). World human rights guide. (3rd ed.) New York: Oxford University Press. Inglehart, R., Basan˜ez, M., & Moreno, A. (1998). Human values and beliefs: A cross-cultural sourcebook. Ann Arbor, MI: The University of Michingan Press. International Labour Organization (2002). LABORSTA, an International Labour OYce database on labour statistics operated by the ILO Bureau of Statistics [data file]. Available from Laborsta Web site,http://laborsta.ilo.org. Levine, R. V., & Norenzayan, A. (1999). The pace of life in 31 countries. Journal of CrossCultural Psychology, 30, 178–205. Lynn, R. (1991). The secret of the miracle economy: DiVerent national attitudes to competitive and money. Exeter, UK: Social AVairs Unit. McCrae, R. R. (2002). NEO-PI-R data from 36 cultures: Further intercultural comparisons. In R. R. McCrae & J. Allik (Eds.), The five-factor model of personality across cultures (pp. 105–125). New York: Kluwer Academic/Plenum Publishers. Population Crisis Committee (1988). Country rankings of the status of women: Poor, powerless and pregnant. Population Briefing Paper, No. 20, June. Washington, DC: Author. Shackelford, T. K., & Schmitt, D. P. (2002). Universal dimensions of human mate preferences. Manuscript submitted for publication. Smith, P. B., Dugan, S., Peterson, M. F., & Leung, K. (1998). Individualism: Collectivism and the handling of disagreement. A 23 country study. International Journal of Intercultural Relations, 22, 351–367. Smith, P. B., Peterson, M. F., & Schwartz, S. H. (2002). Cultural values, sources of guidance, and their relevance to managerial behavior: A 47-nation study. Journal of Cross-Cultural Psychology, 33, 188–208. Swanbrow, D. (1997, December 10). Study of worldwide rates of religiosity, church attendance. News release, The University of Michigan. Retrieved, March 3, 2003, from http:// www.umich.edu/newsinfo/Releases/1997/Dec97/r121097a.html

SOCIAL AXIOMS

191

United Nations (2002). Social indicators [data file]. Available from United Nation Statistics Division Web site, http://unstats.un.org/unsd/demographic/social/default.htm United Nations Development Programme (2000). Human development report 2000. New York: Oxford University Press. United Nations Development Programme (2001). Human development report 2001. New York: Oxford University Press. United Nations Development Programme (2002). Human development report 2002. New York: Oxford University Press. Van de Vliert, E., & Janssen, O. (2002). ‘‘Better than’’ performance motives as roots of satisfaction across more and less developed countries. Journal of Cross-Cultural Psychology, 33, 380–397. Van de Vliert, E., Schwartz, S. H., Huismans, S. E., Hofstede, G., & Daan, S. (1999). Temperature, cultural masculinity, and domestic political violence: A cross-national study. Journal of Cross-Cultural Psychology, 30, 291–314. World Economic Forum (2002). Environmental Sustainability Index 2002 Rankings. Retrieved July 20, 2003, from http://www.ciesin.columbia.edu/indicators/ESI/rank.html World Health Organization (2002). World health statistics annual (online edition) [data file]. Available from World Health Organization Web site, http://www3.who.int/whosis/ menu.cfm

References Abelson, R. P. (1981). Psychological status of the script concept. American Psychologist, 36, 715–729. Abelson, R. P. (1988). Conviction. American Psychologist, 43, 267–275. Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A psychological study of the strange situation. Hillsdale, NJ: Lawrence Erlbaum. Allik, J., & McCrae, R. R. (2004). Towards a geography of personality traits: Patterns of profiles across 36 cultures. Journal of Cross-Cultural Psychology, 35, 13–28. Andersson, L. M. (1996). Employee cynicism: An examination using a contract violation framework. Human Relations, 49, 1395–1418. Applegate, J. L., Kline, S. L., & Delia, J. G. (1991). Alternative measures of cognitive complexity as predictors of communication performance. International Journal of Construct Psychology, 4, 193–213. Argyle, M. (2000). Psychology and religion: An introduction. London: Routledge. Bandura, A. (1982). The psychology of chance encounters and life paths. American Psychologist, 37, 747–755. Bandura, A. (2001). Self-eYcacy: The exercise of control. New York: Freeman. Bandura, A. (2002). Social cognitive theory in cultural context. Applied Psychology: An International Review, 51, 269–290. Bar-Tal, D. (1990). Group beliefs: A conception for analyzing group structure, processes, and behavior. New York: Springer-Verlag. Bar-Tal, D. (2000). Shared beliefs in a society. Thousand Oaks, CA: Sage. Becker, G. (1996). The meta-analysis of factor analyses: An illustration based on the cumulation of correlation matrices. Psychological Methods, 1, 341–353. Bem, D. J. (1970). Beliefs, attitudes, and human aVairs. Belmont, CA: Brooks/Cole.

192

KWOK LEUNG AND MICHAEL HARRIS BOND

Berger, P. L. (1967). The sacred canopy: Elements of a sociological theory of religion. New York: Doubleday. Berger, P. L., & Luckmann, T. (1966). The social construction of reality: A treatise in the sociology of knowledge. New York: Doubleday. Bond, M. H. (1988). Finding universal dimensions of individual variation in multi-cultural studies of value. Journal of Personality and Social Psychology, 55, 1009–1015. Bond, M. H. (1998). Social psychology across cultures: Two ways forward. In J. G. Adair, D. Belanger, & K. L. Dion (Eds.) Advances in psychological science: Social, personal, and cultural aspects (Vol. 1, pp. 137–150). Sussex, UK: Psychology Press Ltd. Bond, M. H., Leung, K., Au, A., Tong, K. K., & Chemonges-Nielsen, Z. (in press). Combining social axioms with values in predicting social behaviors. European Journal of Personality Bond, M. H., Leung, K., Au, A., Tong, K. K., Reimel de Carrasquel, S., Murakami, F., et al. (in press). Culture-level dimensions of social axioms and their correlates across 41 cultures. Journal of Cross-Cultural Psychology Bond, M. H., Leung, K., & Schwartz, S. (1992). Explaining choices in procedural and distributive justice across cultures. International Journal of Psychology, 27, 211–225. Bond, M. H., & Tornatsky, L. C. (1973). Locus of control in students from Japan and the United States: Dimensions and levels of response. Psychologia, 16, 209–213. Bosland, N. (1985). The cross-cultural equivalence of the power distance-, uncertainty avoidance-, individualism-, and masculinity-measurement scales. The Netherlands: Working paper, Institute for Research on Intercultural Cooperation. Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. New York: Basic Books. Brewer, M. B., & Harasty, A. S. (1996). Seeing groups as entities: The role of perceiver motivation. In R. M. Sorrentino & E. T. Higgins (Eds.), Handbook of motivation and cognition (Vol. 3, pp. 347–370). New York: Guilford. Brockner, J. (2003). Unpacking country eVects: On the need to operationalize the psychological determinants of cross-national diVerences. In R. M. Kramer & B. M. Staw (Eds.), Research in organizational behavior (Vol. 25, pp. 333–367). Oxford, UK: Elsevier Brophy, D. R. (2001). Comparing the attributes, activities, and performance of divergent, convergent, and combination thinkers. Creativity Research Journal, 13, 439–455. Cacioppo, J. T., Petty, R. E., Feinstein, J. A., & Jarvis, W. B. G. (1996). Dispositional diVerences in cognitive motivation: The life and times of individuals varying in need for cognition. Psychological Bulletin, 119, 197–253. Chan, W., Ho, R. M., Leung, K., Chan, D. K. S., & Yung, Y. F. (1999). An alternative method for evaluating congruence coeYcients with Procrustes rotation: A bootstrap procedure. Psychological Methods, 4, 378–402. Chang, W. C., Wong, W. K., & Koh, J. B. K. (2003). Chinese values in Singapore: Traditional and model. Asian Journal of Social Psychology, 6, 5–29. Cheung, F. M., & Leung, K. (1998). Indigenous personality measures: Chinese examples. Journal of Cross-Cultural Psychology, 29, 233–248. Cheung, F. M., Leung, K., Zhang, J. X., Sun, H. F., Gan, Y. Q., Song, W. Z., & Xie, D. (2001). Indigenous Chinese personality constructs: Is the five factor model complete? Journal of Cross-Cultural Psychology, 32, 407–433. Christie, R., & Geis, F. L. (Eds.) (1970). Studies in Machiavellianism. New York: Academic Press. Costa, P. T., Jr., & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO PI-R) and NEO Five-Factor Inventory (NEO-FFI). Odessa, FL: Psychological Assessment Resources. Dake, K. (1992). Myths of nature: Culture and the social construction of risk. Journal of Social Issues, 48, 21–37.

SOCIAL AXIOMS

193

Dasgupta, N., Banaji, M. R., & Abelson, R. P. (1999). Group entitativity and group perception: Association between physical features and psychological judgment. Journal of Personality and Social Psychology, 77, 991–1003. DeRidder, R., Hendriks, E., Zani, B., Pepitone, A., & SaYotti, L. (1999). Additional crosscultural evidence on the selective usage of nonmaterial beliefs in explaining life events. European Journal of Social Psychology, 29, 435–442. Diener, E., & Diener, M. (1995). Cross-cultural correlates of life satisfaction and self-esteem. Journal of Personality and Social Psychology, 68, 653–663. Durkheim, E. (1984). The division of labor in society (W. D. Halls, Trans.). London: MacMillan (Original work published 1893). Duveen, G. (2001). The power of ideas. In G. Duveen (Ed.), Social representations: Explorations in social psychology (pp. 1–17). Washington Square, NY: New York University Press. Echter, T., Kim, U., Kau, C. J., Li, H. C., Simmons, C., & Ward, C. (1998). A comparative study in the levels of human values: People’s Republic of China, Singapore, Taiwan, and the United States. Asian Journal of Social Psychology, 1, 271–288. Feather, N. T. (1982). Expectations and actions: Expectancy-value models in psychology. Hillsdale, NJ: Erlbaum. Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press. Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley. Fiske, A. P. (1992). The four elementary forms of sociability: Framework for a unified theory of social relations. Psychological Review, 99, 689–723. Fletcher, G., Danilovics, P., Fernandez, G., Peterson, D., & Reeder, G. D. (1986). Attributional complexity: An individual diVerences measure. Journal of Personality and Social Psychology, 51, 875–884. Foa, U. G. (1971). Interpersonal and economic resources. Science, 171, 345–351. Fu, P. P., Kennedy, J., Yukl, G., Bond, M. H., Peng, T. K., Srinivas, E. S., et al. (in press). Exploring the eVect of cultural values on the relationship between social beliefs and managerial influence strategies in twelve cultures: A meso approach using HLM. Journal of International Business Studies. Fukuyama, F. (1995). Trust: The social virtues and the creation of prosperity. New York: Free Press. Furnham, A. (1988). Lay theories. London: Pergamon Press. Furnham, A., Bond, M. H., Heaven, P., Hilton, D., Lobel, T., Masters, J., et al. (1993). A comparison of Protestant work ethic beliefs in thirteen nations. Journal of Social Psychology, 133, 185–197. Garfield, J. L. (1988). Belief in psychology: A study in the ontology of mind. Cambridge, MA: MIT Press. Georgas, J., Van de Vijver, F. J. R., & Berry, J. (2004). The ecocultural framework, ecosocial indices and psychological variables in cross-cultural research. Journal of Cross-Cultural Psychology, 35, 74–96. Gorsuch, R. L. (1983). Factor analysis. Hillsdale, NJ: Lawrence Erlbaum. Hamilton, D. L., & Sherman, S. J. (1996). Perceiving persons and groups. Psychological Review, 103, 336–355. Harkness, S., & Super, C. (Eds.) (1996). Parents’ cultural belief systems: Their origins, expressions, and consequences. New York: Guilford Press. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley. Heine, S. J., Lehman, D. R., Peng, K., & Greenholtz, J. (2002). What’s wrong with crosscultural comparisons of subjective Likert scales? The reference group eVect. Journal of Personality and Social Psychology, 82, 903–918.

194

KWOK LEUNG AND MICHAEL HARRIS BOND

Hofstede, G. (1980). Culture’s consequences: International diVerences in work-related values. Beverly Hills, CA: Sage. Humphrey, N. K. (1983). The adaptiveness of mentalism. Behavioral and Brain Sciences, 6, 343–390. Inglehart, R. (1997). Modernization and post-modernization: Cultural, economic, and political change in 43 societies. Princeton, NJ: Princeton University Press. Ip, G. W. M., & Bond, M. H. (1995). Culture, values, and the spontaneous self-concept. Asian Journal of Psychology, 1, 30–36. Jahoda, G. (1988). Critical notes and reflections on ‘‘social representations.’’ European Journal of Social Psychology, 18, 195–209. Kashima, Y. (2001). Culture and social cognition: Toward a social psychology of cultural dynamics. In D. Matsumoto (Ed.), The handbook of culture and psychology (pp. 325–360). New York: Oxford University Press. Katz, D. (1960). The functional approach to the study of attitudes. Public Opinion Quarterly, 24, 163–204. Keller, H. (1997). Evolutional approaches. In J. W. Berry, Y. H. Poortinger, & J. Pandey (Eds.) Handbook of cross-cultural psychology (Vol. 1, pp. 215–255). Boston: Allyn and Bacon. Kelly, G. (1963). A theory of personal constructs. New York: W. W. Norton. Keung, D. K. Y., & Bond, M. H. (2002). Dimensions of political attitudes and their relations with beliefs and values in Hong Kong. Journal of Psychology in Chinese Societies, 3, 133–154. Kluckhohn, F. R., & Strodtbeck, F. L. (1961). Variations in value orientations. Evanston, IL: Row, Peterson. Kohn, M. L., Slomczynski, K. M., & Schoenbach, C. (1986). Social stratification and the transmission of values in the family. Sociological Forum, 1, 73–102. Kruglanski, A. W. (1989). Lay epistemics and human knowledge: Cognitive and motivational basis. New York: Plenum. Kurman, J., & Ronen-Eilen, C. (2004). Lack of knowledge of a culture’s social axioms and adaptation diYculties among immigrants. Journal of Cross-Cultural Psychology, 35, 192–208. La Piere, R. I. (1959). The Freudian ethic. New York: Duell, Sloan, and Pierce. Landes, D. S. (1999). The wealth and poverty of nations. New York: Norton. Lau, S., & Leung, K. (1992). Relations with parents and school and Chinese adolescents’ self-concept, delinquency, and academic performance. British Journal of Educational Psychology, 62, 193–202. Lazarus, R., & Folkman, S. (1984). Personal control and stress and coping processes: A theoretical analysis. Journal of Personality and Social Psychology, 46, 839–852. Lerner, M. (1980). The belief in a just world: A fundamental delusion. New York: Plenum. Leung, K. (in preparation). Social cynicism and job attitudes and perception of superiors. City University of Hong Kong Leung, K. (1987). Some determinants of reactions to procedural models of conflict resolution: A cross-national study. Journal of Personality and Social Psychology, 53, 898–908. Leung, K. (1989). Cross-cultural diVerences: Individual-level vs. culture-level analysis. International Journal of Psychology, 24, 703–719. Leung, K., & Bond, M. H. (1989). On the empirical identification of dimensions for crosscultural comparisons. Journal of Cross-Cultural Psychology, 20, 133–151. Leung, K., & Bond, M. H. (1998). Cultural beliefs about conflict and peace. Paper presented at the 24th International Congress of Applied Psychology, San Francisco, CA, Aug. 4–9. Leung, K., Bond, M. H., Reimel de Carrasquel, S., Mun˜oz, C., Herna´ndez, M., Murakami, F., et al. (2002). Social axioms: The search for universal dimensions of general beliefs about how the world functions. Journal of Cross-Cultural Psychology, 33, 286–302.

SOCIAL AXIOMS

195

Leung, K., Bond, M. H., & Schwartz, S. H. (1995). How to explain cross-cultural diVerences: Values, valences, and expectancies? Asian Journal of Psychology, 1, 70–75. Levine, R. V., & Norenzayan, A. (1999). The pace of life in 31 countries. Journal of CrossCultural Psychology, 30, 178–205. Lickel, B., Hamilton, D. L., Wieczorkowska, G., Lewis, A., Sherman, S. J., & Uhles, A. N. (2000). Varieties of groups and the perception of group entitativity. Journal of Personality and Social Psychology, 78, 223–246. Lonner, W. J., & Adamopoulos, J. (1997). Culture as antecedent to behavior. In J. W. Berry, Y. H. Poortinger, & J. Pandey (Eds.) Handbook of cross-cultural psychology (Vol. 1, pp. 43–83). Boston: Allyn and Bacon. Luk, C. L., & Bond, M. H. (1992). Chinese lay beliefs about the causes and cures of psychological problems. Journal of Clinical and Social Psychology, 11, 140–157. Markus, H. R., & Plaut, V. C. (2001). Social representations: Catching a good idea. In K. Deaux & G. Philoge`ne (Eds.), Representations of the social (pp. 183–189). Oxford, UK: Blackwell. Martin, M. M., & Anderson, C. M. (1998). The cognitive flexibility scale: Three validity studies. Communication Reports, 11, 1–9. McClosky, H., & Zaller, J. (1984). The American ethos: Public attitudes toward capitalism and democracy. Cambridge, MA: Harvard University Press. McCrae, R. R. (2002). NEO-PI-R data from 36 cultures: Further intercultural comparisons. In A. J. Marsella, R. R. McCrae, & J. Allik (Eds.), The five-factor model across cultures (pp. 105–125). Amsterdam: Kluwer Academic Publishers. McCrae, R. R., Zonderman, A. B., Costa, P. T., Jr., Bond, M. H., & Paunonen, S. V. (1996). Evaluating replicability of factors in the Revised NEO Personality Inventory. Journal of Personality and Social Psychology, 70, 552–566. Menon, T., Morris, M. W., Chiu, C. Y., & Hong, Y. Y. (1999). Culture and the construal of agency: Attribution to individual versus group dispositions. Journal of Personality and Social Psychology, 76, 701–717. Miller, D. (1991). Handbook of research design and social measurement. Newbury Park, CA: Sage. Morris, M. W., & Peng, K. (1994). Culture and cause: American and Chinese attributions of social and physical events. Journal of Personality and Social Psychology, 67, 949–971. Morris, M. W., Menon, T., & Ames, D. R. (2001). Culturally conferred conceptions of agency: A key to social perception of persons, groups, and other actors. Personality and Social Psychology Review, 5, 169–182. Moscovici, S. (2001). Why a theory of social representations? In K. Deaux & G. Philpgene (Eds.), Representations of the social: Bridging theoretical traditions (pp. 8–35). Oxford, UK: Blackwell Publishers. Nowicki, S., & Duke, M. P. (1983). The Nowicki-Strickland life span locus of control scales: Construct validation. In H. M. Lefcourt (Ed.) Research with the locus of control construct (Vol. 2, pp. 9–43). New York: Academic Press. Oltmanns, T. F., & Maher, B. A. (Eds.) (1988). Delusional beliefs. New York: Wiley. Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning. Urbana, IL: University of Illinois Press. Pepitone, A., & SaYotti, L. (1997). The selectivity of nonmaterial beliefs in interpreting life events. European Journal of Social Psychology, 27, 23–35. Phares, J. E. (1965). Internal-external locus of control as a determinant of amount of social influence. Journal of Personality and Social Psychology, 2, 642–647. Robinson, J. P., Shaver, P. R., & Wrightsman, L. S. (1991). Measures of personality and social psychological attitudes. San Diego, CA: Academic Press. Rokeach, M. (1973). The nature of human values. Glencoe, IL: Free Press.

196

KWOK LEUNG AND MICHAEL HARRIS BOND

Roseman, I. J. (1994). The psychology of strongly held beliefs: Theories of ideological structure and individual attachment. In R. C. Schank & E. Langer (Eds.), Beliefs, reasoning, and decision making: Psych-Logic in honor of Bob Abelson (pp. 175–208). Hillsdale, NJ: Lawrence Erlbaum. Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs, 80, 1–28. Rumelhart, D. E. (1984). Schemata and the cognitive system. In R. S. Wyer & T. K. Srull (Eds.) Handbook of social cognition (Vol. 1, pp. 161–188). Hillsdale, NJ: Lawrence Erlbaum. Rupf, M., & Boehnke, K. (2002). Hierarchic self-interest and political delinquency: Do social axioms serve as moderators? Poster presented in the regional conference of International Association for Cross-Cultural Psychology, Winchester, England, July 7–11. Safdar, S., Lewis, J. R., Greenglass, E., & Daneshpour, M. (2003, July). The relationship between religious aYliation, social beliefs, and coping. Paper presented at the VIII European Congress of Psychology, Vienna, Austria. Sagiv, L., & Schwartz, S. H. (1995). Value priorities and readiness for out-group social contact. Journal of Personality and Social Psychology, 69, 437–448. Schwartz, S. H. (1992). The universal content and structure of values: Theoretical advances and empirical tests in 20 countries. In M. Zanna (Ed.) Advances in Experimental Social Psychology (Vol. 25, pp. 1–65). New York: Academic Press. Seeman, M. (1997). The neglected, elusive situation in social psychology. Social Psychology Quarterly, 60, 4–13. Seligman, M. E. P. (1975). Helplessness. San Francisco: Freeman. Simonton, D. K. (1976). The sociopolitical context of philosophical beliefs: A transhistorical causal analysis. Social Forces, 54, 513–523. Singelis, T. M., Her, P., Aaker, J., Bhawuk, D. P. S., Gabrenya, W., Gelfand, M., Harwood, J., Tanaka-Matsumi, J., & Vandello, J. (2003). Ethnic and regional diVerences in social axioms Manuscript submitted for publication. Singelis, T. M., Hubbard, C., Her, P., & An, S. (2003). Convergent validation of the Social Axioms Survey. Personality and Individual DiVerences, 34, 269–282. Smith, P. B., & Bond, M. H. (1998). Social psychology across cultures (2nd ed.). London: Prentice Hall International (also 1999, Needham Heights, MA: Allyn and Bacon). Smith, P. B., & Bond, M. H. (2004). Honoring culture scientifically when doing social psychology. In M. A. Hogg & J. Cooper (Eds.), Sage handbook of social psychology (pp. 43–61). London: Sage. Snyder, M. (1984). When beliefs create reality. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 18, pp. 247–305). Orland, FL: Academic Press. Solomon, S., Greenberg, J., & Pyszczinski, T. (1991). A terror management theory of social behavior: The psychological functions of self-esteem and cultural worldviews. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 24, pp. 93–159). New York: Academic Press. Spector, P. E. (1982). Behavior in organizations as a function of employee’s locus of control. Psychological Bulletin, 91, 482–497. Spini, D. (2003). Measurement equivalence of 10 value types from the Schwartz Value Survey across 21 countries. Journal of Cross-Cultural Psychology, 34, 3–23. Stewart, B., Hetherington, G., & Smith, M. (1984). Survey item banks: Vol. 12. West Yorkshire, England: MCB University Press. Sullivan, J. L., & Transue, J. E. (1999). The psychological underpinnings of democracy: A selective review of research on political tolerance, interpersonal trust, and social capital. Annual Review of Psychology, 50, 625–650.

SOCIAL AXIOMS

197

Super, C. M., & Harkness, S. (1986). The developmental niche: A conceptualization at the interface of the society and the individual. International Journal of Behavioral Development, 9, 545–570. Torrance, E. P. (1974). Torrance tests of creative thinking. Lexington, MA: Personal Press. Triandis, H. C. (1964). Cultural influences upon cognitive processes. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 1, pp. 1–48). New York: Academic Press. Triandis, H. C. (1973). Work and non-work: Intercultural perspectives. In M. D. Dunnette (Ed.), Work and non-work in the year 2001 (pp. 29–52). Monterey, CA: Brooks/Cole. Triandis, H. C. (1995). Individualism and collectivism. Boulder, CO: Westview Press. Triandis, H. C., & Suh, E. M. (2002). Cultural influence on personality. Annual Review of Psychology, 53, 133–160. Van Bavel, J., Noels, K. A., & Williams, R. (2002). Examining the predictive validity of social axioms on behavioral outcomes: A cross-cultural perspective. Poster Session, 63rd Annual Canadian Psychological Association Convention, Vancouver, BC. Van de Vijver, F., & Leung, K. (1997). Methods and data analysis for cross-cultural research. Newbury Park, CA: Sage. Van de Vliert, E. (2003). Thermoclimate, culture, and poverty as country-level roots of workers’ wages. Journal of International Business Studies, 34, 40–52. Voegelin, E. (1956). Order and history: Vol. 1. Israel and revelation. Baton Rouge, LA: Louisiana State University Press. Vroom, V. H. (1964). Work and motivation. New York: Wiley. Wagner, W., & Kronberger, N. (2001). Killer tomatoes: Collective symbolic coping with biotechnology. In K. Deaux & G. Philoge`ne (Eds.), Representations of the social (pp. 147–164). Oxford, UK: Blackwell. Ward, C., & Ramakrishnan, M. (2003). Fear of losing out? The correlates, causes and cultural context of kiasu. Manuscript submitted for publication. New Zealand: Victoria University of Wellington. Whiten, A., & Byrne, R. W. (Eds.) (1997). Machiavellian intelligence II: Extensions and evaluations. Cambridge, England: Cambridge University Press. Wicker, A. W. (1992). Making sense of environments. In W. B. Walsh, K. H. Craik, & R. H. Price (Eds.), Person-environment psychology: Models and perspectives (pp. 157–192). Hillsdale, NJ: Lawrence Erlbaum. Williams, K. D. (1997). Social ostracism. In R. M. Kowalski (Ed.), Aversive interpersonal behaviors (pp. 133–170). New York: Plenum Press. Wong, S. W., & Bond, M. H. (2003). Socializing emotionality: A study of emotional length, recency and intensity across the citizens of 30 nations. Manuscript submitted for publication. Wrightsman, L. S. (1992). Assumptions about human nature. Newbury Park, CA: Sage. Wrightsman, L. S., Richard, W. C., & Noble, F. C. (1966). Attitudes and attitude change in guidance institute participants. Counselor and Educator Supervision, 5, 212–220. Yang, K. S. (1988). Will societal modernization eventually eliminate cross-cultural psychological diVerences? In M. H. Bond (Ed.), The cross-cultural challenge to social psychology (pp. 67–85). Newbury Park, CA: Sage. Yik, M. S. M., & Bond, M. H. (1993). Exploring the dimensions of Chinese person perception with indigenous and imported constructs: Creating a culturally balanced scale. International Journal of Psychology, 28, 75–95.

This Page Intentionally Left Blank

VIOLENT VIDEO GAMES: SPECIFIC EFFECTS OF VIOLENT CONTENT ON AGGRESSIVE THOUGHTS AND BEHAVIOR

Craig A. Anderson Nicholas L. Carnagey Mindy Flanagan Arlin J. Benjamin, Jr. Janie Eubanks Jeffery C. Valentine

Three experimental studies, one correlational study, and a meta-analysis tested key hypotheses concerning the short-term and long-term impact of exposure to violent video games. Experiment 1 found that violent video games in general increase the accessibility of aggressive thoughts. Experiments 2 and 3 found that playing violent video games increased aggression, even when arousal and aVect were controlled. Experiments 2 and 3 also found that trait hostility and trait aggression were positively related to laboratory aggression. Furthermore, there was correlational evidence of a link between repeated exposure to violent video games and trait aggressiveness. Mediational analyses suggested that the trait eVects and the violent video game eVects on laboratory aggression were partially mediated by revenge motivation. The correlational study uncovered links between habitual exposure to violent video games, persistent aggressive cognitions, and self-reported aggressive behavior. A destructive testing regression approach found that the video game violence/aggression link remained significant even after stressing the link by partialling out sex, narcissism, emotional susceptibility, and Big Five personality factors. However, consistent with prior empirical and theoretical work emphasizing the importance of media violence in the creation of habitual aggressive patterns of thought, partialling out aggressive attitudes reduced the video game violent/aggression link to nonsignificance. The meta-analyses revealed significant eVects of violent video 199 ADVANCES IN EXPERIMENTAL SOCIAL PSYCHOLOGY, VOL. 36

Copyright 2004, Elsevier Inc. All rights reserved. 0065-2601/04 $35.00

200

CRAIG A. ANDERSON et al.

games on aggressive behavior, aVect, and cognition; on cardiovascular arousal; and on prosocial behavior. A best-practices meta-analytic approach revealed that contrary to media industry claims, better conducted studies tend to yield stronger eVects of violent video games on aggression and aggression-related variables than do more poorly conducted studies.

I. Introduction A. CRITICAL INCIDENTS Violent video games are popular with adolescent and adult males and females, are marketed to youth in ways that violate the video game industry’s own standards, and are easily obtained regardless of age (e.g., Buchman & Funk, 1996; Federal Trade Commission, 2000; Walsh, 1999). School shootings by boys with a history of playing violent video games [e.g., West Paducah, KY (December 1997); Jonesboro, AR (March 1998); Springfield, OR (May 1998), Littleton, CO (April 1999), Santee, CA (March 2001), Wellsboro, PA (June 2003) and Red Lion, PA (April 2003)] heightened public debate about the role played by this relatively new violent entertainment medium, including a hearing by the U.S. Senate Committee on Commerce, Science, and Transportation held on March 21, 2000. Other recent violent crimes linked to violent video games include a violent crime spree in Oakland, California (January 2003); five homicides in Long Prairie and Minneapolis, Minnesota (May 2003); beating deaths in Medina, Ohio (November 2002) and Wyoming, Michigan (November 2002); and the Washington, D.C. ‘‘Beltway’’ sniper shootings (Fall 2002). However, such public incidents and outcries do not constitute scientific evidence of a true causal link. This chapter explores current work by media violence researchers and presents five new studies. B. BRIEF HISTORY OF VIOLENT VIDEO GAMES Video games first emerged in the late 1970s, but in the 1990s violent games came of age, with the first-person shooter ‘‘Wolfenstein 3D’’ and the thirdperson fighter ‘‘Mortal Kombat’’ leading the way. By the end of the 20th century, even more graphically violent games were available to virtually anyone who wanted to play them, regardless of age (Walsh, 1999). As early as the mid-1990s, fourth grade girls reported playing video games more than 5 1/2 hours a week, and boys reported playing more than 9 hours a week (Buchman & Funk, 1996). Furthermore, this same sample of fourth graders reported that the majority of their favorite games were violent ones (58.9%

VIOLENT VIDEO GAMES

201

for girls, 72.9% for boys). A survey of eighth and ninth grade students found boys playing about 13 hours a week and girls about 5 hours a week (Gentile, Lynch, Linder, & Walsh, 2004). Data from the Cooperative Institutional Research Program (CIRP, 1998, 1999), which surveys entering college freshmen from more than 600 two- and four-year colleges, reveal that older students also are playing a lot of video games and that their time with such games is also increasing. In 1998 13.3% of the young men reported playing video games at least 6 hours per week during their senior year in high school. By 1999 that figure had increased to 14.8%. Increases are also occurring at the high end of the game playing distribution. In 1998, 2% of the young men reported playing video games more than 20 hours per week. By 1999, that figure had increased to 2.5%. Another troubling aspect involves the lack of parental or societal oversight. A recent survey of teens in grades 8 through 12 (Walsh, 2000) found that 90% of their parents never check the ratings of video games before allowing a purchase, and only 1% reported that their parents had ever kept them from getting a game based on its rating. Furthermore, ratings provided by the video game industry do not match those provided by other adults and game-playing youngsters. Specifically, many games involving violence by cartoon-like characters are classified by the industry as being appropriate for general audiences, a classification with which adults and youngsters disagree (Funk, Flores, Buchman, & Germann, 1999). Also, 89% of the teens in Walsh’s survey (2000) reported that their parents never limit the amount of time they are allowed to play video games. Finally, many of the most violent games have ‘‘demo’’ versions on the Internet that can be downloaded for free by anyone. Of the boys in the sample who play video games, 32% reported downloading them from the Internet (Walsh, 2000).

C. MEDIA VIOLENCE RESEARCH Concern over video game violence would be misplaced if playing such games had little impact on aggression. Decades of research have revealed that viewing television and movie violence can cause short-term increases in aggression and long-term changes in trait aggressiveness (e.g., Bushman & Anderson, 2001; Bushman & Huesmann, 2001; Hearold, 1986; Huesmann & Miller, 1994; Paik & Comstock, 1994; Wood, Wong, & Chachere, 1991). The research literature on video games is smaller and less complete. Despite its relatively small size and the methodological diYculties inherent in the first studies of any ‘‘new’’ phenomenon, a consensus is emerging that violent video games can cause increases in aggressive behavior in children and in young adults (Anderson, 2000; Anderson & Bushman, 2001; Dill & Dill, 1998; Sherry, 2001; Walsh, 2000). We review this literature after considering theoretical issues.

202

CRAIG A. ANDERSON et al.

II. The General Aggression Model Good theoretical reasons support the belief that exposure to violent video games will increase aggressive behavior (Anderson & Dill, 2000; Dill & Dill, 1998). The General Aggression Model (GAM) integrates existing theory and data concerning the learning, development, instigation, and expression of human aggression (Anderson & Bushman, 2002; Anderson & Carnagey, 2004; Anderson & Huesmann, 2003). It does so by noting that the enactment of aggression is based largely on knowledge structures, such as scripts or schemas, created by social learning processes. In brief, GAM describes a multistage process by which two kinds of input variables lead to aggressive (or nonaggressive) behavior. Figure 1 shows a simplified version of the single-episode portion of GAM. Both personological (e.g., trait hostility) and situational (e.g., recent violent video game play) variables influence behavior by aVecting the person’s present internal state, represented by cognitive, aVective, and arousal variables. Playing a violent video game may influence aggression by means of the cognitive route, for example, if it primes aggressive thoughts or scripts, leading to hostile perception, expectation, and attributional biases (e.g., Bushman & Anderson, 2002; Calvert & Tan, 1994; Crick & Dodge, 1994; Dill, Anderson, Anderson, & Deuser, 1997; Kirsh, 1998). The three aspects of present internal state are themselves interrelated, as indicated by the dashed lines connecting them. For example, priming aggressive thoughts might subsequently increase feelings of anger and a desire for revenge if the person is provoked. In such a

Fig. 1. The general aggression model: single episode cycle. Source: Anderson and Bushman, 2002.

VIOLENT VIDEO GAMES

203

case the cognitive eVect is regarded as the primary route of impact and the aVective eVect as a secondary route of impact. The present internal state influences how a person perceives events, interprets their meaning, and chooses behavioral responses. Whatever action the person takes (e.g., aggressive or nonaggressive) influences the present social encounter. This sets the stage for the next round in the social interaction cycle. Of particular relevance to the present article is the fact that finding significant links between exposure to violent video games and aggression does not by itself reveal the primary route of the obtained eVect; it could have occurred via cognition, aVect, or arousal, or some combination. However, such fine distinctions are crucial to theoretical development and to the public policy debate over whether parents should be given tools to help them control their children’s access to violent games. In both the theory and the public policy domains, the precise route(s) of primary impact are important because of the developmental aspects of GAM. Long-term eVects of media violence involve learning processes, such as learning how to perceive, interpret, judge, and respond to events in the physical and social environment. Various types of knowledge structures (e.g., perception, interpretation, judgment, and action) develop over time and are based on day-to-day observations of and interactions with other people, real (e.g., family) and imagined (e.g., media). Each violent episode, as outlined in Fig. 1, is essentially one more learning trial. Short-term eVects become ingrained through the development of aggression-related knowledge structures, which persistently color the person’s expectations and perceptions concerning social interactions, especially those with conflictual content. In a very real sense, a person’s set of chronically accessible knowledge structures defines that person’s personality. Figure 2 displays this developmental aspect of GAM and identifies five types of variables that contribute to the development of an aggressive personality—aggressive beliefs and attitudes, aggressive perceptual schemata, aggressive expectation schemata, aggressive behavior scripts, and aggression desensitization. Four of these variables involve aggressive cognitions. For this reason, short-term eVects of violent media on aggression via the cognitive route are particularly important. Temporary mood states and arousal dissipate over time, but rehearsal of aggressive cognitions can lead to long-term changes in multiple aspects of aggressive personality. Furthermore, the literature on the development of behavioral scripts suggests that even a few rehearsals can change a person’s expectations and intentions involving important social behaviors (Anderson, 1983; Anderson & Godfrey, 1987; Marsh, Hicks, & Bink, 1998). Figure 3 illustrates the dynamic aspects of the episodic and developmental portions of GAM. Exposure to violent video games can serve as a proximate situational cause, increasing the likelihood that an aggressive behavior will occur

204

CRAIG A. ANDERSON et al.

Fig. 2. The general aggression model: developmental/personality processes. Source: Anderson and Carnagey, 2004.

shortly after the exposure, but it also can serve as a distal environmental modifier, influencing the development of aggression-related knowledge structures and hence, aggressive personality.

III. GAM and Violent Video Games A. BASIC ISSUES Two levels of questions emerge from consideration of GAM and violent video games. First, can violent video games cause increases in aggression? Answering this question does not require a careful analysis of the multiple processes by which input variables can influence the expression of aggressive behavior. It merely requires a body of research in which the eVects of violent games are compared to nonviolent games or other appropriate control

VIOLENT VIDEO GAMES

Fig. 3.

205

The general aggression model: overall view. Source: Anderson and Carnagey, 2004.

conditions. Much of the existing video game literature is of exactly this nature, and it is at this level that a review of the literature reveals considerable support for the hypothesis that playing violent video games can increase aggression (Anderson & Bushman, 2001). Both experimental and correlational studies, on average, yield significant positive relations between exposure to violent video games and aggressive behavior, with average eVect sizes in the r þ ¼ 0.20 range (Anderson & Bushman, 2001). The experimental studies demonstrate that a brief exposure to violent video games causes an immediate (and presumably short-lived) increase in aggressive behavior. The correlational studies link repeated exposure to violent video games with a variety of types of real-world aggressive behavior, including violent criminal behavior. In sum, despite its relatively small size and recent history, the research literature has demonstrated that violent video games can increase aggression. The second level of questions to emerge concerns specificity of violent content eVects on aggression via the cognitive route. Nonviolent games can

206

CRAIG A. ANDERSON et al.

also increase aggressive feelings if, for example, they produce high levels of frustration. They also can increase arousal, if they are suYciently demanding and engaging. Thus, both violent and nonviolent games may influence aggression in the immediate situation if they increase aggressive feelings or arousal. The real crux of the debate over eVects of violent video games lies in their unique ability to directly increase aggressive cognitions, cognitions that can have a much longer lasting impact if their repeated instantiation (by repeated violent video game play, for instance) leads to persistent changes in key knowledge structures, such as more positive attitudes toward aggression.

B. KEY QUESTIONS Three key second-level questions remain unanswered by existing research on violent video games. First, do violent video games generally increase aggressive cognitions? Several studies have found significant increases in aggressive thoughts as a function of exposure to violent video games (see Anderson & Bushman, 2001), but most have not explicitly controlled for other potential diVerences between the target video games, such as diVerences in aVective or arousal properties. In fact, only one published experimental study has successfully controlled for both arousal and aVective features (Anderson & Dill, 2000, Study 2); the violent game yielded higher aggressive cognition scores. Second, does violent video game content by itself cause short-term increases in aggressive behavior tendencies? As GAM makes clear, to cleanly test this specific violent content question in an experimental setting the comparison violent and nonviolent video games should be equated on arousal level and aVective factors such as enjoyment, frustration, and state hostility. Such controls can be done by selecting violent and nonviolent games that do not diVer on these factors, or by including measures of arousal and aVect and using them as statistical controls. Only two published experimental studies meet these criteria. Graybill, Strawniak, Hunter, and O’Leary (1987) pretested several video games and equated them on diYculty, excitement, and enjoyment. Interestingly, this is one of the few experimental studies that failed to find a significant eVect of video game violence on aggressive behavior. One potential problem with this study is that the pretest ratings of the various games were done by graduate students, whereas the participant population was second through sixth graders. In other words, the ‘‘equating’’ process was somewhat less than optimal. The other experimental study meeting these control criteria was Study 2 of Anderson and Dill (2000), which found a significant increase in aggression attributable to diVerential violent content. Third, is repeated exposure to violent video games associated with higher aggression levels, and is this mediated by persistently elevated levels of

VIOLENT VIDEO GAMES

207

aggressive thoughts? To show such an eVect, significant positive associations between violent video game exposure, aggressive behavior, and persistent aggressive thoughts would have to be found. Furthermore, the link between video game violence exposure and aggression must be significantly weakened when the persistent aggressive thoughts measure is statistically controlled. As of this writing, there are no published correlational studies of this type.

IV. Overview of the Present Studies Experiment 1 used 10 video games to examine the eVects of violent content on the accessibility of aggressive thoughts, physiological arousal, and aggressive aVect. The results of Experiment 1 were used to select a pair of violent and nonviolent video games matched on arousal and aVective dimensions but diVering in violent content for use in Experiments 2 and 3. Experiment 2 tested whether these two games produce diVerent levels of short-term aggressive behavior. Experiment 3 used two violent and two nonviolent games and a diVerent aggression paradigm in an attempt to replicate the specific violent video game content eVect on aggression. It also provided our first test of one factor that might increase or decrease the violent video game eVect, specifically the realism of the video game targets of aggression. Experiments 2 and 3 also examined trait hostility and revenge motivation eVects on aggression. Correlation Study 1 assessed violent video game exposure, self-reported aggressive behavior, Big five personality factors, and attitudes toward violence. It tested a basic-personality-as-artifact hypothesis as well as an aggressive-cognition-mediation hypothesis. The final new study was an updated meta-analysis of violent video game eVects on aggressive behavior, thoughts, and aVect; physiological arousal; and prosocial behavior. This analysis also compared average eVect sizes of the methodologically best studies to those with significant weaknesses.

V. Experiment 1 A. METHOD 1. Participants Participants were 61 male and 69 female undergraduate students who participated in partial fulfillment of an introductory psychology research requirement. Participants were asked to refrain from alcohol, caVeine,

208

CRAIG A. ANDERSON et al.

tobacco products, and exercise for 12 hours prior to the study start time. Participants were randomly assigned to play one of 10 video games. 2. Materials a. Video Games. Ten video games were selected through a review of video game sites on the World Wide Web, popular magazines, and retail outlets. The five violent games were: Dark Forces, Marathon 2, Speed Demon, Street Fighter, and Wolfenstein 3-D. The five nonviolent games were: 3-D Ultra Pinball, Glider Pro, Indy Car II, Jewel Box, and Myst (see the Appendix for a description of the 10 games). We attempted to find games that did not require extensive practice to achieve at least marginal proficiency and that were relatively involving for a college student population. Street Fighter was played on the Super Nintendo Entertainment System, using a 19-inch Sony color television. All other games were played on a Macintosh computer. b. Video Game Experience. Participants estimated the average number of hours per week spent playing video games in the past few months. Each participant also rated his or her experience with the 10 video games used in this study and an additional four games (SimCity, You Don’t Know Jack, Computer Cribbage, A10 Attack) using a 7-point unipolar scale anchored at 1 (Never have played), 4 (Have played some), and 7 (Have played often). c. Video Game Ratings. After playing the assigned video game, participants completed a 6-item rating scale about the game, rating how diYcult the game was to learn, how enjoyable the game was to play, how much action the game had, how violent the game graphics were, how violent the game content was, and how frustrating the game was (Anderson & Ford, 1986). For all items, a response of 1 indicated ‘‘low’’ and a response of 7 indicated ‘‘high’’ on the adjective of interest. Preliminary analyses revealed that the two items measuring the violence of video game content and graphics were highly correlated, so they were averaged to form a composite measure of perceived video game violence. d. Word Completion Task. The word completion task (Anderson, Carnagey, & Eubanks, 2003; Roediger, Weldon, Stadler, & Riegler, 1992) involves examining a list of 98 words with one or more letters missing, and filling in the missing letters. The missing letters are strategic, such that each item can make more than one word. For instance, one item is ‘‘explo_e,’’ which may be completed as ‘‘explore’’ or ‘‘explode.’’ Participants were told that their task was to fill in the blanks to make complete words. Participants were given 3 minutes to complete as much of the task as they could. An accessibility of aggressive thoughts score was calculated for each participant by dividing the number of aggressive word completions by the total number

VIOLENT VIDEO GAMES

209

of completions. Forty-nine of the items can yield an aggressive word when completed. e. Cardiovascular Measures. Heart rate (HR), systolic blood pressure (SBP), and diastolic blood pressure (DBP) were measured with an A & D Medical automatic digital blood pressure monitor (model UA-751). The blood pressure cuV was attached to each participant’s nondominant upper arm, approximately 1 inch above the elbow. All measurements were obtained while the participants were seated. At each of three measurement periods, HR, SBP, and DBP were collected twice, with approximately 1 minute elapsing between the end of the first measurement and the beginning of the second. The first measurement period (baseline) was after signing the consent form but before game play began. The second measurement was during video game play, and the third was after the video game task. 3. Procedure Participants were told that the study concerned the ways people learn diVerent types of computer tasks. They were informed that we were interested in possible diVerences between simple and complex tasks. After completing consent procedures, participants entered a cubicle and completed the Video Game Experience questionnaire. Participants then read directions for the assigned game. The experimenter started the game for the participants. Participants played for approximately 20 minutes. Physiological measures were collected before the start of the game, approximately 10 minutes into game play, and immediately after the game. Participants next performed the word completion task and the Video Game Ratings measure. Participants were debriefed and thanked for their participation.

B. RESULTS AND DISCUSSION This study had two main goals. One was to compare violent and nonviolent video games on key dimensions so that a pair of games, matched on all dimensions except for violence, could be selected for further in-depth research. The second was to test the hypothesis that playing violent video games primes aggressive thoughts. For all analyses we tested a planned contrast that compared the mean score of participants who had played one of the violent games to the corresponding mean of those who had played one of the nonviolent games. Sex eVects were expected to occur on some measures (e.g., cardiovascular measures, enjoyment of the games) and not on others. When preliminary analyses revealed no sex eVects, sex was dropped from the statistical model.

210

CRAIG A. ANDERSON et al.

1. Cardiovascular Measures a. Blood Pressure. A 10 (Video game)  2 (Sex: female vs. male)  2 (Type: diastolic vs. systolic)  3 (Time of assessment: before video game vs. during video game vs. after video game) ANOVA, with type and time as repeated factors, revealed several interesting eVects. First, there were significant main eVects of sex [F(1, 86) ¼ 26.11, P < .001, d ¼ 0.43], time [F(2, 172) ¼ 9.82, P < .001], and type [F(1, 86) ¼ 2312.64, P < .001, d ¼ 4.02]. Males had higher blood pressure than females (Ms ¼ 93.46 and 86.27).1 Blood pressure increased from the baseline period (before playing the video game, M ¼ 89.92) to the video game play period (M ¼ 91.73) and then decreased after game play was complete (M ¼ 87.94). Of course, SBP was greater than DBP (Ms ¼ 110.52 and 69.21, respectively). There was no main eVect of which game was played [F(9, 86) ¼ 1.67, P > .10], nor was there any hint of a violent vs. nonviolent game eVect [F(1, 86) ¼ 0.02, diVerence not significant]. However, the sex, time, and type main eVects were all moderated by various two-way interactions. The type X sex interaction [F(1, 86) ¼ 18.33, P < .001, d ¼ 0.32] resulted from the fact that the sex eVect was larger for SBP (Mmales ¼ 115.95 vs. Mfemales ¼ 105.8) than for DBP (Mmales ¼ 70.97 vs. Mfemales ¼ 67.45). The type X time interaction [F(2, 172) ¼ 11.88, P < .001] resulted from the fact that SBP remained relatively constant across the three time periods (Mbefore ¼ 111.64 vs. Mduring ¼ 111.09 vs. Mafter ¼ 108.83), whereas DBP was highest during video game play (Mbefore ¼ 68.21 vs. Mduring ¼ 72.37 vs. Mafter ¼ 67.06). The omnibus time X game interaction [F(18, 172) ¼ 1.93, P < .02] is a bit diYcult to comprehend. However, the more specific contrast testing the time  violent vs. nonviolent game interaction was also significant [F(2, 172) ¼ 5.37, P < .01] and accounted for much of the omnibus interaction.2 On average, participants who played one of the nonviolent games showed a decline in blood pressure across the three time periods (Mbefore ¼ 91.41 vs. Mduring ¼ 90.41 vs. Mafter ¼ 88.07). However, those who played one of the violent games showed an increase in blood pressure during video game play, followed by a decrease (Mbefore ¼ 88.44 vs. Mduring ¼ 93.04 vs. Mafter ¼ 87.81). In other words, within the present sample of 10 games the violent ones increased arousal, as measured by blood pressure, whereas the nonviolent ones did not. This point is especially important to remember when 1 All reported means are the appropriate adjusted means. All P-levels are based on two-tailed tests. Sample sizes diVer somewhat in diVerent analyses as the result of occasional missing values. This occurred most frequently for the cardiovascular measures because of occasional equipment malfunctions. 2 In fact, the residual omnibus time  game interaction was nonsignificant [F(16, 172) ¼ 1.50, P > .10].

VIOLENT VIDEO GAMES

211

selecting pairs of games for more in-depth research on the relation between violent content and aggressive behavior. It is also important to note, however, that the postgame assessment of blood pressure showed nearly identical means for the violent and nonviolent games. Therefore, the potential problem of arousal being confounded with violent vs. nonviolent content may not be as severe as many might assume. b. Heart Rate. Preliminary analyses yielded no sex eVects. A 10 (Video game) X 3 (Time of assessment) ANOVA with time as a repeated factor yielded only one marginally significant eVect, the time main eVect [F(2, 190) ¼ 2.82, P < .07]. HR was highest when assessed during game play (M ¼ 77.50), and was at about the same relatively low level before (M ¼ 74.73) and after (M ¼ 74.28) game play. Indeed, a specific contrast comparing HR during game play to the average HR before and after game play was statistically significant [F(1, 95) ¼ 4.43, P < .04, d ¼ 0.30]. None of the video game eVects (omnibus or violent vs. nonviolent contrast) approached significance (all P values > .40). 2. Video Game Ratings The mean ratings of the games on diYculty, enjoyment, action, frustration, and violence are shown in Table I. There was considerable variability on all five dimensions, which made selection of a pair of games diVering primarily on violent content possible. Ratings of game diYculty were aVected by game [F(9, 120) ¼ 7.86, P < .001]. The specific contrast comparing violent to nonviolent games was also significant and revealed the violent games to be more diYcult (M ¼ 3.60) than the nonviolent games (M ¼ 2.80) [F(1, 120) ¼ 8.67, P < .01, d ¼ 0.52]. However, there was considerable variability within each game type and considerable overlap between the violent and nonviolent games. The violent game diYculty means ranged from 2.09 (Wolfenstein 3D) to 4.38 (Street Fighter). The nonviolent game diYculty means ranged from 1.54 (Ultra Pinball) to 4.83 (Myst). Participants’ enjoyment ratings were also aVected by game [F (9, 110) ¼ 2.23, P < .05]. However, the comparison between violent and nonviolent games showed that participants enjoyed the violent and nonviolent games equally (F < 1). In fact, the nonviolent games were enjoyed slightly more than the violent games. The game  sex interaction was also significant [F(9, 110) ¼ 1.97, P < .05]. However, the specific contrast testing the sex  game violence interaction was only marginally significant [F(1, 110) ¼ 3.33, P < .08]. The slight preference for the nonviolent games was marginally greater for females (Mnonviolent ¼ 3.81, Mviolent ¼ 3.53) than for males (Mnonviolent ¼ 4.23, Mviolent ¼ 4.11). Ratings of game action were significantly aVected by game [F(9, 120) ¼ 8.35, P < .001]. As shown in Table I, participants found the violent games to

212

CRAIG A. ANDERSON et al.

TABLE I Mean Rating of Video Game Difficulty, Enjoyment, Action, Frustration, and Violence as a Function of Game, and Averages for Nonviolent and Violent Games Nonviolent games

DiYculty

Enjoyment*

Action

Frustration

Violence*

Myst

4.83

3.06

1.67

5.33

1.25

Jewel Box

1.77

4.38

2.62

3.69

1.00

Indy Car II

2.21

3.31

2.71

3.79

1.43

3-D Ultra Pinball

1.54

5.40

4.08

2.69

1.62

Glider Pro

3.62

3.94

2.31

4.75

1.41

Average

2.80

4.02

2.68

4.05

1.34

Violent games

DiYculty

Enjoyment

Action

Frustration

Violence

Speed Demon

4.00

3.86

3.42

5.33

3.27

Dark Forces

3.27

4.10

3.18

4.27

4.35

Wolfenstein 3D

2.09

3.70

5.36

3.82

5.92

Marathon 2

4.25

3.69

3.67

4.25

4.86

Street Fighter

4.38

3.74

4.81

4.31

4.94

Average

3.60

3.82

4.09

4.40

4.67

Mean square error

2.37

2.35

2.01

2.38

.97

*Adjusted for sex eVects. Possible range of ratings was 1 to 7.

contain more action (M ¼ 4.09) than the nonviolent games (M ¼ 2.68) [F(1, 120) ¼ 31.67, P < .001, d ¼ 0.99]. Participants’ ratings of frustration showed a significant game eVect [F(9, 120) ¼ 3.40, P < .001]. However, the contrast between violent and nonviolent video games was nonsignificant, suggesting that participants found the violent games (M ¼ 4.40) and nonviolent games (M ¼ 4.05) equally frustrating [F(1, 120) ¼ 1.61, P > .20, d ¼ 0.23]. Ratings of the violence of game content showed a significant sex eVect [F(1, 110) ¼ 4.82, P < .05, d ¼ 0.40]. Females rated the games as more violent than did males (Ms ¼ 3.20 & 2.81, respectively). More importantly, the violence ratings also yielded a significant game eVect [F(9, 110) ¼ 45.81, P < .001]. As expected, most of this omnibus game eVect (87%) was due to the much higher violence ratings given for the violent games than for the nonviolent games (Ms ¼ 4.68 & 1.35, respectively,) [F(1, 110) ¼ 350.75, P < .001, d ¼ 3.38]. Furthermore, the sex by game type interaction was not significant [F(1, 110) ¼ 0.92, P > .3], indicating that the sex eVect did not systematically vary as a function of whether the game was violent or nonviolent.

VIOLENT VIDEO GAMES

213

3. Selection of a Matched Pair We conducted a number of additional analyses to select a violent/nonviolent game pair most closely matched on irrelevant dimensions (i.e., diYculty, enjoyment, action, and frustration) and diVering greatly on violence. Although several pairings appeared to meet our criteria fairly well, we ultimately chose Glider Pro and Marathon 2. We conducted a 2 (game)  2 (sex)  4 (rating dimension) ANOVA on ratings of these two games, treating the four ‘‘irrelevant’’ rating dimensions (diYculty, enjoyment, action, frustration) as a repeated measures factor. DiVerential eVects of game would show up either as a game main eVect or a game  rating dimension interaction. The only eVect that approached statistical significance was the main eVect of rating dimension [F(3, 72) ¼ 3.93, P < .05]. The adjusted means for diYculty, enjoyment, action, and frustration were 3.96, 3.81, 2.98, and 4.51, respectively. This eVect, of course, is irrelevant to the issue of game selection. All other eVects had P values > .15. We also compared cardiovascular eVects of these two games. DiVerential game eVects would appear as an interaction involving the game and time of assessment variables. None of the game  time interactions (2-way or higher) were significant. Finally, we compared these two games on the violence ratings. The only significant eVect was a large eVect of game on the violence rating [F(1, 24) ¼ 53.81, P < .001, d ¼ 2.82]. Marathon 2 was rated as considerably more violent than Glider Pro [Ms ¼ 4.86 and 1.41, respectively]. The sex main eVect approached significance [F(1, 24) ¼ 4.24, P < .06, d ¼ 0.79], with females rating the games as more violent than males [Ms ¼ 3.62 and 2.65, respectively]. The sex  game interaction did not approach significance [F(1, 24) ¼ 0.11, P > .70]. Thus, Glider Pro and Marathon 2 were well matched on the irrelevant dimensions and diVered greatly on the desired dimension of violence. Of note, these specific game comparisons were based on a relatively small sample size. Therefore, in Experiment 2 on aggressive behavior (which used these two games), we measured these same rating dimensions, as well as blood pressure and HR, as additional statistical controls. 4. Accessibility of Aggressive Thoughts The second major goal of this study was to test the hypothesis that playing a violent video game can increase the relative accessibility of aggressive thoughts. A series of analyses was conducted to compare the percentage of aggressive word completions after violent versus nonviolent video game play. Each analysis included a specific contrast comparing the average of the nonviolent game conditions to the average of the violent game conditions.

214

CRAIG A. ANDERSON et al.

Preliminary analyses indicated that sex of participant did not have any significant impact on the accessibility of aggressive thoughts measure. a. All 10 Games. Participants produced a significantly higher percentage of aggressive words after violent games (M ¼ 14.7) than after nonviolent games (M ¼ 12.5) [F(1, 120) ¼ 4.26, P < .05, d ¼ 0.37]. Thus, as predicted by GAM, playing violent games increased accessibility of aggressive thoughts, relative to playing nonviolent games. We conducted several similar analyses with various covariates added to the model. When we controlled for the rating dimensions of diYculty, enjoyment, action, and frustration in this way, the violent vs. nonviolent contrast was still significant [F(1, 116) ¼ 5.29, P < .05]. Similarly, the game violence eVect was still significant when we controlled for video game experience (hours per week and average experience with 10 specific games) [F(1, 110) ¼ 4.54, P < .05]. When physiological arousal changes were statistically controlled (baseline to after video game play), the game violence eVect was still significant [F(1, 103) ¼ 5.74, P < .05]. However, when we controlled for rated violence of the games, the game violence eVect became nonsignificant [F(1, 119) ¼ 2.35, P > .12], as expected. In fact, controlling for the rated violence of the games reduced the game violence eVect on aggressive thoughts by 45%. Finally, we performed an analysis that treated the specific games chosen to represent the violent and nonviolent types as random eVects, rather than fixed (Winer, 1971). This entails using the between groups sums of squares, within each of the violent and nonviolent game types, to estimate measurement error, with 8 degrees of freedom. This tests the generalizability of the video game violence eVect across specific games (Wells & Windschitl, 1999). It also yielded a significant game violence eVect [F(1, 8) ¼ 8.96, P < .05]. These results strongly support the hypothesis that violent content in video games causes increases in the accessibility of aggressive thoughts, independent of arousal and aVective influences. This is the first study to conclusively demonstrate this specific eVect of violent video game content. b. Glider Pro vs. Marathon 2. We further examined the violent video game eVect on aggressive thoughts by using only the two matched games Glider Pro and Marathon 2. Sex eVects did not approach significance, so sex was dropped. Despite the small sample size, Marathon 2 participants produced a significantly higher percentage of aggressive words (M ¼ 15.4) than did Glider Pro (M ¼ 10.8) [F(1, 26) ¼ 6.03, P < .05, d ¼ 0.95]. A similar statistical model that included changes in physiological arousal (baseline to after video game play) as covariates yielded similar results. The eVect of game was significant [F(1, 21) ¼ 10.96, P < .01]. We then ran a series of analyses in which each of the video game rating dimensions were entered as covariates. The analyses with diYculty, enjoyment,

VIOLENT VIDEO GAMES

215

action, and frustration each yielded a significant game eVect on aggressive thoughts [Fs(1, 25) > 5.00, P values < .05]. As expected, when rated violence of the game was entered as a covariate the game eVect became nonsignificant [F(1, 25) ¼ 2.42, P > .10], again demonstrating the specificity of violent content eVects.

VI. Experiment 2 A. METHOD 1. Participants Ninety-seven female and 93 male undergraduates participated, selected on the basis of their responses to the Trait Hostility (TH) scale, administered at the beginning of the semester as part of a battery of questionnaires. Half of the males and half of the females were selected from the top and bottom thirds of the TH distribution. All participated in return for partial course credit for their introductory psychology class.3 2. Design The experiment can be conceived as a 2 (TH: high vs. low)  2 (Video game: violent vs. nonviolent)  2 (Provocation pattern: increasing vs. ambiguous)  2 (Sex: male vs. female) between subjects design. However, TH was used as a continuous variable in all analyses, rather than a two-level categorical variable. This regression approach is statistically more appropriate and more powerful. For all analyses that included a covariate (e.g., TH), we also tested all possible interaction terms involving the covariate. None of these interactions were statistically reliable, so they were dropped from the final statistical models reported in this article. For dependent variables that

3 Six participants were dropped because they reported unusually high levels of video game experience or number of hours per week spent playing video games. Our concern was that participants who had extensive video game experience might respond somewhat diVerently than the normal population. We used the standard procedures described by Tukey (1971) for identifying outliers and ‘‘far outliers.’’ Six far outliers were identified, two on the basis of the video game experience scale, four on the hours per week measure. For example, the far outliers on the latter measure reported playing video games from 18 to 24 hours per week in recent months. Five of the six far outliers were males. Supplemental analyses that included these participants yielded the same patterns of aggression as those reported in the Results section, albeit the eVect sizes were slightly smaller.

216

CRAIG A. ANDERSON et al.

were not significantly influenced by sex of participant, the reported results are based on analyses that dropped sex from the statistical model. 3. Materials a. Trait Hostility. The Trait Hostility Scale is a 30-item self-report inventory designed to measure chronic individual diVerences in aggressiveness or irritability, adapted from the irritability scale developed by Caprara, Cinanni, D’Imperio, Passerini, Renzi, and Travaglia (1985) (see Anderson, 1997; Dill et al., 1997). Sample items include (1) Whoever insults me or my family is looking for trouble, (2) Sometimes I shout, hit and kick and let oV steam. The internal reliability of this scale is generally high; in the present sample coeYcient alpha ¼ .86. b. Video Games. As described earlier, Marathon 2 (violent) and GliderPro (nonviolent) were selected because they were similar on a variety of dimensions in Experiment 1. c. Competitive Reaction Time Task. A modified version of the Taylor Competitive Reaction Time (CRT) task was used to assess aggressive behavior. The CRT is a widely used and externally valid measure of aggressive behavior (Anderson & Bushman, 1997; Anderson, Lindsay, & Bushman, 1999; Carlson, Marcus-Newhall, & Miller, 1989; Giancola & Chermack, 1998). Participants believe they are competing with another person to see who can respond first upon presentation of a tone. In the standard version of the game, after each trial the ‘‘loser’’ receives an aversive punishment (e.g., loud noise), the intensity of which is supposedly set by the opponent. Prior to each trial, each participant sets the punishment level that supposedly will be delivered to the opponent, if the participant wins the trial. The possible settings range from 0 (no noise) to 10 (100 db). These settings constitute the measure of aggressive behavior. In the present experiment we used a two-phase version of the task (Anderson, Anderson, Dorr, DeNeve, & Flanagan, 2000; Bartholow & Anderson, 2002; Lindsay & Anderson, 2000). In Phase 1 participants were told that their opponent would set the intensity of the noise blast that the participant would receive as ‘‘punishment’’ on ‘‘lose’’ trials, but that the opponent would not be punished on trials that the participant won. It was further explained that in Phase 2, the roles would be reversed so that the participant would set the intensity of the noise blasts for the opponent, but the participant would not receive punishments on ‘‘lose’’ trials. In actuality, a computer program determined wins and losses as well as the noise intensities and durations delivered to participants in Phase 1. After each trial the participant also saw what noise level was ‘‘set’’ by the opponent, displayed on the computer

VIOLENT VIDEO GAMES

217

screen. All participants were given sample noise blasts of level ‘‘2’’ (60 db) and ‘‘8’’ (90 db) before beginning Phase 1 of the task. d. Provocation Pattern Manipulation. In Phase 1 of the CRT task, all participants received the same randomly ordered series of 13 wins and 12 losses determined by the computer program. They also received blasts of noise on the ‘‘lose’’ trials. The pattern of noise blasts was either an ambiguous or an increasing provocation pattern (Anderson et al., 2000). Participants in both provocation conditions saw exactly the same punishment settings—8 in the low range, 9 in the middle range, and 8 in the high range, and actually received the same punishments (4 in each range). In the ambiguous provocation condition the pattern of noise intensities was random, whereas the increasing provocation pattern consisted of mostly low noise intensities on the early trials, middle intensities on the middle trials, and high intensities on later trials. On completion of Phase 1, the experimenter reminded participants that in Phase 2 they would set noise intensities to be delivered to their opponent, and that they would not receive noise blasts on any trial in this phase. 4. Procedure Participants were randomly assigned to one of the four experimental conditions (video game  provocation), with the constraint that equal proportions of males and females and that equal proportions of high and low TH participants were run in each condition. Same-sex pairs were run in individual cubicles. On arrival, they were seated in cubicles and asked to read and sign the consent form and to then read instructions for the video game. Participants were told that the study was concerned with the ways people learn diVerent types of computer tasks. Participants were informed that we were interested in possible diVerences between simple and complex tasks, and that we would be measuring HR and blood pressure throughout the session. Participants played the assigned video game for 20 minutes. Then the CRT task was explained. All participants were told that their opponent would set the noise blasts in Phase 1 while they would set the noise blasts for their opponent to hear in Phase 2. After the CRT task, participants completed a questionnaire that included a manipulation check, some motivation and aVect items, some demographic information, and several suspicion check items. Blood pressure and pulse were measured at five diVerent points in time: (1) before playing the video game, (2) about 10 minutes into the video game, (3) after playing the video game, (4) after Phase 1 of the CRT task, and (5) after Phase 2 of the CRT task. Finally, the participants were debriefed. Care

218

CRAIG A. ANDERSON et al.

was taken to ensure that each participant knew it was the computer and not the other participant in the session that set the noise blasts in Phase 1 and that the noise blasts they sent in Phase 2 were received only by the computer. 5. Measures a. Aggressive Behavior. Aggressive behavior was operationalized as the noise intensity (0–10) sent by the participants to their opponents in Phase 2 of the CRT task. As is common with the CRT task, we examined four diVerent intensity measures: intensity setting on the first trial, and the average settings on trials 2–9, 10–17, and 18–25. The early trials are the most important in this two-phase version of the CRT, especially Trial 1, because it is the first opportunity the participant has to retaliate after being provoked. Note that in the standard one-phase version Trial 1 occurs prior to any provocation, provocations continue throughout the 25 trials, and therefore all trials are of theoretical interest. b. Cardiovascular Arousal. At each of the five measurement periods, blood pressure and pulse were assessed twice, with an interval of 1 minute between the completion of the first measurement and the beginning of the second.4 c. Questionnaires. After the last blood pressure and pulse measurements, participants answered a number of questions about the experiment. One item asked participants if they were ever ‘‘angry’’ during the reaction time task. Responses were on a 5-point unipolar scale anchored at 1 (not at all), 2 (a little bit), 3 (somewhat), 4 (quite a lot), and 5 (a lot). Also included were six items on which participants were to ‘‘indicate the extent to which this motive describes your motive when deciding on where to set the noise levels.’’ These items used the same 5-point scale described above. The six items were (1) I wanted to impair my opponent’s performance in order to win more; (2) I wanted to control my opponent’s level of responses; (3) I wanted to make my opponent mad; (4) I wanted to hurt my opponent; (5) I wanted to pay back my opponent for the noise levels he/she set; (6) I wanted to blast him/her harder than he/she blasted me. The first two items represent instrumental reasons for aggressing. Responses to these two 4 Heart rate and blood pressure were assessed for two reasons. First, it was an important part of the cover story. Second, we wanted to be able to statistically control for any eVects of the video game manipulation on aggression that might be due to changes in arousal. As expected, the video games did not produce diVerential changes in arousal. Including the cardiovascular change measures as covariates in later analyses of aggressive behavior did not alter the pattern of results in any appreciable way (but did reduce the error term and slightly increased the eVect sizes of the video game and provocation manipulations). Thus, for the sake of simplicity these arousal measures are not further discussed.

VIOLENT VIDEO GAMES

219

items were positively correlated (r ¼ .57, P < .001), and were combined to form a scale labeled ‘‘Instrumental Aggressive Motivation.’’ The latter four items represent a clearly revengeful type of aggressive motive, were highly correlated, and were combined to form a scale labeled ‘‘Revenge Motivation.’’ CoeYcient alpha for this scale was .74. One question asked ‘‘Did the pattern of noise levels that you received appear to be increasing, decreasing, or random?’’ This item was coded as þ1 for increasing, 1 for decreasing, and 0 for random. We expected this manipulation check to yield smaller scores in the ‘‘ambiguous’’ than in the ‘‘increasing’’ provocation condition, because the actual pattern was random in the former and increasing in the latter. A 2 (video game)  2 (provocation pattern) analysis of variance (ANOVA) confirmed this expectation. Those in the increasing provocation conditions reported an increasing pattern of noises (M ¼ .43), whereas those in the ambiguous condition reported that the pattern seemed random [(M ¼ .05), F(1, 178) ¼ 43.91, P < .001, d ¼ 0.95]. An open-ended question asked ‘‘What do you think the purpose of this experiment was?’’ Responses to this item and to the final oral debriefing were used by the experimenter to rate each participant as either not suspicious (0), slightly suspicious (1), or suspicious (2). Fourteen participants were classified as suspicious. However, suspicion was unrelated to performance on the main dependent variables, so all were kept. Participants then completed a ‘‘Background Questionnaire’’ assessing demographic information including height, weight, age, year in school, academic major, and time since using alcohol, caVeine, and exercising. Participants also estimated how many hours per week they had spent playing video games ‘‘in recent months.’’ They then indicated ‘‘How much you have ever played each of . . . ’’ 12 specific video games, using a 5-point unipolar scale anchored at 1 (Never have played) and 5 (Have played a lot). The 12 games listed were Wolfenstein 3D, You Don’t Know Jack, Civilization II, A-10 Attack, Computer Cribbage, Street Fighter, Dark Forces, Myst, Indy Car II, Speed Demon, 3-D Ultra Pinball, and Jewelbox. Ratings on these 12 items were combined into a Video Game Experience composite, with a coeYcient alpha of .63. Finally, a thorough oral debriefing was given, with notes recorded by the experimenter.

B. RESULTS 1. Aggressive Behavior Four measures of aggressive behavior were based on the noise punishment levels that participants set for their opponents during Phase 2 of the CRT task. Recall that the key behavioral measure was noise intensity on Trial 1,

220

CRAIG A. ANDERSON et al.

because it was the participant’s first opportunity to retaliate for noise blasts received during Phase 1. The noise settings for the remaining 24 trials were averaged into three blocks of eight trials each. The key prediction was that exposure to the violent video game would increase aggressive behavior in the two-phase CRT task. We believed that this eVect was most likely to occur in the ambiguous provocation condition, especially on the first or early trials (a time  video game  provocation pattern interaction). This is based on an outburst/social justice model of aggressive behavior in the two-phase CRT first described in the Anderson et al. (2000) studies of temperature eVects. Experiment 5 of that work used a very similar two-phase CRT task with both an ambigous and an increasing provocation pattern and found that most of the interesting temperature eVects occurred in the ambiguous pattern condition, especially on the early trials. We also expected that TH would be positively related to aggression. All of these predictions were borne out. We conducted a 2 (video game)  2 (provocation pattern)  4 (time: trial 1 vs. block 1 vs. block 2 vs. block 3) ANOVA, with the last factor as a repeated measures factor, and with TH as a covariate. We then conducted a set of planned contrasts examining the violent video game eVect separately for the two provocation conditions, for each of the four aggression measures. Both the provocation and the TH main eVects were significant [Fs(1, 179) ¼ 5.81 and 4.69, respectively, Ps < .05]. Participants in the increasing provocation condition set lower noise punishments than those in the ambiguous provocation condition (Ms ¼ 5.06 and 5.51, respectively, d ¼ 0.18). Trait hostility was positively related to noise punishment level (b ¼ 0.23). This latter finding provides additional evidence that the two-phase CRT task validly assesses aggressive behavior. There was also a significant time eVect, a time  provocation pattern interaction, and a time  video game  provocation pattern interaction [Fs(3, 537) ¼ 27.90, 19.81, and 4.06; Ps < .01]. Figure 4 illustrates both the time main eVect and the time  provocation pattern interaction. A simple summary of these eVects is that participants tended to produce the same pattern of noise punishments for their opponents in Phase 2 as they had received from their opponents in Phase 1. On average, those in the ambiguous condition gave moderate punishments across the four times, whereas those in the increasing provocation condition delivered punishments that started low and increased across time. Interestingly, this same aggression pattern has occurred in the other two experiments that used the two-phase CRT and both the ambiguous and the increasing provocation pattern (Anderson et al., 2000, Experiment 5; Lindsay & Anderson, 2000, Experiment 4).

221

VIOLENT VIDEO GAMES

Fig. 4. Noise punishment settings as a function of provocation pattern and time of aggression in Phase 2 of the competitive reaction time task, Experiment 2.

TABLE II Aggressive Behavior (Average Noise Punishment Settings) as a Function of Video Game, Provocation Pattern, and Time of Aggression Video game

Provocation pattern

n

Trial 1

Block 1

Block 2

Block 3

Violent

Ambiguous

44

5.93

5.64

5.52

5.66

Nonviolent

Ambiguous

46

4.86

5.55

5.48

5.47

—

1.07

0.08

0.04

0.19

Video game eVect Violent

Increasing

46

3.87

5.16

5.75

5.57

Nonviolent

Increasing

48

3.98

4.96

5.38

5.81

—

0.11

0.20

0.38

0.24

Video game eVect

To more thoroughly examine the video game eVects in the context of our main predictions as well as in context of the three-way interaction, we conducted 2 (video game)  2 (provocation pattern) ANOVAs with TH as a covariate on each measure of aggressive behavior separately, along with two planned contrasts testing separately the video game eVect in ambiguous and increasing provocation pattern conditions. Table II presents the adjusted means. Figure 5 plots the mean diVerences between the violent and nonviolent video game conditions. As can be seen, the pattern was essentially as predicted. On Trial 1, there were significant main eVects of provocation pattern and trait hostility, and a marginally significant video game  provocation pattern interaction

222

CRAIG A. ANDERSON et al.

Fig. 5. Violent video game eVect on noise punishment settings (violent–nonviolent game conditions) as a function of provocation pattern and time of aggression, controlling for trait hostility, experiment 2.

[Fs(1, 179) ¼ 22.35, 4.03, and 3.56, Ps < .001, .05, & .07, respectively]. The ambiguous provocation pattern led to more aggression than the increasing pattern (Ms ¼ 5.39 and 3.92, d ¼ 0.70). Trait hostility was positively related to aggression (b ¼ 0.35). The specific planned contrasts testing the video game eVect separately for the ambiguous and increasing provocation pattern conditions revealed that those who played the violent video game and who had received the ambiguous provocation pattern delivered significantly higher noise punishments to their opponents than did those in the corresponding nonviolent game condition [F(1, 179) ¼ 5.72, P < .02, d ¼ 0.5]. Interestingly, those who received the increasing provocation pattern were unaVected by the video games (F < 1, d ¼ .05). In Block 1 (Trials 2–9) only the provocation and TH main eVects were even close to significant. Those in ambiguous provocation conditions delivered greater punishments to their opponents than those in the increasing provocation conditions [Ms ¼ 5.60 and 5.06, respectively; F(1, 179) ¼ 7.02, P < .01, d ¼ 0.39]. Trait hostility was again positively related to aggression [b ¼ 0.29, F(1, 179) ¼ 6.67, P < .02]. The video game eVect was not significant under either provocation condition (Fs < 1). In Blocks 2 and 3, none of the eVects of video game, provocation, or TH were significant (Ps > .10). 2. Supplementary Analyses. Table III presents the correlations among questionnaire variables. The most interesting findings involved the measure of revenge motivation. Revenge motivation correlated positively with feeling angry during the CRT task, instrumental aggressive motivation, and

223

VIOLENT VIDEO GAMES TABLE III Correlations Among Questionnaire Variables

Variable Trait hostility Revenge motivation Video game experience Hours per week

Revenge motivation

Video game experience

Hours per week

Angry

Instrumental aggressive motivation

.25*

.02

.05

.06

.09

—

.15{

.08

.44*

.26*

—

.46*

.05

.15{

—

.06

.12

—

.07

Angry Hours per week indicates hours per week spent playing video games. *P < .001. NS range from 174 to 184. y P < .01. { P < .05.

experience with the 12 video games included in the questionnaire. To further explore the latter relationship we split the video game experience scale into two subscales and correlated each with revenge and instrumental aggressive motivation. Experience with the six violent video games correlated significantly with revenge motivation (r ¼ .21, P < .01), and with instrumental aggressive motivation (r ¼ .15, P < .05). Experience with the six nonviolent video games did not correlate significantly with either type of motivation (Ps > .5). Revenge motivation also correlated positively with TH, which was measured several weeks prior to the laboratory session. This suggests that the TH eVect on aggressive behavior may have been mediated by revenge motivation. To test this we ran the same regression ANOVAs on the four measures of aggression but with revenge motivation as a covariate. The TH eVect disappeared. In the repeated measures ANOVA the main eVect of revenge motivation was highly significant [F(1, 178) ¼ 52.08, P < .001], but the TH eVect did not approach significance (F < 1). Similarly, on both of the individual aggression measures that had previously yielded significant eVects of TH (Trial 1, Block 1) the revenge motivation eVect was quite strong [Fs(1, 178) ¼ 26.83 and 35.52, bs ¼ 0.89 & 0.65, respectively; Ps < .001], and the trait hostility eVects became nonsignificant (Ps > .2). These results suggest that trait hostility influenced aggression through its eVect on revenge motivation. Interestingly, revenge motivation was not aVected by the video game manipulation (F < 1). Furthermore, the violent video game eVect seen on Trial 1 aggression in the ambiguous provocation condition was not diminished by the inclusion of revenge motivation in the statistical model [F(1, 178) ¼ 6.96, p < .01]. The diVerence in adjusted means between the violent and nonviolent video game conditions was essentially the same when

224

CRAIG A. ANDERSON et al.

revenge motivation was in the model (Ms ¼ 5.89 vs. 4.80) as when it was not in the model (Ms ¼ 5.93 vs. 4.86). Thus, revenge motivation did not mediate the video game eVect on aggression.

C. DISCUSSION This experiment has four main findings. First and foremost, playing a violent video game for 20 minutes led to significantly more aggression than did playing a nonviolent game, in exactly the circumstances expected—on the first retaliation opportunity after an ambiguous pattern of provocations. This video game eVect on aggression was specifically due to violent content; both the aVective and arousal routes were controlled by the selection of games that do not diVer on these two dimensions. Second, TH was positively related to aggressive behavior in the CRT task. Third, this eVect was mediated by revenge motivation. Finally, the positive correlations between past experience with violent video games and both of the aggressive motive measures (revengeful and instrumental) suggest that repeated exposure to violent video games might increase the likelihood that minor provocations will elicit revengeful and instrumentally aggressive responses. It is important to note that the violent video game eVect occurred only on the first trial in the ambiguous provocation condition. This is the most critical set of circumstances for testing the violent content hypothesis in the two-phase CRT. The lack of video game eVects in the increasing provocation condition confirms earlier reports that in the two-phase CRT the ambiguous provocation pattern results in more sensitive tests of other eVects than does the increasing provocation pattern (Anderson et al., 2000; Lindsay & Anderson, 2000). Therefore, the lack of a video game eVect in the increasing provocation pattern condition should not be seen as a disconfirmation of the violent content hypothesis. Nonetheless, additional tests of the violent content hypothesis are needed. Experiment 3 was conducted in part to provide such a test.

VII. Experiment 3 One major goal was conceptual replication. We made several changes to maximize the gain from this replication test. First, we added two modified games (one violent and one nonviolent) that have not been used in previous studies, in addition to the two games used in Experiment 2 (Wells & Windschitl, 1999). Second, we used the standard one-phase version of the CRT task in which the participant and the ‘‘opponent’’ set punishment levels for each other

VIOLENT VIDEO GAMES

225

on the same set of 25 trials. In this paradigm, aggressive behavior emitted during all 25 trials is relevant, unlike the later trials in the two-phase version. Two additional changes allowed for a reasonable correlational test of the hypothesis that exposure to media violence is positively associated with aggressive behavior. We used the physical aggression subscale of the Buss-Perry Aggression Questionnaire (Buss & Perry, 1992) as a measure of ‘‘trait aggressiveness.’’ We also created an overall media violence exposure variable by combining a shortened version of the violent video games measure used by Anderson and Dill (2000) with a measure of exposure to TV violence. Finally, Experiment 3 provided a first attempt to examine the eVects of varying the realism or the ‘‘humanness’’ of the targets of aggression within a game. It has often been assumed that cartoon-like characters in violent video games (as well as television and movies) would elicit less aggression than human characters, especially among participants older than 10 or 12 years (i.e., participants who can clearly distinguish between the fantasy world of cartoons and the real world). Half of the violent game participants played the original version of Marathon 2, in which the targets are humanoid aliens with green blood. The other half played the same game except that the targets were given a human appearance and spouted red blood when shot. A. METHOD 1. Participants Two hundred fourteen undergraduate female and male students enrolled in an introductory psychology class participated for course credit. Participants were randomly assigned to one of four conditions. Two of the conditions involved playing a violent video game; the other two involved playing a nonviolent video game. Ten participants were dropped for a variety of reasons, including failure to complete the materials, high confusion about how to play the games, declining to play the assigned video game, and giving impossible answers to some questionnaire items. For instance, one participant reported watching 70 hours of television per week while taking a full course load. Thus, the total final sample size consisted of 134 females and 70 males. The ratio of female/male participants was approximately equal across experimental conditions. 2. Materials & Design a. Video Games. All video games were played on G3 Power Macintosh computers. The CRT task was conducted with LCIII Macintosh computers. Two of the games were essentially the same as those played in Experiment

226

CRAIG A. ANDERSON et al.

2 (the green-blooded alien version of Marathon 2, labeled ‘‘Alien’’ hereafter in this article, and Glider Pro). One minor change was made in the Alien version; the human allies that appeared in the opening scene were removed. The other two games were major modifications of Marathon 2. The violent modified version (labeled ‘‘Human’’) was identical to Alien except that human figures with red blood replaced the green-blooded aliens. The nonviolent version consisted of the same ‘‘world’’ as Marathon 2, but there were no enemies at which to shoot, and the player’s main task was to explore the world, find replacement oxygen cylinders, and eventually find the transporter in order to return to the ship. This game had a time element built into it, such that the player could ‘‘die’’ if he or she failed to find suYcient oxygen supplies while searching for the transporter. This was labelled the ‘‘Explore’’ condition.5 b. Aggression Paradigm. After playing the assigned video game, the participants performed a standard version of the CRT task. Participants were led to believe that they were competing against another college student, and that they would be setting punishment levels for each other prior to each trial. They were to click the mouse button as quickly as possible after hearing a tone. Participants were told, ‘‘Once you both have hit the mouse button and the winner is determined, the computer will display a sign that says either ‘YOU WON!’ or ‘YOU LOST!’ If you lose the trial, you will receive the noise your partner chose for you.’’ There were 25 such trials, the outcomes of which were actually controlled by the computer. The ambiguous win/loss and punishment patterns (described in Experiment 2) were used for all participants. The main measure of aggression was the average noise intensity level set by the participant. c. Questionnaire. A questionnaire was administered at the end of the CRT task. It contained the following measures that were used in Experiment 2: revenge motivation (coeYcient alpha ¼ .85), instrumental aggressive motivation (two items r ¼ .40), and angry feelings. Trait hostility was replaced by the nine-item Physical Aggression subscale of the Buss-Perry Aggression Questionnaire (coeYcient alpha ¼ .84). The format and style of the items is very similar to the TH scale used in Experiment 2, with scores ranging from 1 to 5. Sample items are ‘‘There are people who pushed me so far that we came to blows’’ and ‘‘I get into fights a little more than the 5 We also modified the opening scenes of the Alien and Human versions to make it somewhat easier, so that novice game players didn’t ‘‘die’’ right away. We thank Brian C. Anderson for making these various changes to the Marathon 2 games. Note that to maximize the power of the comparisons of the two violent game conditions (Alien vs. Human), we set up our randomization schedule so that there would be about twice as many participants in the violent conditions as in the nonviolent conditions. The final sample sizes were: Alien ¼ 70; Human ¼ 62; Glider Pro ¼ 36; Explore ¼ 36.

VIOLENT VIDEO GAMES

227

average person.’’ Although previous research has shown that this scale loads on the same general factor as the TH scale used in Experiment 2 (Dill et al., 1997), the advantage of the Buss-Perry physical aggression subscale is that all nine items involve aggressive behavior, whereas the TH scale used in Experiment 2 includes several diVerent types of items. Video game experience was replaced with a composite measure of exposure to video game and television violence. To assess exposure to violent video games, we used a shortened version of the measure reported in Anderson and Dill (2000). Participants listed up to three favorite video games, indicated how often they have played each in recent years, and rated the level of violence in the games. The video game violence exposure measure is created by multiplying the violence rating for each game by the amount of time spent playing that game, and averaging across the three games. Participants also estimated how many hours per week they watch television, and what proportion of time they watch violent television shows. The television violence measure is simply the product of those two estimates. To create the media violence exposure composite, we standardized the video game and television violence scores and summed them. d. Design. The overall design is thus a 2 (Sex: male vs. female)  2 (Content: violent vs. nonviolent)  2 (Game version: Old (Alien & Glider Pro) vs. New (Human & Explore). The Physical Aggression scale was used as a covariate.6 3. Procedure Participants were escorted into the laboratory on arrival. After being seated by the video game equipment, participants read and signed a consent form. The cover story explained that the study involved examining how people learn simple and complex computer tasks. They were also told that they would be playing two games, one a single-player game and the other a competitive two-person game. Each participant was given an explanation of how the controller worked and how to play the single-player game, followed by instructions on how to play the CRT task. They were then told that they had been randomly assigned to play the single-player game first and were instructed to play the designated game for 20 minutes. After playing for 20 6 This scale was not administered at the beginning of the session because we believed that doing so would increase suspicion about the true purpose of the study. To check on the possibility that the experimental manipulations might have systematically influenced scores on this ‘‘trait’’ measure, we conducted a 2 (Content: violent vs. nonviolent)  2 [Game version: Old (Alien & Glider Pro) vs. New (Human & Explore)] ANOVA on the trait physical aggression scores. None of the eVects approached significance [all Fs < 0.15, all Ps > .70]. Thus, using this variable as a trait physical aggression covariate seemed appropriate.

228

CRAIG A. ANDERSON et al.

minutes, the participant completed the CRT task. Participants were told that they would not meet, see, or learn who their opponent was, but that their opponent was the same sex. After completion of the CRT task, a questionnaire was administered to assess a number of aVective and motivational variables, as well as suspicion. During the final debriefing, the experimenter probed for suspicion, explained all procedures, answered any questions, and thanked the participant.7

B. RESULTS 1. Aggressive Behavior The main measure of aggressive behavior in this experiment was the average noise intensity level that participants set as punishments for their opponent, averaged across all 25 trials of the CRT task. Both the trait physical aggression eVect and the main eVect of video game violent content were significant, replicating the main findings of Experiment 2. Trait physical aggression was positively associated with aggression [F(1, 195) ¼ 15.32, P < .001, b ¼ .42]. More importantly, as shown in the left two columns of Fig. 6, participants who had played a violent video game set higher punishment levels than those who had played a nonviolent game [F(1, 195) ¼ 7.17, P < .01, d ¼ .38, Ms ¼ 5.41 and 4.83, respectively]. None of the other main eVects or interactions were significant [all Fs(1, 195) < 3.20, all Ps > .07].

7 As in Experiment 2, the final questionnaire included items asking participants about the true purpose of the study. In addition, the Experimenter conducted a structured interview designed to detect suspicion as well as to ease into the debriefing. The second author later examined the participants’ written comments and the Experimenters’ notes and rated each participant on a four-point suspicion scale (0–3). Twenty-two participants indicated that they knew (or strongly believed) that the study was about video games and aggression. These participants were excluded from all analyses. During the year in which this experiment was conducted, a number of stories about violent video games appaeared in national, regional, and student newspapers and in a wide array of electronic media, which may account for the increased suspicion rate. Unlike Experiment 2, in which only a few participants were so suspicious and in which suspicion was unrelated to the key dependent variables, in the present study suspicion was highly related to the key measure of aggression. Specifically, the 22 highly suspicious participants produced significantly lower levels of aggression than did the remaining participants [F(1, 238) ¼ 26.47, P < .0001, d ¼ .67]. This finding is similar to other research on suspicion in aggression studies, in which suspicious participants tend to be relatively unaVected by experimental manipulations and tend to display low levels of aggression, presumedly as a result of their deciding to not display aggressive inclinations (Berkowitz & Donnerstein, 1982; Carlson, Marcus-Newhall, & Miller, 1990; Kruglanski, 1975). Fortunately, there was no systematic relation between condition of the study and suspicion level [F(7, 238) ¼ 1.04, P > :40].

VIOLENT VIDEO GAMES

229

Fig. 6. Covariate adjusted average noise intensity settings as a function of video game violence, without (unmediated) and with revenge motivation as a mediating variable, Experiment 3.

2. Supplementary Analyses a. Correlations. Table IV presents the correlations among the key questionnaire measures. The pattern of correlations was very similar to that found in Experiment 2. Revenge motivation was positively correlated with feeling angry, instrumental aggressive motivation, and trait physical aggression. The latter suggests that the trait physical aggression eVect on aggressive behavior reported earlier may have been mediated by revenge motivation, similar to the TH eVect in Experiment 2. We return to this idea shortly. One of the most interesting correlational findings concerned the relation between media violence exposure and the trait physical aggression measure. As shown in Table 4, the correlation was positive, statistically significant, and in the small to medium range. Trait physical aggression also correlated positively with the number of hours spent with electronic entertainment in general. To further test the hypothesis that violent content is most important in this relation, rather than simply number of hours spent on electronic entertainment, we conducted a simple regression analysis that included media violence exposure and number of hours per week spent on video games and television as predictors of trait physical aggression. The media violence measure remained a significant predictor [F(1, 200) ¼ 18.79, P < .001, b ¼ .43]. When sex was also added to the model, the media violence eVect remained significant [F(1, 199) ¼ 6.68, P < .02, b ¼ .27].

230

CRAIG A. ANDERSON et al. TABLE IV Correlations Among Key Questionnaire Variables

Variable Trait physical aggression

Revenge motivation

Hours per week

Angry

Instrumental aggressive motivation

Media violence

.39*

.23*

.13

.10

.36*

Revenge motivation

—

.05

.48*

.49*

.07

Hours per week

—

—

.02

.55*

Angry

—

—

—

.18y

.00

Instrumental aggressive motivation

—

—

—

—

.03

.04

Hours per week indicates hours per week spent playing video games and watching television. *P < .001. NS range from 203 to 204. y P < .01. { P < .05.

b. Mediation by Revenge. Revenge motivation was further analyzed with a 2 (Sex: male vs. female)  2 (Content: violent vs. nonviolent)  2 (Game version: Old (Alien and Glider Pro) vs. New (Human and Explore) ANCOVA, with trait physical aggression as a covariate. There was a significant main eVect of violent content on revenge motivation [F(1, 195) ¼ 8.24, P < .01, d ¼ .41]. Participants in the violent game conditions reported higher levels of revenge motives than did those in the nonviolent game conditions [Ms ¼ 1.97 and 1.63, respectively]. There was also a significant eVect of trait physical aggression [b ¼ .43, F(1, 195) ¼ 31.83, P < .001]. This suggests that both the trait physical aggression eVect and the video game eVect on aggression may have been mediated by revenge motivation. To test the mediation hypothesis, we added revenge motivation to the ANCOVA model and examined the change in aggression score variance accounted for by trait physical aggression and by the video game violence manipulation. In both cases, the proportion of unique variance attributed to the predictor dropped significantly. Revenge motivation accounted for 77% of the variance attributed to the trait physical aggression [F(1, 194) ¼ 13.27, P < .001]. Nonetheless, the unique variance accounted for by trait physical aggression still remained statistically significant [F(1, 194) ¼ 3.92, P < .05]. Thus, it appeared that revenge motivation mediated most but not all of the trait physical aggression eVect on aggression. Revenge motivation accounted for 61% of the the violent video game variance [F(1, 194) ¼ 4.89, P < .05]. However, when revenge motivation was partialled out, the video game eVect was still marginally significant [F(1, 194) ¼ 3.16, P < .08]. Revenge motivation appeared to mediate much

VIOLENT VIDEO GAMES

231

but not all of the video game eVect on aggression. The two right columns of Fig. 6 show the video game violence eVect after revenge motivation is partialled out of the adjusted means. c. Alien vs. Human Targets. The two violent versions of Marathon 2 used in this experiment diVered only in whether the enemy targets were green-blooded aliens or red-blooded humans. Supplemental analyses were performed on these two violent game conditions. The first set involved ratings of the games provided by participants on the standard dimensions used to select games in Experiment 1: diYculty, enjoyment, violence of graphics and content (averaged), frustration, and action. The 2 (sex)  2 (game) ANCOVAs with trait physical aggression as a covariate yielded no significant game main or interaction eVects. Indeed, the only reliable eVects to emerge from these analyses were main eVects of sex on the enjoyment, violence, and frustration ratings [Fs(1, 127) ¼ 6.78, 8.94, and 4.06, respectively; all Ps < .05]. Males enjoyed these two violent games more than females (Ms ¼ 3.88 and 3.09). They also rated the games as less violent (Ms ¼ 4.03 and 4.87) and less frustrating (Ms ¼ 3.76 and 4.38) than females. In sum, the two violent games were essentially equal on these game dimensions, although men and women viewed them somewhat diVerently. The most important question concerning the two violent game conditions is whether the substitution of human characters with red blood increased aggression in this short-term context. It did not [F(1, 127) ¼ 0.36, P > .50], although the means were slightly in that direction (Ms ¼ 5.49 and 5.33 for the Human and Alien conditions, respectively).

C. DISCUSSION Experiment 3 replicated the main findings of Experiment 2 using two similar violent games (the Alien and Human versions of Marathon 2) versus two nonviolent games (Glider Pro and the exploration version of Marathon 2) and did so with the more standard version of the CRT task. This replication strengthens the hypothesis that violent content in a video game can increase aggressive behavior after a minor provocation, even when the arousal and aVective sequela are controlled. Furthermore, the positive association between trait physical aggression and CRT aggression further validates this noise–aggression paradigm. Finally, the trait physical aggression eVect on aggressive behavior was partially mediated by revenge motivation, similar to the TH eVect in Experiment 2. Experiment 3 also found essentially equivalent eVects of the two violent versions of Marathon 2—the green-blooded alien or red-blooded human

232

CRAIG A. ANDERSON et al.

enemies. This strongly contradicts conventional wisdom and some research on television violence, both of which suggest that more realistic violence would have a relatively greater impact. Of course, failure to find a statistically significant eVect is not suYcient to claim that the eVect does not exist. Two alternative explanations for the lack of a ‘‘realism’’ eVect are quite plausible. First, it may be that the realism eVect occurs in repeated exposure long-term contexts but not in a short-term context. This would be true if the realism eVect operates as a kind of systematic desensitization eVect (Carnagey, Bushman, & Anderson, under review). Second, the high realism condition in Experiment 3 (i.e., the Human-red blood version of Marathon 2) may not have been suYciently realistic. Further work (not speculation) is needed on the realism eVect. Finally, Experiment 3 found additional correlational support for the link between exposure to media violence and general level of trait physical aggression. By themselves, such correlational data are not conclusive regarding causality, but they lend support for the hypothesis that repeated exposure to violent media leads to relatively high levels of trait aggressiveness.

VIII. Correlation Study 1 A. OVERVIEW The primary purpose was to test the hypothesis that repeated exposure to violent video games would be positively associated with aggressive behavior and persistent aggressive cognitions (e.g., attitudes toward aggression), and that these persistent aggressive cognitions would at least partially mediate the violent video game exposure–aggression link. A secondary purpose was to examine associations between violent video game exposure and several key personality indicators. Specifically, we included Goldberg’s Big Five scales, the Narcissism scale, and Caprara’s Emotional Susceptibility scale. If repeated exposure to violent video games does influence basic personality development, then associations should be found between such exposure and personality indicators known to be related to habitual aggressive behavior tendencies, such as Goldberg’s agreeableness and conscientiousness factors, narcissism, and emotional susceptibility. Of course, such associations could reflect the opposite causal direction. That is, basic personality factors favoring aggressiveness might well predispose some individuals to enjoying and playing violent video games. A cross-sectional correlation study of this type cannot, by itself, conclusively rule out this alternative. However, this alternative further implies that any significant correlations between violent video

VIOLENT VIDEO GAMES

233

game exposure and aggression should become nonsignificant when basic personality factors are statistically controlled. In sum, strong support for the hypothesis that repeated exposure to violent video games increases aggressive behavior tendencies and does so through changes in persistent aggressive cognitions would consist of three parts. First, there must be significant associations between violent video game exposure and aggressive behavior tendencies. Second, these associations should persist even when basic personality factors are statistically controlled. Third, these associations should be substantially reduced when persistent aggressive cognitions are statistically controlled. Adequate testing of these ideas requires a large sample size, so that lack of statistical power does not lead to type I inferential errors.

B. METHOD 1. Participants Three hundred seventeen male and 489 female college students at a large midwestern university participated in a large mass testing questionnaire session for partial credit toward introductory psychology course requirements.

2. Measures a. Video Game Violence Exposure. Video game violence exposure (VGV) was assessed using the same shortened version of the Anderson and Dill (2000) measure described earlier in Experiment 3. Across the three video game items, coeYcient alpha was .83, only slightly lower than the .86 alpha reported by Anderson and Dill (2000) for their five-game version of this measure. b. Basic Personality. We measured seven basic personality factors, the Big Five as well as two more specific factors that have been theoretically and empirically related to aggression (narcissism and emotional susceptibility). Goldberg’s (1992) Big Five measure of basic personality structure consists of 20 items for each factor. CoeYcient alphas were: surgency (extraversion) ¼ .87; agreeableness ¼ .91; conscientiousness ¼ .89; emotional stability (neuroticism) ¼ .80; intellect (openness) ¼ .85. The 40-item Narcissism scale (Raskin & Terry, 1988) yielded an alpha of .84. The 27-item Emotional Susceptibility scale (Caprara et al., 1985) yielded an alpha of .91. c. Attitudes towards Violence. Two measures of attitudes toward violence were used. One was the recently revised 39-item Attitudes towards Violence scale (ATVS; Anderson, Benjamin, Wood, & Bonacci, in press),

234

CRAIG A. ANDERSON et al.

with an alpha of .92. The other was the 15-item Adolescent Attitudes towards Violence scale (AATVS; Funk, Elliott, Urman, Flores, & Mock, 1999), with an alpha of .76. d. Aggressive Behavior. Three measures of trait aggressive behavior were used. The nine-item physical and the five-item verbal aggression subscales of the Aggression Questionnaire (Buss & Perry, 1992) yielded alphas of .85 and .79, respectively. We also used the same standardized 10 physical aggression items used by Anderson and Dill (2000) from the National Youth Survey (Elliot, Huizinga, & Ageton, 1985), which includes behaviors that would be considered criminal if known to police (e.g., assault, robbery). For ease of exposition, the Buss-Perry physical aggression measure will be referred to as mild physical aggression, whereas the National Youth Survey measure will be referred to as severe physical aggression.

C. RESULTS AND DISCUSSION 1. Zero-order Correlations As expected, VGV was positively related to each of the three aggression measures and to both attitudes towards violence measures, as can be seen in Table V. VGV was also negatively associated with the Big Five factors of agreeableness and conscientiousness, as expected. VGV was also slightly positively correlated with the Big Five factor of emotional stability. Finally, VGV was positively correlated with narcissism and negatively correlated with emotional susceptibility. There are a number of additional interesting correlations in Table V, but because they are less central to the main hypotheses we leave them for the reader to discover. 2. VGV–Aggression Link and Basic Personality If the strong correlation between VGV and aggression is due solely to spurious relationships with basic personality factors, such that aggressive people also happen to like violent video games, then statistically controlling for basic personality factors should eliminate the VGV–aggression relationship. We used the destructive testing approach to assess this question (Anderson & Anderson, 1996; Anderson & Dill, 2000). In this approach, one first determines whether a specific predicted relationship exists (e.g., the VGV–aggression correlation). If so, then theoretically meaningful competitor variables (e.g., Big Five) are entered into the regression mode to determine whether these competitors break the target relation. Of primary interest is not whether the initial target link can be broken (i.e., made

235

VIOLENT VIDEO GAMES TABLE V Correlations Among Key Questionnaire Variables Variable MPA VGV MPA

VA

SPA ATVS AATV

.312* .196* .166* .237* —

.438* .315* .478* .131* .209*

B5Su

B5Ag

B5Co

B5ES

.321*

.012

.161*

.121*

.076y

.696*

.012

.319*

.202*

.080y

y

.091

y

B5In

NPI

ES

.029

.159*

.170*

.043

.197*

.070y

VA

—

—

.358*

.180*

.223*

.078

.055

.296*

.068

SPA

—

—

—

.250*

.322*

.003

.187*

.130*

.021

.100y

.112y

.039

ATVS

—

—

—

—

.527*

.097y

.307*

.137*

.009

.220*

.132*

.007

AATV

—

—

—

—

—

.068

.357*

.235*

.026

.108y

.167*

.045

B5Su

—

—

—

—

—

—

.320*

.198*

.137*

.380*

.499*

.098y

B5Ag

—

—

—

—

—

—

—

.607*

.205*

.489*

.048

.066

B5Co

—

—

—

—

—

—

—

—

.222*

.415*

.087y

.146*

B5ES

—

—

—

—

—

—

—

—

—

.055

.056

.566*

B5In

—

—

—

—

—

—

—

—

—

—

.262*

.050

NPI

—

—

—

—

—

—

—

—

—

—

—

.212*

VGV, Video game violence exposure; MPA, mild physical aggression; VA, verbal aggression; SPA, severe physical aggression; ATVS, Attitudes towards Violence scale; AATV, Adolescent Attitudes towards Violence scale; B5Su, surgency; B5Ag, agreeableness; B5Co, conscientiousness; B5ES, emotional stability; B5In, intellect; NPI, narcissism; ES, emotional susceptibility. *P < .001. y P < .05. n ¼ 806.

nonsignificant), because even strong causal links between measured variables can eventually be broken by adding more correlated competitors into the model. Instead, the focus is the durability of the link, with consideration given to the theoretical and empirical strength of the competitor variables. Table VI presents the results of this destructive testing procedure, displaying the raw slopes linking VGV to each of the three aggression measures, when only VGV is in the model (first column of slopes) and when other competitor variables are partialled out (columns 2–5). We first added sex to the model. The slopes decreased in magnitude but remained statistically significant.8 In turn, we added the Big Five factors, then the narcissism measure, and finally the emotional susceptibility measure. As shown in Table VI, the VGV–aggression link survived all of these competitor variables. These

8

Because sex is highly correlated with exposure to violent video games, partialling out sex eVects likely overcorrects for gender diVerences. Thus, the resulting estimate of the VGV– aggression relation is an extremely conservative one.

236

CRAIG A. ANDERSON et al. TABLE VI Destructive Testing Results on the VGV/Aggression Relation VGV slopes with various predictor variables in the model

Aggression measure

VGV

þSex

þBig 5

þNPI

þES

Mild physical aggression

.0286*

.0098y

.0082y

.0078{

.0081y

Verbal aggression

.0182*

.0098y

.0075{

.0068{

.0071{

Severe physical aggression

.0097*

{

.0058

{

.0056

{

.0053

.0055{

df for F test of VGV eVect

1, 804

1, 803

1, 798

1, 797

1, 796

VGV, Video game violence exposure; Big Five, five basic personality factors; NPI, narcissism personality inventory; ES, emotional susceptibility. *P < .001. y P < .01. { P < .05. n ¼ 806.

results support the view that the VGV–aggression link is not a spurious artifact of basic personality structure. 3. Persistent Aggressive Thoughts as a Mediator Our next analysis used the same destructive testing approach, but here the meaning is considerably diVerent because the ‘‘competitor’’ variables are potential mediators of the long-term VGV eVect on aggressive tendencies. Our application of GAM to the video game domain explicitly states that the content of violent video games can create long-term changes in a host of aggression-related knowledge structures, many of which can be indexed by measures of general attitudes toward violence. Thus, we expected that adding such attitude measures to the statistical model would result in substantial reductions in the VGV–aggression relation. Once again, we began with a model having only VGV as the predictor, and then added sex to the model (see columns 1 and 2 of Table VI). Next, we added both attitudes toward violence measures to the model (instead of the personality factors). As anticipated, the VGV aggression slope decreased substantially for each of the three aggression measures. For mild physical aggression the VGV slope dropped from a significant .0098 (P < .005) to a nonsignificant .0013 [F(1, 801) ¼ 0.28], an 87% decrease in the slope [(.0098 - .0013)/ .0098]. For verbal aggression the VGV slope dropped from a significant .0098 (P < .01) to a nonsignificant .0055 [F(1, 801) ¼ 2.44], a 44% decrease. For severe physical aggression the VGV slope dropped from a significant .0058 (P < .02) to a nonsignificant .0031 [F (1, 801)1.96], a 47% decrease.

VIOLENT VIDEO GAMES

237

In sum, these results support the hypothesis that long-term eVects of repeated exposure to violent video games on aggressive behavior tendencies are at least partially mediated by changes in persistent aggressive cognitions. Of course, as noted earlier the correlational nature of these results warrants some interpretative caution and further integrative research.

IX. Updated Meta-Analysis A number of new studies have become available since Anderson and Bushman’s (2001) meta-analysis of violent video game eVects. We added all of the new studies we could locate to our database and made a few procedural changes to address two specific questions. One question concerned the possibility (frequently oVered by media representatives and other critics) that the average eVect sizes reported in prior meta-analyses might be inflated by the inclusion of studies with potentially important methodological weaknesses. In the present meta-analytic study, we identified nine methodological weaknesses found in at least some violent video game studies, and we categorized each sample as having none of them (the ‘‘Best Practices’’ studies) or at least one (‘‘Not Best Practices’’ studies). We compared the average eVect size results of this distinction. The second question concerned potential diVerences between experimental studies (which allow stronger causal statements on the basis of even a few studies) and correlational studies (which typically involve more serious and realistic forms of aggression, and address the long-term eVect issue).

A. METHODS 1. Study Sample A complete list of included studies can be found at the following web page: http://www.psychology.iastate.edu/faculty/caa/abstracts/2000–2004/04AESPref.pdf. We included all studies that we could locate that had data testing a possible link between exposure to violent video games and one of five types of outcome variables: aggressive behavior (defined as behavior intended to harm another person), aggressive cognition, aggressive aVect, helping behavior, and physiological arousal. A given ‘‘study’’ might contain more than one independent ‘‘sample’’ of research participants. For example, some studies reported results separately for male and female participants. We used one eVect size for each sample for each of the available five dependent

238

CRAIG A. ANDERSON et al.

variables. For example, if a sample had three diVerent (and valid) measures of aggressive behavior and a composite of the three, we used the composite measure. If a composite could not be obtained, we used the average of the three separate eVect sizes.

2. Best Practices Coding The following potential methodological problems were examined for each sample: 1. Nonviolent video game condition contained violence, and there was no suitable nonviolent control condition. 2. Violent video game condition contained little or no violence. 3. In a correlational study, the measure of video game exposure was not specifically tied to violent video games (e.g., the amount of time spent on any kind of video game was measured instead of time spent on violent video games). 4. Evidence that the violent and nonviolent conditions diVered significantly in ways that could contaminate the conditions, such as the nonviolent condition being more diYcult, boring, or frustrating than the violent condition. 5. A pre-post design was used, but only the average of the pre- and postmanipulation measures was reported. 6. Each research session involved both a video game player and an observer, but only the average of the player-observer measures was reported. 7. The aggressive behavior measure was not aggression against another person (e.g., aggression against a non-human character, or against objects). 8. The outcome variable was physiological arousal, but arousal diVerences between the violent and nonviolent video game conditions were already controlled by pretesting, game selection, or both (i.e., equally arousing violent and nonviolent games were intentionally chosen by the researchers to control for potential arousal eVects on other outcome measures such as aggressive behavior). 9. The outcome variable was aggressive aVect, but aVective diVerences between the violent and nonviolent video game conditions were already controlled by pretesting, game selection, or both (i.e., violent and nonviolent games were intentionally chosen by the researchers to have the same aVective impact, to control for potential aVective influences on other outcome measures such as aggressive behavior). Some of these ‘‘weaknesses’’ are actually strengths for other aspects of the same research. For example, if one wants to study whether violent video game

VIOLENT VIDEO GAMES

239

content (relative to a nonviolent video game) can increase aggressive behavior even when there are no arousal diVerences between the games, pretesting and selecting violent and nonviolent video games that produce equivalent levels of arousal is an excellent methodological feature (as in our Experiments 1–3). However, that same sample does not allow a good test of whether violent video games on average increase arousal. Thus, for aggressive behavior this sample would be coded as a ‘‘best practice’’ one, whereas it would be coded as a ‘‘not best practice’’ sample for physiological arousal. For several samples it was possible to get eVect sizes for both a best practices procedure and a not best practices procedure on the same outcome variable. For example, several correlational studies reported both a best practices measure of time spent on violent video games and a not best practices measure of time spent on any type of video game. B. RESULTS AND DISCUSSION 1. Best vs. Not Best Practices Figure 7 presents the best vs. not best practices results. In each case the methodologically best samples yielded average eVect sizes that were larger than methodologically weaker samples. This was especially pronounced for aggressive behavior and aggressive aVect, wherein the 95% confidence intervals for the best and not best samples did not overlap. These results suggest that eVect size estimates that include methodologically weaker studies (e.g., Anderson & Bushman, 2001) underestimate the true eVect sizes of exposure to violent video games. 2. Best Practices Samples: Experimental vs. Correlational Figure 8 presents the average eVect sizes of the best practices samples categorized by type of study. There are no consistent diVerences in eVect sizes for the experimental versus correlational samples. Correlational studies yielded a slightly larger average eVect on aggressive and helping behavior than did experimental studies, whereas the opposite was true for aggressive cognition and aVect. In all four of these cases the experimental and correlational 95% confidence intervals overlap. Furthermore, both the experimental and correlational eVect sizes were significantly diVerent from zero for each outcome variable except physiological arousal, and that was because there are no best practices correlational studies of this type. In sum, despite the relatively small size of this research domain, there is considerable correlational and experimental evidence linking exposure to violent video games with increases in aggressive behavior and to several aggression-related variables.

240

CRAIG A. ANDERSON et al.

Fig. 7. EVects of violent video games on aggressive behavior, aggressive cognition, aggressive aVect, helping behavior, and physiological arousal by best practices methodology. rþ, average eVect size; K, number of independent samples; N, total number of participants. Vertical capped bars are the upper and lower 95% confidence intervals.

X. General Discussion These five studies (three experiments, one correlational study, one metaanalytic study) contribute to our understanding of human aggression from both a personality processes and a situational eVects perspective. We believe these results can best be understood within the GAM theoretical framework described earlier. A. SITUATIONAL EFFECTS The main situational finding was that brief exposure to violent video games increased aggressive behavior relative to nonviolent video games matched on arousal and aVective dimensions. This occurred in both experiments that measured aggression, regardless of whether the violent game had

VIOLENT VIDEO GAMES

241

Fig. 8. EVects of violent video games on aggressive behavior, aggressive cognition, aggressive aVect, helping behavior, and physiological arousal by type of study for best practices samples. r+, average eVect size; K, number of independent samples; N, total number of participants. Vertical capped bars are the upper and lower 95% confidence intervals.

green-blooded aliens or red-blooded humans as the enemy targets, and occurred in both versions of the CRT task. The lack of reliable sex X game violence interactions suggests that the eVect was similar in men and women. A second situational finding concerns the cognitive eVects of violent video games. Experiment 1 demonstrated that violent games in general produce increases in the relative accessibility of aggressive thoughts. The present findings make a very strong case for the hypothesis that violent video games can (and do) cause increases in aggression because of the violent content of such games, not just because of their arousal or aVective properties. The present empirical results in combination with our theoretical analysis also lend support to the concern that repeated exposure to violent video games (or other violent media) might lead to development of an increasingly aggressive personality, and that much of this developmental eVect may be the direct result of the violent content. In short, repeatedly thinking

242

CRAIG A. ANDERSON et al.

about violent characters, choosing to be aggressive, enacting that aggressive choice, and being rewarded for it can be conceived as a series of learning trials influencing a variety of types of aggressive knowledge structures. Violent video games may well teach players to become more aggressive people. One interesting diVerence between the results of Experiments 2 and 3 concerns revenge motivation. There was little evidence that the video game eVect in Experiment 2 was mediated by an increase in desire for revenge, but there was considerable evidence that much of the video game eVect was mediated by revenge motivation in Experiment 3. The various procedural diVerences between the two studies may account for these diVerences, particularly the timing of various measurements in the two-phase versus the standard one-phase CRT task. In any case, it may be useful in future research to explore the possibility that the violent content of violent video games may increase aggression by first priming aggressive cognitions, which in turn increase desire for revenge when mildly provoked.

B. PERSONALITY EFFECTS Experiments 2 and 3 yielded several findings of general interest from a personality perspective. First, trait hostility (Experiment 2) and trait physical aggression (Experiment 3) were positively related to aggression in the CRT task. This further validates these trait measures and the two versions of the CRT task. Second, both of these eVects were largely mediated by revenge motivation. Thus, it appears that one way in which highly hostile people are predisposed to be aggressive against others is through increased revenge motives that are aroused when mildly provoked. We believe that this is the first demonstration that such trait eVects on aggressive behavior operate through increases in desire for revenge. A third finding concerns the question of who is most susceptible to violent video game eVects. There is some evidence from the television/movie violence literature that highly aggressive people tend to be more strongly influenced by exposure to violent media than nonaggressive people (e.g., Bushman & Huesmann, 2001). Such person  situation interactions do not always occur in the media violence literature; sometimes the opposite pattern occurs (e.g., Anderson, 1997). Also, even when this pattern occurs, it is not the case that nonaggressive people are entirely unaVected. A common claim of skeptics and media industry representatives is that only a very few disturbed individuals might be aVected at all. In the present studies neither trait hostility (Experiment 2) nor trait physical aggressiveness (Experiment 3) interacted with the violent video game manipulation. That is, the violent video game eVect on aggression was not reliably bigger (or smaller) for those

VIOLENT VIDEO GAMES

243

participants who scored high on these traits than for those who scored low. Along these same lines, we did not obtain significant sex  video game violence interactions on aggressive behavior. We suspect that with larger sample sizes such person  situation interactions will emerge in some contexts. Nonetheless, the lack of such interactions in the present studies suggests that violent video games influence a sizeable proportion of people. A fourth finding of interest, from Experiment 2, was that experience with violent video games correlated positively (and significantly) with both revenge motivation and instrumental aggressive motivation, whereas experience with nonviolent video games did not. Although Experiment 2 was not explicitly designed to test these relations, it is interesting that this pattern fits exactly what would be expected if repeated exposure to violent video games does create more aggressive individuals. It also fits with prior research designed to test such eVects (e.g., Anderson & Dill, Study 1, 2000). Fifth, the correlational findings of Experiment 3 that general media violence exposure is positively associated with trait physical aggression, even when time spent on electronic entertainment and sex were statistically controlled, support a long and increasingly strong line of research on media violence eVects (e.g., Bushman & Anderson, 2001). The correlational study provided support for the hypothesized link between repeated exposure to violent video games and increased aggressive tendencies, and did so for three types of aggression: mild physical, verbal, and severe physical aggression. This study also provided the first correlational support for the contention that such long-term eVects on aggressive behavior are mediated by persistent aggressive thoughts, here indexed by two diVerent attitudes toward violence measures. Furthermore, this study provided the first correlational evidence that the violent video game exposure link to aggression persists even when a host of basic personality factors are statistically controlled. Finally, the meta-analysis further revealed consistent eVects of violent video game exposure on aggressive behavior, cognition, and aVect, as well as on arousal and prosocial behavior. The fact that these eVects were stronger in the methodologically strongest studies, and occurred in both experimental and correlation designs (with the exception of arousal), lends further support for our application of GAM to both long- and short-term media violence eVects.

C. IMPLICATIONS There are many questions requiring additional empirical work. One is the need to identify specific features of violent video games that increase or decrease their impact on aggressive thoughts, feelings, and behaviors. Experiment 3

244

CRAIG A. ANDERSON et al.

provided one test of the hypothesis that the realism or ‘‘humanness’’ of the game target might exacerbate the eVect. It yielded no support for that hypothesis. Further tests using more extremely realistic and gory graphics are needed. A second question concerns the long-term eVects of repeated exposure to violent video games, especially on children and teens. Based on over 40 years of research on television and movie violence, one reasonable expectation is that repeatedly exposing youth to violent video games over a period of years will have a sizeable negative impact on their development. Indeed, there is reason to believe that the video game violence eVect will be larger than television violence eVects because of the highly engaging and active nature of video games compared with the relatively passive nature of watching TV. Nonetheless, longitudinal research is badly needed to test this prediction and to delineate protective and exacerbating factors. A third set of questions concerns possible positive eVects of games designed to promote prosocial behaviors. Do such games increase prosocial and decrease antisocial behavior? Virtually no research exists on this topic. However, video games are going to remain a major source of entertainment. Therefore we believe it is important to oVer empirical evidence and quality theory on which types of features promote a prosocial gaming experience, as well as highlighting the potential antisocial eVects of games with violent themes.

XI. Appendix A. VIOLENT GAME DESCRIPTIONS  Dark Forces. This is a standard first-person shooter. The player assumes the role of a special ops guy in the Rebellion with the objectives of stealing the Death Star plans and getting out alive. This game has a fairly high level of violence, with weapons like a blaster rifle and laser pistol to kill enemy guards and stormtroopers.  Marathon 2. This is a standard first-person shooter, in which the player assumes the role of a space marine trapped in a base that has been taken over by aliens. Your main goal is to retake it and not die. This game has a high violence level, with the basic, underlying premise of the game being to shoot anything that moves and kill or be killed. The targets are mainly aliens, with some ‘‘compilers’’ or robots that the aliens use as slaves.  Speed Demon. This is a 3D combat driving game. The player drives a heavily armed vehicle in a race with other similarly armed vehicles,

VIOLENT VIDEO GAMES

245

shooting at and crashing into them as they do likewise. One gets points for destroying other vehicles.  Street Fighter. This is a third-person fighting game, similar in many ways to Mortal Kombat. The player chooses a character and then engages in a series of fights with other characters. Each character has specific strengths and weaknesses.  Wolfenstein 3D. This is a first-person shooter. The player assumes the role of B. J. Blascowitz, an American soldier caught and taken prisoner trying to infiltrate a top-secret Nazi experimentation lab. The goal is to shoot your way out of the prison and kill everything that moves. You can pick up various weapons, including various guns. There is a very high violence rate, with you shooting dogs and Nazi guards, with gory bullet hits and ‘‘death poses.’’ B. NONVIOLENT GAME DESCRIPTIONS  3D Ultra Pinball. This is simply an electronic version of a pinball game, complete with flippers, buzzers, bells, and various visual and auditory eVects.  Glider Pro. Players of this game control the forward and backward motion of paper airplanes through a house. By flying over air ducts one can gain lift. The glider can turn various items on and oV (such as light switches, computers). One can earn points by flying over certain objects. If the glider hits the floor or certain other items, it crumples and is replaced by another glider.  Indy Car 2. The player assumes the role of driver in an Indy car race with the goal of winning the race. If you bump other cars or drive too fast on turns, you crash. This game can be played from the keyboard, but is much easier with a steering wheel and pedals.  Jewel Box. This game is a colorful version of Tetris. Variously colored shapes drop from the top of the screen. The player manipulates the objects as they fall, trying to create filled rows. When a row is completely filled, it disappears and the player receives points.  Myst. This is a nonviolent exploration/mystery/adventure game with a first-person perspective. It begins on the Island of Myst. Players begin by appearing on the island with no knowledge of how or why they ended up there, or how to proceed. In an ancient library, players discover two mysterious books, which lay out the basic mystery. Players must then travel to several diVerent worlds (‘‘ages’’) to unravel the mystery.

246

CRAIG A. ANDERSON et al.

Acknowledgments We thank Mary Ballard, Brad Bushman, and Jeanne Funk for their helpful comments on an earlier version of this work.

References Anderson, C. A. (1983). Imagination and expectation: The effect of imagining behavioral scripts on personal intentions. Journal of Personality and Social Psychology, 45, 293–305. Anderson, C. A. (1997). Effects of violent movies and trait irritability on hostile feelings and aggressive thoughts. Aggressive Behavior, 23, 161–178. Anderson, C. A. (2000). Violent video games increase aggression and violence. U.S. Senate Committee on Commerce, Science, and Transportation hearing on ‘‘The Impact of Interactive Violence on Children.’’ Tuesday, March 21, 2000. Available at: http://www.psychology. iastate.edu/faculty/caa/abstracts/1995–1999/96AA.html. Anderson, C. A., & Anderson, K. B. (1996). Violent crime rate studies in philosophical context: A destructive testing approach to heat and southern culture of violence effects. Journal of Personality and Social Psychology, 70, 740–756. Anderson, C. A., Anderson, K. B., Dorr, N., DeNeve, K. M., & Flanagan, M. (2000). Temperature and aggression. In M. Zanna (Ed.), Advances in Experimental Social Psychology (Vol. 32, pp. 63–133). New York: Academic Press. Anderson, C. A., Benjamin, A. J., Wood, P. K., Bonacci, A. M. (in press). Development and testing of the attitudes toward violence scale: Evidence for a four-factor model. Aggressive Behavior. Anderson, C. A., & Bushman, B. J. (1997). External validity of ‘‘trivial’’ experiments: The case of laboratory aggression. Review of General Psychology, 1, 19–41. Anderson, C. A., & Bushman, B. J. (2002). Human aggression. Annual Review of Psychology, 53, 27–51. Anderson, C. A., & Bushman, B. J. (2001). Effects of violent video games on aggressive behavior, aggressive cognition, aggressive affect, physiological arousal, and prosocial behavior: A meta-analytic review of the scientific literature. Psychological Science, 12, 353–359. Anderson, C. A., & Carnagey, N. L. (2004). Violent evil and the general aggression model. In A. Miller (Ed.), The social psychology of good and evil (pp. 168–192). New York: Guilford Publications. Anderson, C. A., Carnagey, N. L., & Eubanks, J. (2003). Exposure to violent media: The effects of songs with violent lyrics on aggressive thoughts and feelings. Journal of Personality and Social Psychology, 84, 960–971. Anderson, C. A., & Dill, K. E. (2000). Video games and aggressive thoughts, feelings, and behavior in the laboratory and in life. Journal of Personality and Social Psychology, 78, 772–790. Anderson, C. A., & Ford, C. M. (1986). Affect of the game player: Short term effects of highly and mildly aggressive video games. Personality and Social Psychology Bulletin, 12, 390–402. Anderson, C. A., & Godfrey, S. (1987). Thoughts about actions: The effects of specificity and availability of imagined behavioral scripts on expectations about oneself and others. Social Cognition, 5, 238–258.

VIOLENT VIDEO GAMES

247

Anderson, C. A., & Huesmann, L. R. (2003). Human aggression: A social–cognitive view. In M. Hogg & J. Cooper (Eds.), Handbook of social psychology (pp. 296–323). London: Sage Publications. Anderson, C. A., Lindsay, J. J., & Bushman, B. J. (1999). Research in the psychological laboratory: Truth or triviality? Current Directions in Psychological Science, 8, 3–9. Bartholow, B. D., & Anderson, C. A. (2002). Examining the effects of violent video games on aggressive behavior: Potential sex differences. Journal of Experimental Social Psychology, 38, 283–290. Berkowitz, L., & Donnerstein, E. (1982). External validity is more than skin deep: Some answers to criticism of laboratory experiments. American Psychologist, 37, 245–257. Buchman, D. D., & Funk, J. B. (1996). Video and computer games in the ’90s: Children’s time commitment and game preference. Children Today, 24, 12–16. Bushman, B. J., & Anderson, C. A. (2001). Media violence and the American public: Scientific facts versus media misinformation. American Psychologist, 56, 477–489. Bushman, B. J., & Anderson, C. A. (2002). Violent video games and hostile expectations: A test of the general aggression model. Personality and Social Psychology Bulletin, 28, 1679–1686. Bushman, B. J., & Huesmann, L. R. (2001). Effects of televised violence on aggression. In D. Singer & J. Singer (Eds.), Handbook of children and the media (pp. 223–254). Thousand Oaks, CA: Sage Publications. Buss, A. H., & Perry, M. (1992). The aggression questionnaire. Journal of Personality and Social Psychology, 63, 452–459. Calvert, S. L., & Tan, S. (1994). Impact of virtual reality on young adults’ physiological arousal and aggressive thoughts: Interaction versus observation. Journal of Applied Developmental Psychology, 15, 125–139. Caprara, G. V., Cinanni, V., D’Imperio, G., Passerini, S., Renzi, P., & Travaglia, G. (1985). Indicators of impulsive aggression: Present status of research on irritability and emotional susceptibility scales. Personality and Individual Differences, 6, 665–674. Carlson, M., Marcus-Newhall, A., & Miller, N. (1989). Evidence for a general construct of aggression. Personality and Social Psychology Bulletin, 15, 377–389. Carlson, M., Marcus-Newhall, A., & Miller, N. (1990). Effects of situational aggression cues: A quantitative review. Journal of Personality and Social Psychology, 58, 622–633. Carnagey, N. L., Bushman, B. J., Anderson, C. A. (under review). Video game violence desensitizes players to real world violence. CIRP (1998, 1999). Cooperative institutional Research Program survey results. Ames, IA: Office of Institutional Research. Crick, N. R., & Dodge, K. A. (1994). A review and reformulation of social informationprocessing mechanisms in children’s social adjustment. Psychological Bulletin, 115, 74–101. Dill, K. E., & Dill, J. C. (1998). Video game violence: A review of the empirical literature. Aggression and Violent Behavior: A Review Journal, 3, 407–428. Dill, K. E., Anderson, C. A., Anderson, K. B., & Deuser, W. E. (1997). Effects of aggressive personality on social expectations and social perceptions. Journal of Research in Personality, 31, 272–292. Elliot, D. S., Huizinga, D., & Ageton, S. S. (1985). Explaining delinquency and drug use. Beverly Hills, CA: Sage. Federal Trade Commission. (2000). Marketing violent entertainment to children: A review of selfregulation and industry practices in the motion picture, music recording, & electronic game industries. Report of the Federal Trade Commission. Available at: www.ftc.gov/reports/ violence/.

248

CRAIG A. ANDERSON et al.

Funk, J. B., Elliott, R., Urman, M. L., Flores, G. T., & Mock, R. M. (1999). The Attitudes Towards Violence scale: A measure for adolescents. Journal of Interpersonal Violence, 14, 1123–1136. Funk, J. B., Flores, G., Buchman, D. D., & Germann, J. N. (1999). Rating electronic games: Violences is in the eye of the beholder. Youth & Society, 30, 283–312. Gentile, D. A., Lynch, P. J., Linder, J. R., Walsh, D. A. (2004). The effects of violent video game habits on adolescent hostility, aggressive behaviors, and school performance. Journal of Adolescence. Giancola, P. R., & Chermack, S. T. (1998). Construct validity of laboratory aggression paradigms: A response to Tedeschi and Quigley (1996). Aggression and Violent Behavior, 3, 237–253. Goldberg, L. R. (1992). The development of markers for the Big-Five factor structure. Psychological Assessment, 4, 26–42. Graybill, D., Strawniak, M., Hunter, T., & O’Leary, M. (1987). Effects of playing versus observing violent versus nonviolent video games on children’s aggression. Psychology: A Quarterly Journal of Human Behavior, 24, 1–8. Hearold, S. (1986). A synthesis of 1043 effects of television on social behavior. In G. Comstock (Ed.), Public communication and behavior (Vol. 1, pp. 65–133). New York: Academic Press. Huesmann, L. R., & Miller, L. S. (1994). Long-term effects of repeated exposure to media violence in childhood. In L. R. Huesmann (Ed.), Aggressive behavior: Current perspectives. New York: Plenum Press. Kirsh, S. J. (1998). Seeing the world through Mortal Kombat-colored glasses: Violent video games and the development of a short-term hostile attribution bias. Childhood, 5, 177–184. Kruglanski, A. W. (1975). The human subject in the psychology experiment: Fact and artifact. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 8, pp. 101–147). New York: Academic Press. Lindsay, J. J., & Anderson, C. A. (2000). From antecedent conditions to violent actions: A general affective aggression model. Personality and Social Psychology Bulletin, 26, 533–547. Marsh, R. L., Hicks, J. L., & Bink, M. L. (1998). Activation of completed, uncompleted, and partially completed intentions. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 350–361. Paik, H., & Comstock, G. (1994). The effects of television violence on antisocial behavior: A meta-analysis. Communication Research, 21, 516–546. Raskin, R., & Terry, H. (1988). A principal-components analysis of the narcissistic personality inventory and further evidence of its construct validation. Journal of Personality and Social Psychology, 54, 890–902. Roediger, H. L., Weldon, M. S., Stadler, M. L., & Riegler, G. L. (1992). Direct comparison of two implicit memory tests: Word fragment and word stem completion. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 1251–1269. Sherry, J. L. (2001). The effects of violent video games on aggression: A meta-analysis. Human Communication Research, 27, 409–431. Tukey, J. W. (1977). Exploratory data analysis. Menlo Park, CA: Addison-Wesley. Walsh, D. (1999). 1999 Video and Computer Game Report Card. Minneapolis: National Institute on Media and the Family. http://www.mediaandthefamily.org/1999vgrc2.html. Walsh, D. (2000). Interactive violence and children: Testimony submitted to the Committee on Commerce; Science, and Transportation, March 21, 2000. Downloaded from http:// www.mediaandthefamily.org/senateviolence-full.html on March 24, 2000.

VIOLENT VIDEO GAMES

249

Wells, G. L., & Windschitl, P. D. (1999). Stimulus sampling in social psychological experimentation. Personality and Social Psychology Bulletin, 25, 1115–1125. Winer, B. J. (1971). Statistical principles in experimental design. New York: McGraw-Hill. Wood, W., Wong, F. Y., & Chachere, J. G. (1991). Effects of media violence on viewers’ aggression in unconstrained social interaction. Psychological Bulletin, 109, 371–383.

This Page Intentionally Left Blank

SURVIVAL AND CHANGE IN JUDGMENTS: A MODEL OF ACTIVATION AND COMPARISON

Dolores Albarracı´n Harry M. Wallace Laura R. Glasman

A model of judgment maintenance and change is proposed that specifies the various processes that take place at the time of making a judgment on the basis of memory-based and online information. This model proposes that attitude maintenance and change depend on three processes: recalling a prior attitude, recalling or receiving other attitude-related information, and comparing the prior attitude with attitude-related information. Unlike prior models, the activation/comparison model assumes that all three processes can elicit attitude change and maintenance under diVerent conditions. For instance, the mere activation of attitude-related information that is consistent with a prior attitude will favor stability, whereas activation accompanied with comparison with a prior attitude will result in polarization of the prior attitude. Furthermore, even when prior attitude accessibility will elicit attitude maintenance in the absence of comparative processes, prior attitude accessibility can accelerate comparison and therefore change when comparative cues are present. Finally, people who are motivated to compare their prior attitudes with new information should by necessity first activate their prior attitude before comparison can take place. Consequently, attitude comparison cues may induce attitude survival if subsequent processing stops at the point of attitude activation and does not proceed to the stage of attitude comparison. Comparative principles are identified and the implications of this model are discussed in relation to prior theorizing on change in attitudes and nonevaluative judgments. 251 ADVANCES IN EXPERIMENTAL SOCIAL PSYCHOLOGY, VOL. 36

Copyright 2004, Elsevier Inc. All rights reserved. 0065-2601/04 $35.00

252

DOLORES ALBARRACI´N et al.

I. Introduction Social psychologists’ current understanding of attitude change and maintenance faces two primary problems. One problem is fragmentation of the literature. Several bodies of research and theory identify mechanisms or thought processes that have diVerent implications for the change and maintenance of attitudes over time. These mechanisms involve recalling a prior attitude about an object, considering online attitude-related information, and evaluating the prior attitude in light of the attitude-related information. However, past research has typically considered these processes at a molecular level, taking into account only one process at a time (for an exception, see Festinger, 1957; Festinger & Carlsmith, 1959). To our knowledge, no prior model of attitude change has attempted to explicate the complex consequences of these rather basic mechanisms for attitude change and survival. The fragmentation of the attitude change literature contributes to a second problem. The predictions that can be made from isolated mechanisms are not the same as those that an integrative view allows. Our model emphasizes that understanding and predicting attitude change requires examination of three processes: (1) activating the prior attitude (retrieving it from memory), (2) activating information related to the prior attitude (which can be from memory or external), and (3) comparing the prior attitude with the related information (Fig. 1).1 None of the processes in Fig. 1 is inevitable, and each process can have diVerent implications for attitude change and maintenance. On one hand, the sole activation of either attitude-consistent information or the prior attitude itself will lead to attitude maintenance. On the other hand, online reconstruction of an attitude based on the sole activation of attitude-inconsistent information, as well as comparison of the prior attitude with attitude-consistent or inconsistent information, should generally produce attitude change. Nevertheless, these two processes do not always occur independently of each other, and better understanding of attitude change emerges from a joint consideration of the 1 We define attitudes as evaluative judgments that are typically generated covertly, and may or may not be expressed to others. Both current and prior attitudes in this case represent the judgment component of attitudes, which is distinct from the representation of this judgment in memory or the representations of attitude-related information in memory (for treatments of the nature of attitude representations in memory, see Bassili & Brown, in press; Fazio, 1986; Wyer & Srull, 1989). We conceptualize change as a diVerence between a present attitude judgment about a topic and an attitudinal judgment about the same topic generated at an earlier time, even when this change results from seemingly fluid and temporary contextual factors (see e.g., Schwarz & Bohner, 2001; Tourangeau & Rasinski, 1988).

SURVIVAL AND CHANGE OF ATTITUDES

Fig. 1.

253

Processes underlying attitude survival and change. Boxes represent processes.

two. Although the processes themselves are not counterintuitive, their joint implications elaborated in our model often contrast with prior assumptions and predictions. Conceptualizing attitude activation and comparison as a sequential process implies that activating a prior attitude will often facilitate comparison. That is, the comparison process requires activation of both the prior attitude and attitude-related information, so spontaneous or externally induced attitude activation (step 1) will foster comparison (step 2) simply because such activation is a prerequisite for comparison. However, activation of both the prior attitude and attitude-related information does not guarantee comparison; attitude activation may not stimulate comparison if situational or individual factors discourage comparison. Just as attitude activation facilitates comparison, comparison cues facilitate activation. When people are motivated to compare their prior attitude with attitude-related information, they may first need to activate the elements required for comparison if these elements were not already available before starting to compare. However, comparison cues do not guarantee comparison even if these cues lead to the activation of the elements required for comparison. The individual may not be suYciently motivated or able to proceed from step 1 (activating attitudes) to step 2 (comparative validation). An example may illustrate the need to investigate attitude activation and comparison in a concurrent fashion. For instance, considering only the activation of a prior attitude leads to the well-supported prediction that highly accessible prior attitudes survive or last longer than prior attitudes with low accessibility (Fazio, 1986). A woman who can easily recall her prior favorable attitude toward COPA airlines is more likely to sustain favorable attitudes than a person who can barely remember that ‘‘COPA’’

254

DOLORES ALBARRACI´N et al.

is a Panamanian company, let alone an evaluation of the service provided by this airline. However, if attitude activation is considered in the context of comparative processes, it is easy to see how activating a prior attitude could lead to attitude change, rather than attitude maintenance. For instance, a woman who clearly remembers having a prior favorable evaluation of COPA airlines and is politely greeted by a COPA flight attendant may reassess the validity of her prior attitude in relation to the new information (comparative validation process). In doing this, the subjective validity of her prior attitude and the new information may increase due to the perceived fit between each cognition and prior knowledge (Wyer & Srull, 1989), her mood (Schwarz & Clore, 1983), or the sheer convenience of holding a given attitude (Festinger, 1964). Furthermore, when compared, her positive prior attitude and the new, also positive information should be mutually validating and combine to increase the positivity of the prior attitude (corroboration). Polarization following subjective confirmation of an earlier conclusion, however, should not occur when people form a current attitude on the spot, ignoring a past attitude. For instance, the woman in our example may fail to retrieve her earlier favorable attitude toward COPA airlines at the time when she is greeted by the polite flight attendant. In this case, failure to retrieve the prior attitude should prevent the comparative validation process we described, leading to the online formation of a current attitude that is equally positive to the prior one. In conclusion, attitude survival may occur in a relatively incidental way even if the prior attitude is inaccessible, provided people consider attitude-relevant information that supports the prior attitude. Attitude maintenance may also emerge from failed attempts to compare new information with a prior attitude. In our example, the woman who is politely greeted by the flight attendant may be asked to complete a consumer survey. In filling out the survey, she may try to recall an evaluation of the airline that she developed years ago and to integrate that attitude with the new information she gathered. However, she might become distracted in the middle of completing the survey and default to the prior attitude she recalled to perform a comparison. With such interruptions, motivation to perform a comparison will increase stability instead of change. Because ability and motivation to think about an issue often are necessary to perform the processes in Fig. 1, our model has implications for the eVects of these variables on attitude change. We argue that a prior attitude is most likely to survive when people activate the prior attitude but do not use evaluatively inconsistent information for either comparisons or online constructions. Therefore, people must have suYcient ability and motivation to retrieve the prior attitude, but not enough ability and motivation to process

SURVIVAL AND CHANGE OF ATTITUDES

255

relevant but inconsistent material.2 Consider the influence of ability and motivation to think about an issue when the attitude-relevant information has the same evaluative implications as the prior attitude. Whereas comparison of a prior attitude with attitude-consistent information should lead to change (polarization), either recycling the old accessible attitude or forming an attitude on the go should lead to attitude survival. Thus, lower and moderate levels of processing ability and motivation should promote attitude survival more than high levels of these variables. This chapter is organized in three sections. The next section (Section II) defines basic concepts pertinent to the model. The predictions described in Section III include an extended presentation of the potential forms of attitude change (e.g., polarization of prior attitudes, boomerang types of eVects, and compromise between the prior attitude and the attitude-related information) based on the evaluative category or implication (good vs. bad) of the judgment and the relevant information, as well as the processing stages that develop (Part A of Section III). The chapter then reviews the influences of attitude activation and comparison (Part B of Section III) and the predicted eVects of general processing ability and motivation on attitude change (Part C of Section III). In Section IV, we contrast the predictions of our model with prior models of attitudes, including algebraic models, such as information-integration theory (Anderson, 1974), Sherif and Hovland’s (1961) social judgment theory, and more recent models of communication and persuasion. II. Attitude Change and Survival: Preliminary Definitions People form attitudes when they link an object to an evaluative category in the process of evaluating the object (good vs. bad, positive vs. negative; see Ajzen & Fishbein, 1980, 2001; Eagly & Chaiken, 1993; Fazio, 1986; Wyer & Srull, 1989; Zanna & Rempel, 1988). The attitude object can be a concrete target, a behavior, an abstract entity, a person, or an event (e.g., Fishbein & Ajzen, 1974). For example, individuals form evaluations of social groups 2 Prior research has shown that people’s ability and motivation to think about an object can induce diVerent inferences about the validity of a given type of information (Chaiken, 1980; Petty & Cacioppo, 1986). For instance, people use their mood as information when they think about their mood to a moderate extent (Albarracı´n & Kumkale, 2003) but not when ability and motivation are either high or low. Moreover, individuals may correct for the influence of mood when they have high ability and motivation, resulting in reverse eVects of mood on judgment (see Ottati & Isbell, 1996). These eVects are orthogonal to the eVects of concern in this chapter, as they aVect the actual validity of the prior attitude and the attitude-related information but not the processes that guide attitude change and maintenance.

256

DOLORES ALBARRACI´N et al.

(e.g., prejudice), their own behaviors (attitude toward the behavior; Fishbein & Ajzen, 1975), their personal attributes (attitudes toward the self, including self-esteem), and other people (person impressions). Moreover, individuals form these attitudes on the basis of diVerent types of information. For instance, Eagly and Chaiken (1993; see also Ajzen, 2001; Albarracı´n, 2002; Petty & Wegener, 1999) maintain that attitudes can be based on cognitive, aVective, and behavioral information. A person who values health and believes that smoking poses health risks is likely to develop a negative attitude toward smoking (Fishbein & Ajzen, 1975). People who experience positive mood may favorably evaluate their lives or conclude that a political candidate is desirable (Clore & Parrott, 1994; Isbell & Wyer, 1999; Schwarz & Clore, 1983, 1996). Moreover, according to Bem (1965, 1972; Bem & McConnell, 1970), individuals often consider their past behavior and infer their attitudes from that behavior, particularly when they lack a firm prior attitude about the object being considered (for relevant evidence, see Chaiken & Baldwin, 1981). Although aVective and behavioral information often gives way to attitudes, attitudes are neither aVective nor behavioral in nature: They are evaluations (Eagly & Chaiken, 1993; Zanna & Rempel, 1988). Three attitude dimensions are relevant to the model’s predictions. Attitudes are first characterized by their evaluative category or implication stemming from people’s classifications of objects as good or bad (e.g., Bargh, Chaiken, Govender, & Pratto, 1992). Moreover, attitudes vary in the extremity or the polarity of the value assigned to the evaluative category (Thurstone, 1959). For example, an individual may moderately or strongly dislike former President Clinton or may be weakly or extremely prejudiced against a societal group. Finally, people’s attitudes are associated with diVerent degrees of confidence or subjective certainty (see Abelson, 1988; Petty, Brin˜ol, & Tormala, 2002). A woman may be very confident that Clinton was a good president, but decrease confidence in this attitude as a result of new information that derogates Clinton’s image.3 This chapter is concerned with stability and changes in the evaluations of an object as time passes. Attitude change denotes a transformation of at least one of the aforementioned attitude dimensions over two diVerent (and arbitrarily defined) time points. Some attitudes change in evaluative category or implication. For instance, individuals who initially favor a president may later receive information that leads them to disapprove of this president. Attitudes also change in extremity and confidence. For example, people with negative attitudes toward smoking may disapprove of smoking even more after making 3

It is important to note that properties such as extremity and confidence are often termed attitude strength. We believe that conceptualizing attitude change requires understanding change in all of these dimensions.

SURVIVAL AND CHANGE OF ATTITUDES

257

new friends who also disapprove of smoking (attitude polarization following corroboration). As an example of change in confidence, people may become more confident in their stereotypes about a minority group if training designed to decrease stereotypes backfires (boomerang type of eVect). In contrast to attitude change, attitude survival (also denoted maintenance and persistence) denotes continuity in the three dimensions of evaluative category, extremity, and confidence. For example, people can automatically access their prior evaluations of political parties and maintain the same exact attitudes over many years (e.g., Fazio, 1990; Fazio, Powell, & Herr, 1983). Attitudes also survive when individuals form an attitude that is identical to a prior one on the basis of relevant information that is available online, even when they did not retrieve their prior attitude from permanent memory. As described by Bem (1965; see also Albarracı´n & Wyer, 2000), people may repeatedly infer an attitude on the basis of their past behavior that happens to be salient at the time (for a review of conditions that facilitate selfperception, see Eagly & Chaiken, 1993). When this behavior is the same over time, the attitude individuals generate is also likely to be stable (for discussions of how chronic accessibility of information can lead to attitudes that are identical to attitudes people formed at an earlier time, see Schwarz & Bohner, 2001; Wilson & Hodges, 1992). We will discuss these diVerent phenomena and their facilitating conditions in the context of empirical evidence relevant to our model. Although attitude extremity and confidence are distinct attitude dimensions, they often go hand in hand. Consider the case in which people recall a general, dichotomous categorization of an object as good or bad. For instance, people may recall that they previously thought that a war against Iraq was a bad idea. Those who become more or less confident in their category assignment should express a more or less extreme judgment, provided that there are enough verbal labels in the context of a broader judgmental scale (e.g., ‘‘very’’ or ‘‘slightly’’ bad). In the example, people who become more confident that the Iraq war is a bad idea are likely to describe the war as ‘‘very bad,’’ whereas people who hesitate with respect to their prior judgment may characterize it as ‘‘slightly bad.’’ However, very diVerent eVects might emerge for changes in confidence and extremity when people start with a more specific judgment, such as ‘‘very’’ bad on a semantic diVerential scale. Under those conditions, becoming more confident in one’s judgment could imply reendorsing ‘‘very bad’’ as one’s judgment rather than changing the extremity of one’s attitude. In addition to ceiling eVects, implicit theories about change may further complicate the situation. If a person’s initial attitude changes in valence, confidence may increase or decrease depending on the circumstances being considered. For example, a person who is initially against the Iraq war may

258

DOLORES ALBARRACI´N et al.

receive information about the advantages of the war, leading that person to manifest a neutral or even positive position. If the person focuses on his or her having changed the earlier attitude in light of more comprehensive information, his or her attitude confidence should increase as extremity decreases. In contrast, if the person focuses on his or her having changed the prior attitude, he or she may conclude that attitudes in this domain are short-lived and decrease his or her confidence in them. In sum, the model discussed in this chapter predicts changes in extremity and confidence in the case of attitude polarization, but it makes no assumptions about confidence in the case of a compromise between the prior attitude and the attitude-related information.

III. Activation/Comparison Model The activation/comparison model stipulates that attitude maintenance and change are a function of (1) prior attitude activation, (2) activation of information related to the prior attitude, and (3) comparison between the prior attitude and the attitude-related information. According to this activation/comparison model, none of these three components is inevitable. People may activate their prior attitude without activating related information. For example, a man planning to vote in an upcoming political election may recall his prior favorable attitude toward the candidate without recalling specific information about the candidate or attending to media portrayals of the candidate. Moreover, concurrent activation of a prior attitude and information related to that attitude does not guarantee comparison between these two components. The man might read a newspaper article describing the behavior of the candidate without considering how the implications of the information in the article fit with his attitude toward the candidate. We describe the nature and predicted eVects of attitude activation and information comparison on attitude change and maintenance below. Figure 2 shows the outcomes of these processes on change and survival, and Table I summarizes the model postulates. A. PROPOSED PROCESSES 1. Activation of Prior Attitude People activate a prior attitude if they retrieve an association between an object and a category that they previously stored in memory (see Fazio, 1986). For instance, a person may recall that a person is ‘‘good’’ or ‘‘bad,’’

Fig. 2. Influence of processes of activation and comparison on attitudinal outcomes. Boxes indicate processes and diamonds indicate decision points. A and B represent diVerent implications or categories (e.g., bad versus good). V and I denote subjectively valid and invalid information, respectively. AA and BB represent attitudes of the same A and B evaluation with increased confidence, extremity, or both.

DOLORES ALBARRACI´N et al.

260

TABLE I Postulates about the Processes and Conditions Underlying Attitude Survival and Change Number

Postulate/principle Independent EVects of Attitude and Attitude-Related Information on Attitude Survival and Change

Postulate 1

Attitude activation without comparison. When comparison processes do not occur, attitude activation alone can lead to attitude survival.

Postulate 2

Entirely online reconstruction. When people do not activate a prior attitude but they activate attitude-related information, they are likely to form a new attitude on the basis of the attitude-related information. In this situation, the prior attitude survives when the attitude-related information is consistent with the prior attitude, but changes when the attitude-related information is inconsistent with the prior attitude. Principles and EVects of Comparison on Attitude Survival and Change

Principle 1

Corroboration principle of comparison. People increase the confidence and extremity of their subjectively valid prior attitudes when the evaluative implications of subjectively valid attitude-related information corroborate their prior attitudes. Correspondingly, they maintain the confidence or extremity of their prior attitudes when the prior attitudes and the attitude-related information are both evaluatively consistent but only one is valid.

Principle 2

Defensive-confidence principle of comparison. People increase the confidence and extremity of a given evaluation (prior attitude or evaluative implication of attitude-related information) when they perceive that evaluation as valid but invalidate the other, incongruent evaluation. In contrast, comparing a prior attitude with inconsistent but valid attitude-related information results in a compromise between the prior attitude and the attitude-related information.

Postulate 3

General eVects of comparison. Comparative processes will generally induce attitude change. Thus, ability (e.g., use of comparative formats for presenting the information) and motivation (e.g., use of comparison instructions) to compare are likely to result in attitude change.

Reciprocal Influences of Attitude Activation and Comparison on Each Other and Their EVects on Attitude Survival and Change Postulate 4

EVects of prior attitude activation in the presence of comparative motivation. Although prior attitude activation (e.g., accessibility) will increase survival in the absence of comparative motivation (Postulate 1), attitude activation may accelerate change in the presence of such motivation.

Postulate 5

EVects of comparative motivation in the absence of attitude activation. Comparative motivation in the absence of attitude activation will facilitate attitude activation and stability but may be insuYcient to produce comparison and the corresponding attitude change. continues

SURVIVAL AND CHANGE OF ATTITUDES

261

TABLE I continued Number

Postulate/principle

EVects of General Processing Ability and Motivation on Attitude Survival and Change Postulate 6

EVects of ability and motivation in the presence of attitude-inconsistent information. When attitude- relevant information conflicts with a prior attitude, high and low levels of ability and motivation to think about the issue should induce more change than moderate levels of general ability and motivation.

Postulate 7

EVects of ability and motivation in the presence of attitude-consistent information. When attitude-relevant information corroborates a prior attitude, high ability and motivation to think about the issue should induce more change than both moderate and low levels of general ability and motivation.

an event is ‘‘desirable’’ or ‘‘undesirable,’’ or an object is ‘‘attractive’’ or ‘‘unattractive’’ (for the processes underlying categorization, see Bargh, 1994, 1997; Markman & Gentner, 2000; Medin & Coley, 1998; Smith, Fazio, & Cejka, 1996; Wyer, 1973; Wyer & Srull, 1989). Attitude retrieval from memory can occur automatically or follow the application of eVortful recall strategies (see Lingle & Ostrom, 1981).4 For example, teachers have less diYculty retrieving their evaluations of a student currently in their class than they do retrieving their evaluations of a student in their class four years ago (for a discussion of goal-directed recall, see Koriat, 2000). Compelling evidence indicates that activation of prior attitudes is important for attitude maintenance, specifically from studies on the attitudebehavior relation. Strong attitude- behavior correlations can be interpreted as manifestations of survival of attitudes over time. For instance, Fazio, Powell, and Williams (1989; Fazio, Chen, McDonel, & Sherman, 1982; Fazio & Williams, 1986) showed that the probability that people will use a given attitude as a basis for their future behavior is a function of the accessibility of the attitude. In this research, participants first reported their attitudes toward food products and then were allowed to take products home. As shown in Fig. 3, the correlation between participants’ attitudes and their later selection of products was higher when attitudes were initially more accessible than when they were initially less accessible. That is, attitudes guided behavioral decisions to a greater extent when at the time of selecting the products, participants could quickly recall the attitudes they reported at the beginning of the experiment than when they could not.

4 Although the model we propose can be applied to online initial attitudes, this chapter concentrates on initial attitudes that people retrieve from permanent memory.

262

DOLORES ALBARRACI´N et al.

Fig. 3. Influence of attitude accessibility on the attitude-behavior relation. Numbers are correlations. Data are from Fazio, Powell, and Williams (1989).

Fig. 4. Path analysis from Glasman and Albarracı´n’s (2003) meta-analysis. The analysis is based on the within-study correlations among each of the three variables in the model. ***p