Advances in Group Processes, Volume 17 (Advances in Group Processes) (Advances in Group Processes)

LIST OF CONTRIBUTORS

Jeffrey W. Bowlby

Applied Research Branch Human Resources Development Canada

John F Dovidio

Department of Psychology Colgate University

Patrick J. Flinn

Alston and Bird LLP Atlanta, Georgia

Margaret Foddy

School of Psychological Science La Trobe University

Samuel L . Gaertner

Department of Psychology University of Delaware

Karen A . Hegtvedt

Department of Sociology Emory University

Gladys Kafati

Department of Psychology Colgate University

Neil]. MacKinnon

Department of Sociology and Anthropology University of Guelph

Phoebe Riches

School of Psychological Science La Trobe University

Cecilia L . Ridgeway

Department of Sociology Stanford University

Judith G. Smetana

Department of Clinical and Social Sciences in Psychology University of Rochester

Vii

vu i

LIST OF CONTRIBUTORS

Frans N . Stokman

ICS University of Groningen

John F. Stolte

Department of Sociology Northern Illinois University

Jonathan H. Turner

Department of Sociology University of California, Riverside

Marcel A .L .M . van Assen

ICS University of Groningen

Jelle van der Knoop

DECIDE University of Groningen

Reinier C.H. van Oosten

DECIDE University of Groningen

Henry A . Walker

Department of Sociology University of Arizona

Elaine Wethington

Departments of Human Development and of Sociology Cornell University

Morris Zelditch

Department of Sociology Stanford University



PREFACE

EDITORIAL POLICY Advances in Group Processes publishes theoretical analyses, reviews, and theorybased empirical chapters on group phenomena . The series adopts a broad conception of "group processes ." This includes work on groups ranging from the very small to the very large, and on classic and contemporary topics such as status, power, exchange, justice, influence, decision making, intergroup relations, and social networks . Contributors have included scholars from diverse fields including sociology, psychology, political science, philosophy, mathematics, and organizational behavior . The series provides an outlet for papers that may be longer, more theoretical, and/or more integrative than those published by standard journals . We place a premium on the development of testable theories and theory-driven research . Chapters in the following categories are especially apropos :

•

•

Conventional and unconventional theoretical work, from broad metatheoretical and conceptual analyses to refinements of existing theories and hypotheses . One goal of the series is to advance the field of group processes by promoting theoretical work . Papers that review and integrate programs of research . The current structure of the field often leads to the piecemeal publication of different parts of ix



x

PREFACE

•

a program of research . This series offers those engaged in programmatic research on a given topic an opportunity to integrate their published and unpublished work into a single paper, Review articles that transcend the author's own work are also of considerable interest . Papers that develop and apply social psychological theories and research to macro-sociological processes . One premise underlying this series is that links between macro- and micro-sociological processes warrant more systematic and testable theorizing . The series encourages development of the macro-sociological implications embedded in social psychological work on groups .

In addition the editors are open to submissions that depart from these guidelines .

CONTENTS OF VOLUME 17 This volume of Advances in Group Processes covers a broad section of theory and research . The first two chapters address stereotyping and discrimination . Both contributions emphasize the importance of intergroup relations in the production of seemingly intrapersonal phenomena. In their chapter "Group Identity and Intergroup Relations : The Common In-group Identity Model," Dovidio, Gaertner, and Kafati illustrate how the same social categorization processes that ordinarily lead to prejudice and discrimination, can also be used to reduce them . Their theoretical and empirical analyses suggest that cooperative intergroup contact can significantly alter the cognitive interpretations responsible for prejudice, stereotyping, and behavioral discrimination . The next chapter utilizes a distinct, yet somewhat complementary, theoretical perspective . In "The Affective Dynamics of Stereotyping and Intergroup Relations," MacKinnon and Bowlby show how affect control theory can be used to analyze the connections among stereotypes, intergroup attitudes, and discrimination . Three studies are shown to be consistent with hypotheses derived from the affect control model . Taken together, these two chapters do an especially nice job of extending and integrating research on prejudice and stereotyping . The next two papers deal with important theoretical issues concerning status in task groups . In "The Formation of Status Beliefs : Improving Status Construction Theory," Ridgeway explicates the basic logic of status construction theory . Upon reviewing the theory and evidence for status construction processes, Ridgeway offers a "belief formation" assumption to more formally explain how social interaction fosters consensual status beliefs . Next, Foddy and Riches extend a theory of status cues by specifying how task and categorical markers of status combine to alter social interaction . They argue that task-related cues (such as verbal fluency) can overwhelm the effect categorical membership cues (such as ethnic accent) when the two are opposed. The reason is that the former are more directly



Preface

xi

related to the task outcome . Two studies involving Anglo-Australian and Greek-Australian speakers were found to support the main hypotheses . Three papers tackle various aspects of power in bargaining and exchange . In "Strategic Decision Making," Stokman and collaborators describe a simulation-based methodology for analyzing decisions wherein stakeholders have differing priorities, values, and power capabilities . This technique uses empirical data gathered by subject area specialists to predict the decision process and outcome . Not only do these models generate highly accurate predictions, they also provide a strategic tool for analyzing the decision-making process and potential points of compromise . The next chapter represents an important follow-up to a theoretical paper that appeared in the first volume of Advances in Group Processes more than a decade ago. In "The Normative Regulation of Power," Zelditch and Walker analyze and examine theoretical propositions based on their elaboration of the Dornbush-Scott theory of authority . They report data from nine experiments which scrutinize various implications of the theory . An important finding is that internal self-restraint is a tenuous basis for the normative regulation of power-external constraints play a major role . This trio of papers is rounded out by Stolte's chapter titled `Beyond the Concept of Value in Power-Dependence Theory : Expanding a Model of the `Whole Actor ."' Here, Stolte takes on the important task of clarifying and extending Emerson's unfinished theory of value . His theoretical analysis yields important insights into the kinds of decisions that are based on rational deliberation, versus those guided by feelings and emotions . This analysis also raises important questions about the relationship between value and environmental uncertainty . Readers interested in rational choice theory and/or emotional processes should find Stolte's chapter especially provocative . The next chapter deals with conflict in familial relations . In "Cultural Contexts and Developmental Processes in Adolescent-Parent Relationships" Smetana describes a program of research directed at understanding adolescent-parent conflict within a broader normative context . It is argued that adolescents and parents experience conflict due to divergent perceptions and interpretations of important issues . In turn, these conflicts play an important role in modifying the overall family structure and furthering adolescent development . Data from European American, African American, and Chinese families are compared and contrasted to illustrate the similarities and differences in conflict across cultures . Two chapters deal with the spread or transmission of social phenomena . Wethington's chapter "Contagion of Stress" addresses how social stressors are spread from one area of life to another . Wethington catalogues the major types of stress contagion and reviews the primary theoretical perspectives that are used to understand them . She then discusses in unusually fine detail how these approaches may be improved, arguing for refined measurement procedures, a combination of qualitative and quantitative data, and longitudinal studies of the contagion process . Next, Hegtvedt and Flinn break fresh ground by pushing justice theory in new directions in their chapter titled "Intergenerational Justice and the Environment :



xii

PREFACE

Determining the Fair Use of Mono Basin Water ." They first theoretically assess how fair distributions come to be viewed as unfair, and detail how actors attempt to resolve competing justice claims . These ideas are then applied to the trans-generational distribution of natural resources . An insightful case study involving California water-rights litigation is offered to illustrate the principles involved . Finally, Jonathan H . Turner offers a broad-ranging conceptual framework for analyzing the relations between micro-, meso-, and macro-level forces . Turner claims that "embedded encounters" are shaped by the larger categorical and corporate units in which they occur. He then develops a number of testable propositions and shows how these might be useful for an array of sociological theorizing . The prospects and strategies for building multilevel theory are discussed . Shane R . Thye Edward J . Lawler Michael W . Macy Henry A . Walker Volume Co-Editors

GROUP IDENTITY AND INTERGROUP RELATIONS THE COMMON IN-GROUP IDENTITY MODEL

John F . Dovidio, Samuel L . Gaertner, and Gladys Kafati

ABSTRACT Discrimination is an intergroup as well as an interpersonal phenomenon and is related to fundamental processes associated with social categorization and the development of in-group favoritism . We propose in the Common In-Group Identity Model that by understanding the factors that underlie the development of these intergroup biases, these forces can be redirected to improve intergroup relations . In particular, if members of different groups are induced to recategorize themselves as a superordinate group rather than as two separate groups, intergroup prejudice, stereotyping, and discrimination will be reduced through the extension of pro-in-group bias to former out-group members . Moreover, recategorization in the form of a dual identity, in which superordinate and subgroup identities are both salient, can achieve the benefits of a strong superordinate identity without requiring groups to forsake valued ethnic and racial identities . Data reported from laboratory studies, field experiments, and surveys involving a range of different types of groups offer converging evidence in support of the model .

Advances in Group Processes, Volume 17, pages 1-35 . Copyright m 2000 by JAI Press Inc. All rights of reproduction in any form reserved . ISBN : 0-7623-0651-3 1



2

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

Overt expressions of prejudice have declined significantly over the past 35 years (Dovidio and Gaertner 1986, 1998 ; Schuman, Steeh, Bobo, and Krysan 1997) . To the extent that prejudice causes intergroup biases (Dovidio . Brigham, Johnson, and Gaertner 1996), corresponding reductions in discrimination would be expected to occur . However, discrimination is shaped by more than personal prejudice (e .g ., LaPiere 1934) . Discrimination is an intergroup as well as an interpersonal phenomenon . Thus, more comprehensive perspectives and strategies for combating discrimination need to consider the nature of the relations between groups as well as between persons . The present chapter therefore examines the role of intergroup processes in both generating and also potentially in improving intergroup relations . In particular, we propose that intergroup contact can reduce social biases by changing the way people view the relations between groups .

THEORETICAL PERSPECTIVES ON RACE RELATIONS Psychological and sociological approaches to race relations have traditionally reflected different perspectives . Psychological theories have commonly emphasized the role of the individual, in terms of personality and attitude, in social biases and discrimination (see Duckitt 1992 ; Jones 1997) . Traditional psychological theories, such as the work on the authoritarian personality (Adorno, Frenkel-Brunswik, Levinson, and Sanford 1950), have considered the role of dysfunctional processes on the overt expression of social biases . In addition, contemporary approaches, such as aversive racism and symbolic racism perspectives, have considered the contributions of normal processes (e .g., socialization and social cognition) in the expression of subtle, and often unconscious, biases (Dovidio and Gaertner 1996, 1998 ; Gaertner and Dovidio 1986a ; Kovel 1970 ; Sears 1988) . Sociological theories, in contrast, have frequently emphasized the role of large-scale social and structural dynamics in intergroup relations in general, and in race relations in particular (Blauner 1972 ; Bonacich 1972 ; Wilson 1978) . These theories have considered the dynamics of race relations largely in economic and class-based terms-and often to the exclusion of individual influences (see Bobo 1999) . Despite the existence of such divergent views, both sociological and psychological approaches have converged to recognize the importance of understanding group functions and collective identies on race relations (see Bobo 1999) . Blumer (1958a, 1958b, 1965a, 1965b), for instance, offered a sociologically based approach in which group competition and threat are fundamental processes in the development and maintenance of social biases . Blumer (1958a) wrote : "Race prejudice is a defensive reaction to such challenging of the sense of group position . . . . As such, race prejudice is a protective device . It functions, however shortsightedly, to preserve the integrity and position of the dominant group" (p . 5) . From a psychological orientation, Sherif, Harvey, White, Hood, and Sherif (1961)



Group Identity and Intergroup Relations

3

similarly proposed in their classic Robbers Cave study that the functional relations between groups is the critical factor determining intergroup attitudes . According to this position, which is related to realistic conflict theory (Bobo 1988 ; Bobo and Hutchings 1996), competition between groups produces prejudice and discrimination, whereas intergroup interdependence and cooperative interaction that result in successful outcomes reduce intergroup bias . Discrimination can serve less tangible collective functions as well as concrete instrumental objectives . Blumer (1958a) acknowledged that the processes for establishing group position may involve goals such as gaining economic advantage, but they may also be associated with the acquisition of intangible resources such as prestige . Theoretical developments in social psychology, stimulated by Social Identity Theory (Tajfel and Turner 1979), further highlight the role of group categorization, independent of actual realistic group conflict, on motivations to achieve favorable group identities ("positive distinctiveness") and consequently on the arousal of intergroup bias and discrimination . Personal prejudice and stereotypes, realistic group threat, and symbolic threats reflect different hypothesized causes of discrimination, but they can operate jointly to motivate discriminatory behavior . Bobo and his colleagues (see Bobo 1999) have demonstrated that group threat and personal prejudice can contribute independently to discrimination against other groups . W . Stephan and his colleagues (Stephan and Stephan 2000 ; Stephan, Ybarra, Martinez, Schwartzwald, and Tur-Kaspa 1998) have also found that personal negative stereotypes, realistic group threat, and symbolic group threat all predict discrimination against other groups (e .g ., immigrants), and each accounts for a unique portion of the effect . The independence of these effects points to different underlying mechanisms . In this chapter we examine how group representations and identities influence the nature of race relations .

INTERPERSONAL AND INTERGROUP PROCESSES The operation of group-level processes is hypothesized to be dynamically distinct from the influence of individual-level processes . Different modes of functioning are involved, and these modes critically influence how people perceive others and experience their own sense of identity . In terms of perceptions of others, for example, Brewer (1988 ; see also Fiske and Neuberg's 1990 Continuum Model) has proposed a "dual process model of impression formation ." The primary distinction in this model is between two types of processing : person based and category based . Person-based processing is bottom-up . It is data-driven, involving the piecemeal acquisition of information that begins "at the most concrete level and stops at the lowest level of abstraction required by the prevailing processing objectives" Brewer 1988, p . 6) . Category-based processing, in contrast, proceeds from global to specific ; it is top-down . In top-down processing, how the external



4

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

reality is perceived and experienced is influenced by category-based, subjective impressions . According to Brewer, category-based processing is more likely to occur than person-based processing, because social information is typically organized around social categories . With respect to one's sense of identity, Social Identity Theory (Tajfel and Turner 1979) and Self-Categorization Theory (Turner 1985) view the distinction between personal identity and social identity as a critical one . When personal identity is salient, a person's individual needs, standards, beliefs, and motives primarily determine behavior. In contrast, when social identity is salient, "people come to perceive themselves as more interchangeable exemplars of a social category than as unique personalities defined by their individual differences from others" (Turner, Hogg, Oakes, Reicher and Wetherell 1987, p . 50) . Under these conditions, collective needs, goals, and standards are primary . Illustrating the dynamics of this distinction, Verkuyten and Hagendoom (1998) found that when individual identity was primed, individual differences in authoritarianism were the major predictor of the prejudice of Dutch students toward Turkish migrants . In contrast, when social identity (i .e ., national identity) was made salient, in-group stereotypes and standards primarily predicted prejudiced attitudes . Thus, whether personal or collective identity is more salient critically shapes how a person perceives, interprets, evaluates, and responds to situations and to others (Kawakami and Dion 1993, 1995) . In general, the social categorization of others and oneself plays a significant role in prejudice and discrimination . We propose, in the Common Ingroup Identity Model, that appropriately engineered social categorization during group contact can play an important role in the reduction of social biases.

SOCIAL CATEGORIZATION AND BIAS Models of category-based processing (Brewer 1988 ; see also Fiske and Neuberg 1990) assume that "the mere presentation of a stimulus person activates certain classification processes that occur automatically and without conscious intent . . . . The process is one of 'placing' the individual social object along well-established stimulus dimensions such as age, gender, and skin color" (Brewer 1988, pp . 5-6) . We have further hypothesized that "a primitive type of categorization may also have a high probability of spontaneously occurring, perhaps in parallel process . This is the categorization of individuals as members of one's ingroup or not" (Dovidio and Gaertner 1993, p . 170) . Because of the centrality of the self in social perception (Higgins and Bargh 1987 ; Kihlstrom et al . 1988), we propose that social categorization involves most fundamentally a distinction between the group containing the self, the in-group, and other groups, the out-groupsbetween the "we's" and the "they' s" (see also Tajfel and Turner 1979 ; Turner et al . 1987) . This distinction has a profound influence on evaluations, cognitions, and behavior .



Group Identity and Intergroup Relations

5

Upon social categorization of individuals into in-groups and out-groups, people favor in-group members directly in terms of evaluations and resource allocations (Tajfel, Billig, Bundy, and Flament 1971), as well as indirectly in valuing the products of their work (Ferguson and Kelley 1964) . Also, in-group membership increases the psychological bond and feelings of "oneness" that facilitates the arousal of promotive tension or empathy in response to their needs or problems (Hornstein 1976) . In part as a consequence, prosocial behavior is offered more readily to in-group than to out-group members (Dovidio et al . 1997 ; Piliavin, Dovidio, Gaertner, and Clark 1981) . In addition people are more likely to be cooperative and exercise more personal restraint when using endangered common resources when these are shared with in-group members than with others (Kramer and Brewer 1984) . In terms of information processing, people retain information in a more detailed fashion for in-group members than for out-group members (Park and Rothbart 1980), have better memory for information about ways in-group members are similar and out-group members are dissimilar to the self (Wilder 1981), and remember more positive information about in-group than out-group members (Howard and Rothbart 1982) . Moreover, people are more generous and forgiving in their explanations for the behaviors of in-group relative to out-group members . Positive behaviors and successful outcomes are more likely to be attributed to internal, stable characteristics (the personality) of in-group than out-group, whereas members negative outcomes are more likely to be ascribed to the personalities of out-group members than of in-group members (Hewstone 1990) . Relatedly, observed behaviors of in-group and out-group members are encoded in memory at different levels of abstraction (Maass, Salvi, Arcuri and Semin 1989) . Undesirable actions of outgroup members are encoded at more abstract levels that presume intentionality and dispositional origin (e .g ., she is hostile) than identical behaviors of in-group members (e .g ., she slapped the girl) . Desirable actions of out-group members, however, are encoded at more concrete levels (e .g ., she walked across the street holding the old man's hand) relative to the same behaviors of in-group members (e .g ., she is helpful) . Because abstract schemas are resistant to modification, these biases lead to a persistence of more positive cognitions and evaluations about ingroup than out-group members . Thus, in-group/out-group categorization not only initiates bias, but it also helps to maintain biases . Given the centrality and spontaneity of the social categorization of people into in-group and out-group members, how can bias be reduced? One of the most influential strategies involves creating and structuring intergroup contact .

INTERGROUP CONTACT AND THE REDUCTION OF BIAS Over the past half-century, Allport's (1954) revised Contact Hypothesis has been a guiding framework for strategies designed to reduce intergroup bias and conflict



6

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

(Pettigrew 1998) . This hypothesis proposes that simple contact between groups is not automatically sufficient to improve intergroup relations . Rather, for contact between groups to reduce bias successfully certain prerequisite features must be present. These characteristics of contact include equal status between the groups, cooperative (rather than competitive) intergroup interaction, opportunities for personal acquaintance between the members, especially with those whose personal characteristics do not support stereotypic expectations, and supportive norms by authorities within and outside of the contact situation . Although it is difficult to establish all of these conditions in intergroup contact situations, this formula is effective when these conditions are met (Cook 1984 ; Johnson, Johnson, and Maruyama 1983 ; Pettigrew 1998) . Recent approaches have extended research on the Contact Hypothesis by attempting to understand the potential common processes and mechanisms that these diverse factors engage to reduce bias (Brewer and Miller 1984) . For example, when intergroup contact is favorable and has successful consequences, psychological processes that restore cognitive balance or reduce dissonance produce more favorable attitudes toward members of the other group and toward the group as a whole to be consistent with the positive nature of the interaction . Also, the rewarding properties of achieving success may become associated with members of other groups, thereby increasing attraction (Gaertner et al . 1999) . Intergroup contact can also influence how interactants conceive of the groups and how the members are socially categorized . Cooperative learning and jigsaw classroom interventions (Aronson and Patnoe 1997), which are designed to increase interdependence between members of different groups and to enhance appreciation for the resources they bring to the task, may reduce bias in part by altering how interactants conceive of the group boundaries and memberships .

CONTACT, CATEGORIZATION, AND IDENTITY As we have argued earlier (Gaertner, Dovidio, Anastasio, Bachman, and Rust 1993), the factors specified by the Contact Hypothesis may reduce bias through the process of reducing the salience of the intergroup boundaries or redefining group boundaries . The former process refers to decategorization, whereas the latter process involves recategorization . Decategorization and Personalization

From the perspective of decategorization strategies, intergroup interaction under the conditions specified by the Contact Hypothesis can individuate members and reduce the salience of intergroup boundaries . Interventions that reveal variability in the opinions of out-group members or have out-group members



Group Identity and Intergroup Relations

7

respond as individuals rather than as a group renders each member more distinctive and thus weakens the prior categorization scheme (Wilder 1986) . Increasing the perception that both in-group and out-group members are "separate individuals" rather than part of a "we-they" categorization scheme further reduces bias by undermining the effects of pro-in-group biases in evaluations, cognitions, and behavior. Another decategorization strategy of repeatedly criss-crossing category memberships by forming new subgroups each composed of members from former subgroups changes the pattern of who is "in" and who is "out" can also weaken the effects of prior categorization (Urban and Miller 1998) . Brewer and Miller (1984) offered a potentially unifying theoretical framework that posits that the features specified by the Contact Hypothesis (e .g ., equal status, cooperative interaction, self-revealing interaction, and supportive norms) share the capacity to decategorize group boundaries and to promote more differentiated and personalized conceptions, particularly of out-group members . Brewer and Miller propose that personalization, involving the exchange of personal information, can complement and enhance the impact of decategorization . With a more differentiated representation of outgroup members, there is the recognition that there are different types of out-group members, thereby weakening the effects of categorization and its subsequent effects on intergroup bias . During this personalization process, interactants focus on information about an out-group member that is relevant to the self (as an individual rather than self as a group member) . Repeated personalized interactions with a variety of out-group members can, over time, undermine the value of the category stereotype as a source of information about members of that group . The effects of personalization may therefore generalize to new situations as well as to other out-group members who are not present during the intergroup contact situation (Bettencourt, Brewer, Croak, and Miller 1992 ; Marcus-Newhall, Holtz, and Brewer 1993 ; Miller, Brewer, and Edwards 1985) . Thus, intergroup contact can reduce bias in both an immediate and generalizable fashion (see also, Pettigrew 1997) . In addition to decategorization, we propose that the features specified by the Contact Hypothesis also reduce bias through recategorization that increases rather than decreases the level of category inclusiveness (Gaertner et al . 1993, 1994, 1996) . That is, these contact conditions facilitate a reduction in bias, in part, because they share the capacity to transform members' representations of the memberships from separate groups to more inclusive social entity . However, we do not necessarily consider the decategorization and recategorization frameworks to be completely mutually exclusive theoretically ; we believe them to be capable of working in parallel and complementary ways . Given the complexity of intergroup attitudes, it is plausible that these features of contact operate through several related as well as different pathways .



8

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI Recategorization

Recategorization, as represented by our own work on the Common In-group Identity Model, involves interventions to change people's conceptions of the memberships from different groups to one, more inclusive group (Gaertner et al . 1993) . The model is rooted in the social categorization perspective of intergroup behavior and recognizes the central role of social categorization in reducing as well as in creating intergroup bias (Tajfel and Turner 1979) . Specifically, if members of different groups are induced to conceive of themselves more as a single, superordinate group rather than as two separate groups, attitudes toward former

Conditions of Contact

Representational Mediators

Consequences

Intergroup Interdependence (e.g., joint outcomes)

One Group Recategorization ("We")

Evaluative Consequences (e.g., prejudice)

Group Differentiation (e.g., similarity) Environmental Context (e.g, egalitarian norms) Pre-Contact Experience (e.g., affective priming)

Two Subgroups

in One Group Recategorization ("Us+Them =We") Two Groups Categorization ("We/They") Separate Individuals Decategorization ("Me/You") /

Cognitive Effects (e.g., stereotyping)

Behavioral Effects (e.g., helping)

Figure 1 . Within the Common In-group Identity Model, elements of an intergroup contact situation (e .g, intergroup interdependence) influence cognitive representations of the groups as one superordinate group (recategorization), as two subgroups in one group (recategorization involving a dual identity), as two groups (categorization), or as separate individuals (decategorization) . Recategorization and decategorization, in turn, can both reduce cognitive (e .g., beliefs), affective (e .g ., feelings of empathy), and behavioral (e .g ., helping) biases, but in different ways .



Group Identity and Intergroup Relations

9

out-group members will become more positive through processes involving pro-in-group bias . As noted earlier, categorization of a person as an in-group member rather than as an out-group member has been demonstrated to produce more positive evaluations and to influence systematically how people process information in ways that reinforce these favorable attitudes . For instance, changing the basis of categorization from race to an alternative dimension can alter who is a "we" and who is a "they," thus undermining racial biases . Thus, cooperative relationships between group, such as those established in Sherif and colleague's (1961) Robbers Cave study, can reduce bias not only by ameliorating realistic group threat but also by establishing a more inclusive, superordinate group identity (Gaertner et al . 2000) . The Common In-group Identity Model is presented schematically in Figure 1 . On the left, situational factors and interventions, such as intergroup interdependence (which produces cooperative behavior) and group similarities or differences, are hypothesized to influence cognitive representations of the memberships (center) as one group (recategorization), two subgroups in one group or on the same team (recategorization with a dual identity), as two groups (categorization), or as separate individuals (decategorization) . Within the model, intergroup cognition involved in social categorization is the critical mediator of subsequent intergroup attitudes and behavior . The model recognizes that recategorization and decategorization strategies can both reduce prejudice, but in different ways . Recategorization, either in terms of a one group or dual identity representation, reduces bias by extending the benefits of in-group favoritism to former out-group members . Attitudes and behavior toward these former outgroup members thus become more favorable, approaching those toward in-group members . Decategorization, in contrast, reduces favoritism toward original ingroup members as they become perceived as separate individuals rather than as members of one's own group .

INTERGROUP CONTACT AND THE COMMON IN-GROUP IDENTITY MODEL As illustrated in Figure 1, we hypothesize that interventions in intergroup contact situations designed to produce a recategorization of people from members of different groups to members of a common group can have a significant impact on intergroup attitudes (prejudice), stereotypes, and behavior . We summarize empirical evidence consistent with the basic propositions of the model . Intergroup Attitudes Cooperative interaction, which results from intergroup interdependence and superordinate goals, is one of the most powerful and robust aspects of intergroup



10

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

contact for reducing bias (Sherif and Sherif 1969) . From our perspective, cooperative interaction may enhance positive evaluations of out-group members, at least in part, by transforming interactants' representations of the memberships from two groups to one group . To test this hypothesis directly, we conducted a laboratory experiment that brought two three-person laboratory groups together under conditions designed to vary independently the members' representations of the aggregate as one group or two groups and the presence or absence of intergroup cooperative interaction (Gaertner, Mann, Dovidio, Murrell, and Pomare 1990) . To develop separate group identities and allegiances, participants first worked on a problem-solving task in three-person groups, unaware of the existence of the other group . We then arranged contact between the groups on another problemsolving task (involving survival in the wilderness) in which structural aspects of the situation were designed to emphasize separate or common group identity (e .g ., segregated or integrated seating ; original group names or a new group name for all six participants) . In addition, we manipulated whether the outcome (and potential reward) for their solution to would be evaluated independently for the two three-person groups or jointly for the six-person group . After the interaction, participants evaluated original in-group and out-group members . The "groupness" and the independent/joint outcome manipulations both reduced intergroup bias . Moreover, even when the groups had independent outcomes, participants induced to feel like one group relative to those whose separate group identities were reinforced reported that the aggregate did feel more like one group . They also had lower degrees of intergroup bias in their evaluations of ingroup and out-group members . Consistent with our hypothesis about how interdependent outcomes reduces bias among participants induced to feel like two groups, the information that the joint solution developed by the six-person group would be evaluated increased the extent to which the participants felt like they belonged to one group. Feelings of belonging to one group, in turn, mediated reduced levels of bias in evaluative ratings . Also supportive of the Common Ingroup Identity Model (see also Brewer 1979), reduced bias associated with introducing interdependent outcomes was due to enhanced favorable evaluations of out-group members . More direct support for the hypothesized mediating role of cognitive representations was revealed by the multiple regression mediation approach, a form of path analysis . The results are presented in Figure 2 . The coefficients associated with the paths are standardized beta weights . This analysis indicated that the influence of interdependent relative to independent outcomes on more positive evaluations of out-group members was substantially reduced (i .e ., from a beta of .62 to one of .09 ; see Figure 2) when the mediating effects of group representations and perceptions of cooperation and competition were considered . Furthermore, consistent with our model, among these potential mediators, only the one-group representation related independently to evaluations of out-group members . That is, intergroup interdependence improved attitudes toward the out-



Group Identity and Intergroup Relations

11

Individuals

.03

-.59 -.74 i \

Interdependent/ Independent Outcomes

Different Groups

One Group

26

/

~~

. •~-r''r,

Attitudes Toward Outgroup

.80

Cooperation t

I .62 Note :

.09

Thick arrows indicate statistically significant paths (p < .05)

Figure 2.

The mediating role of cognitive representations in the effects of cooperation on the reduction of intergroup bias .

group, and thereby reduced intergroup bias, primarily indirectly by producing more inclusive, one-group representations . The advantage of this experimental design is that interdependence preceded changes in participants' representations of the aggregate from two groups to one group and also changes in intergroup bias . In addition, because we manipulated the representations of the aggregate in the absence of interdependence, the development of a one-group representation preceded changes in intergroup bias . Thus, we can be confident about the directions of causality in this study . In other experiments we have manipulated structural aspects of intergroup contact situations, such as segregated or integrated seating, and found that these manipulations directly influenced group representations and, ultimately, intergroup bias (Gaertner and Dovidio 1986b ; Gaertner, Mann, Murrell, and Dovidio 1989) . Taken together, these studies provide convergent evidence for a clear sequence of events and identification of causal relations . In a second series of studies we utilized survey techniques under more naturalistic circumstances . The advantage of these studies is that they may provide more externally valid evidence related to enduring and meaningful groups, often with histories of conflict . One of these studies involved a survey of black, Chinese, Hispanic, Japanese, Korean, Vietnamese, and Caucasian students attending a multi-ethnic high school in the northeastern United States (see Gaertner et al . 1994, 1996) . The primary theoretical question we pursued in this study was



12

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

whether students' perceptions of the student body as one group, different groups "playing on the same team," or separate groups would mediate the proposed relation between their perceptions of the favorableness of the conditions of contact (e .g ., cooperative interaction, equal status ; see Green, Adams, and Turner 1988) and their degrees of intergroup bias . Contact Hypothesis items measuring participants' perceptions of equal status between the groups included items such as, "Teachers at this school are fair to all groups of students ." Group representations were assessed with items such as, "Despite the different groups at school, there is frequently a sense that we are just one group ." One index of intergroup bias included items designed to assess feelings (good, uneasy, bad, respectful) toward one's own ethnic group relative to each of the other ethnic groups . Another index involved rating each ethnic/racial group on a modified "feelings thermometer" (see Abelson, Kinder, Peters, and Fiske 1982 ; Esses, Haddock, and Zanna 1993) . The findings were consistent with the processes we hypothesized in the Common In-group Identity Model . More favorable perceptions of the conditions of intergroup contact predicted stronger representations of the student body as one group and as different groups on the same team, but weaker representations as separate groups . Cognitive representations, in turn, predicted bias . The more the student body was perceived to be one group or different groups on the same team, the lower the bias in affective reactions . The more the student body was perceived as different groups, the greater the bias in favorability . We conducted two additional survey studies of intergroup relations in different contexts-banking mergers (Bachman 1993) and blended families from remarriage (Banker and Gaertner 1998) . In these studies Contact Hypothesis items were adapted, for example, to measure participants' perceptions of equal status between the groups in each context (e .g., "Employees from [the participants' company] and [the other organization] are seen as equal members of the merged organization" for banking mergers ; "My stepparent treats me, my siblings and step-siblings equally" for blended families) . The findings of these studies offered converging support for the hypothesis that the features specified by the Contact Hypothesis improve evaluations of out-group members and decrease intergroup bias by changing members' representations of the groups . In particular, perceptions of more favorable conditions of intergroup contact predicted more inclusive, one-group representations . These representations, in turn, mediated more positive interorganizational and interfamily attitudes . Stereotypes In addition to our evidence that a common superordinate identity increases positive evaluations of people who would otherwise be regarded as out-group members, other researchers (Maass, Ceccarelli, and Rudin 1996 ; Study 2) have demonstrated its capacity to reduce subtle linguistic biases the serve to perpetuate stereotypes (Maass, Ceccarelli, and Rudin 1996 ; Study 2) . Maass and colleagues



Group Identity and Intergroup Relations

13

gave northern and southern Italians living in Switzerland messages that emphasized the differentiation between northern and southern Italians (in a two-groups condition), or between Italians and the Swiss (in a superordinate "Italian" condition) . These participants were then shown cartoons depicting northern and southern Italians performing positive and negative behaviors . Participants were then asked to choose one of four response alternatives corresponding to the four levels of abstraction (see Semin and Fiedler 1988) . When the distinction between northern and southern Italians was emphasized, the results replicated the linguistic bias effect . Higher levels of abstraction were used to describe positive behaviors of in-group members (e .g ., she is helpful) than for out-group members (e .g ., she walked with the old lady) . Also, higher levels of abstraction were used to describe undesirable behavior of out-group members than for in-group members . This linguistic bias was not evident in the superordinate identity condition . Thus, a superordinate identity fundamentally changed the way behavioral information about in-group and out-group members was processed and importantly, in a way that reduced this subtle bias in information processing . In addition, research by Karpinski and Von Hippel (1996) revealed that this linguistic bias mediates the extent to which people maintain stereotypic expectancies in the face of disconfirming information . Thus, the development of a common in-group identity can not only reduce general intergroup prejudice but also, by reducing intergroup linguistic bias, can help to change intergroup stereotypes . Intergroup Behaviors

In addition to producing more positive intergroup attitudes and reducing prejudice, the development of a common in-group identity can facilitate more positive intergroup behaviors and thus reduce discrimination . Although the connection between positive intergroup attitudes and behaviors seems intuitive, the extension from evaluative bias to intergroup behavior is not a trivial one . As we noted at the beginning of this chapter, discrimination is shaped by more than prejudice . Indeed, the empirical relationship between evaluative bias and is not particularly strong . In a review of the literature regarding the relationship between racial prejudice and racial discrimination by whites, Dovidio and colleagues (1996) found that the correlation was significant but relatively modest in magnitude (r = .32) . In addition, Struch and Schwartz (1989) demonstrated that intergroup bias and intergroup aggression were virtually unrelated (r = .07) and had different antecedents . The variables that predicted intergroup aggression strongly, such as perceived conflict and trait inhumanity, predicted intergroup bias only weakly, if at all. Struch and Schwartz (1989) concluded, "These results make clear the danger of generalizing from research on in-group favoritism to intergroup aggression" (p . 371) . Thus, in another set of studies we focused the consequences of promoting a common in-group identity for positive behavioral interactions between groups .



14

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

Although, as Struch and Schwartz's (1989) work demonstrates, intergroup bias and intergroup behavior may have different antecedents, the development of a common in-group identity is likely to influence both evaluations and positive forms of intergroup interaction . With respect to helping, theoretically several factors could contribute to the effects of recategorization on intergroup helping . Recategorization could operate indirectly through several interpersonal processes . Recategorization of a person as an in-group member increases perceptions of shared beliefs (e .g ., Brown 1984; Wilder 1984) . The perception of greater interpersonal similarity, in turn, typically facilitates helping (see Dovidio 1984) . In addition, recategorization may facilitate empathic arousal, whereby a person's motivational system becomes coordinated to the needs of another (Hornstein 1976 ; Piliavin et al . 1981), which may motivate helpful and sometimes altruistic actions (see Batson 1991) . Intergroup processes, as well as interpersonal processes, can promote helping . Hornstein and his colleagues (Flippen, Hornstein, Siegal, and Weitzman 1996 ; Hornstein 1976 ; Hornstein, Masor, Sole, and Heilman 1971) have demonstrated that recognition of common group membership increases helping beyond the mere effects of interpersonal similarity or attraction . Hornstein and his colleagues (Hornstein 1976 ; Sole, Marton, and Hornstein 1975) proposed that factors such as similarity or common fate may give rise to a sense of "we-ness"-a sense of belonging to a common group . This sense of we-ness, in turn, leads to more positive behaviors such as helping, particularly under conditions of threat (Dovidio and Morris 1975 ; Flippen et al . 1996) . Sole and colleagues (1975) suggested that developing this sense of we-ness facilitated helping toward others independent of interpersonal attraction . It is possible that the development of a common in-group identity (the sense of "we-ness") could also facilitate other forms of positive social interaction, such as the exchange of more self-disclosing information among members of the recategorized superordinate group . Recategorization fundamentally involves a more inclusive redefinition of the in-group boundary, and self-disclosure is intimately involved in the regulation of interpersonal boundaries . According to Derlega, Metts, Petronio, and Margulis (1993), "It is useful to consider adjustments in self-disclosure outputs and inputs as similar to the opening and closing of 'boundaries"' (p . 67) . Self-disclosure and interpersonal closeness are reciprocally related. Self-disclosure communicates a desire for intimacy and promotes closeness (Altman and Taylor 1973) ; in turn, as the closeness of the relationship between interactants increases, self-disclosure increases (Dindia and Allen 1992) and can produce even greater involvement with the other person (Davis 1976 ; Derlega and Chaikin 1976 ; Shaffer, Pegalis, and Bazzini 1996). Thus, creating a common in-group identity may produce a bond of closeness that facilitates greater self-disclosure (Derlega et al . 1993), as well as greater trust, trustworthiness, and cooperativeness (Kramer and Brewer 1984) . We therefore hypothesized that recategorization from two groups to one superordinate group would increase both



Group Identity and Intergroup

Relations

15

helping and self-disclosure between members of former out-groups (see Dovidio et al . 1997) . Following the basic paradigm of Gaertner and colleagues (1990) for manipulating group representations (described earlier), members of two three-person laboratory groups were brought together under conditions (e .g ., integrated or segregated seating, same or different t-shirts) designed to induce them to conceive of themselves as one group or to reinforce their separate group identities . Then, some participants were given an opportunity to help or to engage in a self-disclosing interaction with an in-group or an out-group member . For helping, participants listened individually to an audio-recording of another student (described as either an over or under-estimator, from a previous session) describe how illness had prevented her from completing an important survey of student life for a committee on which she served . Subsequently, participants received a note ostensibly from this person that contained an appeal to help her by placing posters recruiting volunteers to participate in the survey in various locations across campus . The other participants were engaged in a self-disclosure task in which they were asked to discuss the topic, "What do I fear most?" During this task, subjects interacted with either an in-group or out-group member who participated earlier in their session . The results for the helping and self-disclosure measures converged to support predictions from our model . In each case, the bias favoring in-group members, which was present in the two-groups condition, was eliminated for those induced to regard the aggregate as one group . We found further evidence of the benefits of emphasizing a common in-group identity on interracial helping in a field experiment (Ward, Rust, Nier, Gaertner, and Carpenter 1995) . In this study, which was conducted at the University of Delaware football stadium prior to a game between the University of Delaware and Westchester State University, black and white male and female students approached white 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 fans who similarly wore clothing that identified their university affiliation, we systematically varied whether fans and our interviewers had common or different university identities in a context in which we expected university identities to be particularly salient . Consistent with the results of our laboratory experiment, white fans' helping behavior toward a black confederate was significantly influenced by making the common university identity salient . The black confederates were helped with the interviewer more when they visibly shared the common group identity than when they did not (60% versus 38%) . The white confederates were helped equivalently across the two conditions (43% versus 40%) . These fans were obviously not color-blind : Only black interviewers who shared common university affiliation with the fans-and who presumably voluntarily assimilated to the fans' salient group identity-were given especially high levels of assistance .



16

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

These effects of recategorization on both helping and self-disclosure have practical, as well as theoretical, implications . They suggest that recategorization manipulations can help to create interactions among group members in the contact situation that can activate other processes, which can subsequently promote more positive intergroup behaviors and attitudes . In particular, both self-disclosure and helping typically produce reciprocity . More intimate self-disclosure by one person normally encourages more intimate disclosure by the other (Archer and Berg 1978 ; Derlega et al . 1993) . As we discussed earlier, the work of Miller, Brewer, and their colleagues (e .g ., Brewer and Miller 1984 ; Miller et al . 1985) has demonstrated that personalized and self-disclosing interaction can be a significant factor in itself in reducing intergroup bias . Considerable cross-cultural evidence also indicates the powerful influence of the norm of reciprocity on helping (Schroeder, Penner, Dovidio, and Piliavin 1995) . According to this norm, people should help those who have helped them, and they should not help those who have denied them help for no legitimate reason (Gouldner 1960) . Thus, the development of a common in-group identity can motivate interpersonal behaviors between members of initially different groups that can initiate reciprocal actions and concessions (see Deutsch 1993 ; Osgood 1962) that will not only reduce immediate tensions but can produce more harmonious intergroup relations beyond the contact situation . Although finely differentiated impressions of out-group members may not be an automatic consequence of forming a common in-group identity, these more elaborated, differentiated, and personalized impressions can quickly develop because the newly formed positivity bias is likely to encourage more open communication (Gaertner et al . 1993) . The development of a common in-group identity creates a motivational foundation for constructive intergroup relations which can act as a catalyst for positive reciprocal interpersonal actions . Thus, the recategorization strategy proposed in our model and decategorization strategies, such as individuating (Wilder 1984) and personalizing (Brewer and Miller 1984) interactions, can potentially operate complementarily and sequentially to improve intergroup relations in lasting and meaningful ways . More positive interpersonal and intergroup relations, which predict more inclusive group representations, may not only relate to more positive attitudes toward others but also more favorable feelings about and commitment to one's organization or institution. Both individually and socially, this commitment can have important consequences . For instance, attrition rate is a serious issue for students of color on college campuses (Steele 1997) with between 50 and 60 percent of black and Hispanic college students dropping out of college before receiving their degree (Carter and Wilson 1997 ; Stangor and Sechrist 1998) . Institutional commitment is an important predictor, over and above academic performance, of retention (Bennett and Okinaka 1990) . To explore the applicability of the Common In-group Identity Model to this issue, we reanalyzed data from a racial climate survey of black and white college students attending a midwestern state

.. c 4)

C)

ono C) 0 U

c ro V c~f

m L O v--

ro

c O

4J

c

17



18

JOHN F . DOVIDIO, SAMUEL L. GAERTNER, and GLADYS KAFATI

university (Snider and Dovidio 1996) . This survey included items assessing satisfaction with the social environment (with other students, faculty, administrators, and the local community), with experience of discrimination (inside and outside the classroom), and with academic factors (high school preparation and current college grade-point average) . Because the survey was designed for other purposes, the group representations identified by the Common In-group Identity Model were not directly included . One item did relate to feeling to a common ingroup identity-"feeling part of the university community ." This item, however, does not distinguish between a one-group representation or a same-team representation (see Figure 1) ; that is, people could feel part of the university community as one, single common group or they could maintain their group's own identity while feeling collectively part of the "same team" (i .e ., the community overall) . Nevertheless, because both of these perspectives represent recategorization within our model, we used this item to reflect the development of a common group identity . The measure of institutional commitment for our analyses was a composite index of intention to complete the degree at the university and willingness to recommend the university to others . Overall, black students were significantly less satisfied with each of the four aspects of the social environment than were white students . Blacks also reported substantially higher rates of discrimination experienced both inside and outside the classroom . In addition, black students felt significantly less a part of the university community and were substantially less committed to the university than were whites . Of relevance to our model, however, was the pattern of mediation . As illustrated in Figure 3, feeling part of the university community was the primary mediator of commitment . When it was considered in the regression equation it significantly predicted commitment while the effects of race and all of the other variables except satisfaction with the administration and with teaching were nonsignificant . The strength of the path from race to institutional commitment, which was initially significant (beta = - .20) was substantially weakened and nonsignificant (beta = - .01) when the measure of feeling part of the university was considered in the equation . When the responses were analyzed separately by race, feeling part of the community was the most influential predictor for both blacks (beta = .52, p < .001) and whites (beta = .42, p < .001) . The only other significant predictor of institutional commitment was satisfaction with the administration for both black (beta = .25, p < .05) and white (beta = .25, p < .01) students . Thus, the development of a common in-group identity may predict not only the nature of the relationship among individuals but also between individuals and the group, which can have important consequences for both . We found a similar pattern of results in a re-analysis of data of white, Latino, and black psychologists employed in academic departments (see Niemann and Dovidio 1998) . A measure of feeling part of the department was created from two related items in the survey : (1) "I have to prove myself to be fully accepted as a



Group Identity and Intergroup Relations

19

member of my department" (reverse scored), and (2) "I feel free to be myself at work ." The outcome of interest, related to commitment, was job satisfaction (see Niemann and Dovidio 1998) . The responses of whites were compared to those of blacks and Latinos (which did not significantly differ from each other) . Analogous to the findings for students, black and Latino psychologists felt less part of their department and were less satisfied in their position than were white psychologists . Again, consistent with our model, the effect of race/ethnicity on job satisfaction was significantly mediated by feeling part of the department . In general, consistent with the Common In-group Identity Model, we have found evidence that a range of aspects involved in intergroup contact, which have been identified by the Contact Hypothesis as key elements, decrease intergroup bias in part through changing cognitive representations of the groups . Moreover, we have shown that developing a common in-group identity can produce more favorable attitudes, reduce stereotyping, and facilitate greater pro-social and personal interactions with people originally identified as out-group members, and enhance group commitment . Nevertheless, for both practical and theoretical reasons there may be limits to the benefits of interventions designed to emphasize solely a one-group representation .

RECATEGORIZATION AND A DUAL IDENTITY Hewstone (1996) has argued, pragmatically, that interventions designed to create a common in-group identity (such as equal status contact) may not be sufficiently potent to "overcome powerful ethnic and racial categorizations on more than a temporary basis" (p . 351) . Moreover, theoretically, based on the assumptions of Social Identity Theory (Tajfel and Turner 1979 ; Turner 1981), introducing equal status interventions in intergroup contact situations may produce threat to the integrity of members' separate group identities and needs for group distinctiveness . Consequently, to re-establish positive and distinctive group identities, members of such equal status, interacting groups may maintain relatively high levels of intergroup bias (Brown and Wade 1987) or even show increased levels of bias (Deschamps and Brown 1983) . Hornsey and Hogg (2000), for instance, found that a condition that emphasized students' common university membership produced even higher levels of bias between humanities and math-science students than did a condition that emphasized their separate group identities . Thus, under some conditions, equal status intergroup interactions may not produce more inclusive group representations nor successfully reduce bias-and may, in fact, exacerbate bias . Under other conditions, however, equal status contact can reduce bias . Hewstone and Brown's (1986 ; see also Hewstone 1996) Mutual Intergroup Differentiation Model suggests that equal status interaction reduces intergroup bias when original group identities are salient and are not threatened by contact, such as



20

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

when the groups have differentiated areas of expertise . Hewstone and Brown (1986) proposed that for contact between members of different groups to reduce bias successfully, "first, that contact should be 'intergroup' not 'interpersonal' ; and second, that an appropriate model of intergroup contact should be based on 'mutual intergroup differentiation"' (Hewstone 1996, p . 333) in which the groups are distinguished by the experience and expertise that they bring to the situation . When the groups are not differentiated in their roles or experience, positive social identity may be threatened (Tajfel and Turner 1979), and bias may be exacerbated rather than reduced (Jetten, Spears, and Manstead 1997) . Furthermore, according to the Mutual Intergroup Differentiation Model, bias will most effectively be reduced when the groups are encouraged "to recognize mutual superiorities and inferiorities, and to accord equal values to dimensions favoring each group" (Hewstone 1996, p . 334 ; see also Tajfel 1969 ; Turner 1981, 1985 ; van Knippenberg 1984) . Differentiation and a Common In-group Identity

The predictions derived from the Mutual Intergroup Differentiation Model are not necessarily incompatible with those based on the Common In-group Identity Model . Recategorization into a single, undifferentiated group (a one-group representation) is only one form considered in our model (see Figure 1) . We further propose that the development of a common in-group identity does not necessarily require each group to forsake its less inclusive group identity completely . As Ferdman (1995) notes, every individual belongs to multiple groups . For example, Rodriguez-Scheel (1980) found that Chicanos in Detroit chose nonethnic categories, such as religion, occupation, and family, as often as ethnic labels to describe themselves . Thus, recategorization can also take the form of a dual identity, in which superordinate and subgroup identities are both salient . This type of recategorization is represented by a "different groups working together on the same team" representation . Establishing a common superordinate identity while simultaneously maintaining the salience of subgroup identities (i .e., developing a dual identity as two subgroups within one group) may be particularly effective because it permits the benefits of a common in-group identity to operate without arousing countervailing motivations to achieve positive distinctiveness . We report two studies investigating this hypothesis . In one study (Dovidio, Gaertner, and Validzic 1998) groups of three students were given feedback indicating that, based on their performance on an earlier task, their group was higher, lower, or equal in status than another group with which they were about to cooperate . This status manipulation was crossed factorially with whether the groups were assigned identical or different task perspectives in preparation for their cooperative interaction . The discussions involved the Winter Survival Problem (Johnson and Johnson 1975) and when the groups had different task perspectives, the members of one group were told to assume that



Group Identity and Intergroup Relations

21

they would hike to safety whereas the members of the other group were asked to assume that they would stay put to await search parties . As predicted, the analyses revealed that when the groups were of equal status and task perspectives were different, intergroup bias was lower and the representation of the aggregate feeling like one group was higher than in each of the other three conditions (i .e., equal status-same task, unequal status-different task, and equal status-same task) . In addition, the one-group representation mediated the relation between the experimental manipulations of status and task perspective on intergroup bias . These findings are consistent with the proposed value of equal status, primarily, as Hewstone and Brown (1986) propose, when the distinctiveness between groups is maintained . In another study (Mottola 1996) relative group status among actual employees of many different companies was varied by asking participants to imagine that their current organization was about to merge with another . In one condition, their present company was described as higher in status than the other in terms of generating greater sales and greater profits . In another condition, their company was lower in status on both dimensions . In a third condition, both companies were described as having equal status in terms of both sales and profit . In a fourth condition, their company was described as higher in status on one dimension (e .g ., profit), but the other company was higher on the other dimension . Consistent with Hewstone and Brown's (1986) ideas about the benefits of maintaining the mutual distinctiveness of groups working together, participants in the fourth condition in which one company had higher profit and the other higher sales anticipated that they would more strongly identify with the merged organization than did participants in each of the other three conditions (which did not differ from one another) . Thus, when each group can maintain positive distinctiveness, we can anticipate greater acceptance of a superordinate identity from the members of both groups . This would be particularly important in organizational settings when subgroups must be formed and only the convergent pattern is realistically possible . Furthermore, when the basis of group differentiation facilitates the merged group's effectiveness, the development of a common group identity could be enhanced . For example, as in this study, when each subgroup contributes a different but necessary component to the merged organization's effectiveness, group differentiation could enhance the acceptance of organizational identification . Beyond generating more positive intergroup attitudes, the degree of organizational unity experienced can have implications for the effectiveness and productivity of a merged organization . For example, we have found that the extent to which members of two groups perceived of themselves as one superordinate group correlated with the actual effectiveness of their task solution (r = .28, p < .02) . This relationship is also consistent with the meta-analytic conclusions of Mullen and Copper (1994) who found that group cohesiveness significantly predicts group productivity (r = .25) . Moreover, Mullen and Copper suggest that the relationship between cohesiveness and productivity may be bidirectional and thus iterative : cohesiveness may enhance productivity, and successful accomplish-



22

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

ment, in turn, further increases cohesiveness . Indeed, the relationship between successful performance and subsequent productivity is even stronger (r = .51) than the relationship between initial cohesiveness and productivity . Thus, the development of a common ingroup identity can help form a basis for more harmonious intergroup relations to develop through mutual success and achievement, which reinforces the common bond and identity between the groups . As depicted by the "subgroups within one group" (i .e ., a dual identity) representation in our model (see Figure 1), we believe that it is possible for members to conceive of two groups (for example, parents and children) as distinct units within the context of a superordinate (i .e ., family) identity . The development of dual identities, in which groups are allowed or encouraged to maintain their previous group identities while at the same time having salient superordinate identities, is an alternative form of recategorization to the one-group representation . A representation of "different groups on the same team" permits mutual intergroup differentiation while recognizing a common, superordinate group membership . Moreover, this type of recategorization may be particularly effective when people have strong allegiances to their original groups . When group identities and the associated cultural values are central to members' functioning or when they are associated with high status or highly visible cues to group membership, it would be undesirable or impossible for people to relinquish these group identities or, as perceivers, to be "color-blind ." Indeed, demands to forsake these group identities or to adopt a color-blind ideology would likely arouse strong reactance and result in especially poor intergroup relations (see Schofield 1986) . If, however, people continued to regard themselves as members of different groups but all playing on the same team or as part of the same superordinate entity, intergroup relations between these "subgroups" would be more positive than if members only considered themselves as "separate groups" (see Brewer and Schneider 1990) . In this respect, the benefits of a dual identity may be particularly relevant to interracial and interethnic group contexts . Majority and Minority Group Perespectives

Racial and ethnic identities are fundamental aspects of individuals' self-concepts and esteem and thus are unlikely to be readily abandoned . Thus, attempts to produce a single superordinate identity at the expense of one's racial or ethnic group identity may ultimately be detrimental to intergroup relations . In contrast, the development of a dual identity, in terms of a bicultural or multicultural identity, may not only improve intergroup relations but may also contribute to the social adjustment, psychological adaptation, and overall well-being of minority group members (LaFromboise, Coleman, and Gerton 1993) . There is evidence intergroup benefits of a strong superordinate identity can be achieved for both majority and minority group members when the strength of the subordinate identity is equivalently high. For example, in a survey study of white



Group Identity and Intergroup Relations

23

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 for affirmative action policies that would benefit blacks and other minorities on relational concerns regarding the fairness of Congressional representatives than on whether these policies would increase or decrease their own well-being . However, for the group who identified themselves more strongly with being white than being American, their position on affirmative action was determined more strongly by concerns regarding the instrumental value of these policies for themselves . 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 . Among people of color, even when racial or ethnic identity is strong, perceptions of a superordinate connection enhance interracial trust and acceptance of authority within an organization . Huo, Smith, Tyler, and Lind (1996) surveyed white, black, Latino, and Asian employees of a public-sector organization . Identification with the organization (superordinate identity) and racial/ethnic identity (subgroup identity) were independently assessed . Regardless of the strength of racial/ethnic identity, respondents who had a strong organizational identity perceived that they were treated fairly within the organization, and consequently they had favorable attitudes toward authority . Huo and colleagues (1996) conclude that having a strong identification with a superordinate group can redirect people from focusing on their personal outcomes to concerns about "achieving the greater good and maintaining social stability" (pp . 44-45), while also maintaining important racial and ethnic identities . We found converging evidence in our multi-ethnic high school survey study (Gaertner et al . 1996), which we introduced earlier . In particular, we compared students who identified themselves on the survey using a dual identity (e .g ., indicating they were Korean and American) with those who used only a single, subgroup identity (e .g ., Korean) . Supportive of the role of a dual identity, students who described themselves as both American and as a member of their racial or ethnic group had less bias toward other groups in the school than did those who described themselves only in terms of their subgroup identity . Also, the minority students who actually identified themselves using a dual identity reported lower levels of intergroup bias relative to those who only used their ethnic or racial group identity . Not only do whites and people of color bring different values, identities, and experiences to intergroup contact situations, these different perspectives can shape perceptions of and reactions to the nature of the contact . Blumer (1958a) proposed that group status is a fundamental factor in the type of threat that different groups experience . Recent surveys, for example, reveal that blacks show a



24

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

higher level of distrust and greater pessimism about intergroup relations than do whites (Dovidio and Gaertner 1998 ; Hochschild 1995) . Majority group members tend to perceive intergroup interactions as more harmonious and productive than do minority group members (Gaertner et al . 1996 ; Islam and Hewstone 1993 ; see also the college student survey discussed earlier), but they also tend to perceive subordinate and minority groups as encroaching on their rights and prerogatives (Bobo 1999) . In addition, majority and minority group members have different preferences for the ultimate outcomes of intergroup contact. Whereas minority group members often tend to want to retain their cultural identity, majority group members favor the assimilation of minority groups into one single culture (a traditional "melting pot" orientation)-the dominant culture (Horenczyk 1996) . Berry (1984; Berry, Poortinga, Segall, and Dasen 1992) presents four forms of cultural relations in pluralistic societies that represent the intersection of "yes-no" responses to two relevant questions : (1) Are cultural identity and customs of value to be retained? (2) Are positive relations with the larger society of value, and to be sought? These combinations reflect four adaptation strategies for intergroup relations : (1) integration, when cultural identities are retained and positive relations with the larger society are sought ; (2) separatism, when cultural identities are retained but positive relations with the larger society are not sought ; (3) assimilation, when cultural identities are abandoned and positive relations with the larger society are desired ; and (4) marginalization, when cultural identities are abandoned and are not replaced by positive identification with the larger society . Although this framework has been applied primarily to the ways in which immigrants acclimate to a new society (van Oudenhoven, Prins, and Buunk 1998), we propose that it can be adapted to apply to intergroup relations between majority and minority groups more generally . Substituting the separate strengths of the subgroup and subordinate group identities for the answers to Berry's (1984) two questions, the combinations map onto the four main representations considered in the Common In-group Identity Model : (1) same team (subgroup and superordinate group identities are high, like integration) ; (2) different groups (subgroup identity is high and superordinate identity is low, like separatism) ; (3) one group (subgroup identity is low and superordinate group identity is high, like assimilation ; and (4) separate individuals (subgroup and superordinate group identities are low, like marginalization) . Research in the area of immigration suggests that immigrant groups and majority groups have different preferences for these different types of group relations . Van Oudenhoven and colleagues (1998) found in the Netherlands that Dutch majority group members preferred an assimilation of minority groups (in which minority group identity was abandoned and replaced by identification with the dominant Dutch culture), whereas Turkish and Moroccan immigrants most strongly endorsed integration (in which they would retain their own cultural identity while also valuing the dominant Dutch culture) . These preferences also apply to the preferences of whites and of people of color about racial and ethnic group



Group Identity and Intergroup

Relations

25

relations in the United States . We have found that whites most prefer assimilation, whereas racial and ethnic minorities favor pluralistic integration (Kafati 1999) . Moreover, these preferred types of intergroup relations for majority and minority groups, a one-group representation for whites and same-team representation for people of color, may differentially mediate the consequences of intergroup contact for the different groups . To explore this possibility, we examined how different types of cognitive representations might mediate the relationship between more positive perceptions of intergroup contact and higher levels of institutional commitment among white students and students of color (Kafati 1999) . As in the Snider and Dovidio (1996) survey study of college students considered earlier in this chapter, commitment was chosen as the main outcome variable because attrition rates by race/ethnicity are a widespread concern on college campuses . Perceptions of favorable intergroup contact were measured in the same way as in our multi-ethnic high school study (Gaertner et al . 1996) and included ratings on the dimensions of equal status, supportive norms, personal interaction, and cooperative interaction . Commitment included items about students' intentions to complete their education at the institution and their willingness to recommend the institution to others, as well as questions about the academic and social climate . Whereas in the Snider and Dovidio (1995) survey, only a single item, feeling part of the university community, assessed recategorized representations, this study examined all four cognitive representations identified by the Common In-group Identity Model and hypothesized to represent different models of intergroup relations : perceptions of being (1) one group (assimilation), (2) separate groups (separatism), (3) separate individuals (decategorization/marginalization), and (4) different groups on the same team (integration/multiculturalism) . In addition, participants' identification with their racial or ethnic group was assessed . We hypothesized that the positive relation between perceptions of more favorable intergroup contact and institutional commitment would be mediated by different representations of the groups . Whereas we expected that the relation for whites would be mediated by one-group representations, reflecting an assimilation perspective, we anticipated that the relation for people of color would be mediated by same-team representations, a multicultural perspective that recognizes both one's racial or ethnic group identity and a superordinate identity-a dual identity . The results confirmed the hypothesis . For whites, more positive perceptions of intergroup contact related to stronger perceptions of students consisting of one group, different groups on the same team, and separate individuals, as well as weaker perceptions of different groups . However, when considered simultaneously, only the one-group representation was significantly related to commitment (see Figure 4, top panel) . Supportive of the hypothesized mediation, the relationship between perceptions of more favorable conditions of intergroup contact and greater institutional commitment, which was initially significant (beta = .36), was greatly reduced by the inclusion of one-group representations in the



26

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

White Students Individuals

Different Groups

Conditions of Contact

One Group

.04

.36

Students ofColor Individuals

Different Groups

Conditions of Contact

Same Team

34

.37 One Group

.24

Figure 4 .

.04

Intergroup Relations and Satisfaction at College for White Students and for Students of Color



Group Identity and Intergroup Relations

27

equation (beta = .04) . This pattern of results is also consistent with the finding in our field experiment, described earlier in this chapter, that white football fans responded particularly favorably to blacks who apparently adopted their university affiliation as a superordinate identity . Conditions of contact also significantly predicted each of the representations for students of color . In contrast to the pattern for whites, though, the same-team representation-but not the one-group representation-predicted commitment (see Figure 4, bottom panel) . In general, these effects were stronger for people higher in racial ethnic identification, both for whites and people of color . These findings have practical as well as theoretical implications . Although correlational data should be interpreted cautiously, it appears that, for both whites and people of color, favorable intergroup contact may contribute to their commitment to their institution . However, strategies and interventions designed to enhance satisfaction need to recognize that whites and people of color may have different ideals and motivations . Because white values and culture have been the traditionally dominant ones in the United States, American whites may see an assimilation model, in which members of other cultural groups are absorbed into the "mainstream," as the most comfortable and effective strategy . For people of color, this model, which denies the value of their culture and traditions, may not only be perceived as less desirable but also as threatening to their personal and social identity-particularly for people who strongly identify with their group . Thus, efforts to create a single superordinate identity, although well-intentioned, may threaten one's social identity, which in turn can intensify intergroup bias and conflict .

CONCLUSION Discrimination and bias are group-level as well as individual-level phenomena . Categorizing people as members of one's own group (the in-group) and other groups (out-groups) automatically initiates biases that produce and maintain more positive attitudes, stereotypes, and actions toward in-group than out-group members . We propose that by understanding the processes that underlie the development of these intergroup biases, these forces can be redirected to reduce prejudice and discrimination. In particular, the Common In-group Identity Model involves interventions to change people's conceptions of the memberships from different groups to one, more inclusive group . If members of different 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 . In general, data from laboratory studies, field experiments, and surveys involving a range of different types of groups offer converging evidence in support of the model .



28

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

The development of a common in-group identity does not necessarily require each group to forsake its less inclusive group identity completely . Recategorization can also take the form of a dual identity (such as by a "different groups working together on the same team" representation), in which superordinate and subgroup identities are both salient, and the benefits of a strong superordinate identity can be achieved for both majority and minority group members when the strength of the subordinate identity is high . Moreover, because of the salience and importance of racial and ethnic identities among minority group members relative to majority group members, there tend to be differences in these groups' preferences for recategorization strategies and intergroup relations . We have found, for example, that whites most prefer assimilation (reflecting a one-group identity), whereas racial and ethnic minorities favor pluralistic integration (involving a dual-identity or same-team representation) . Furthermore, these preferred types of intergroup relations for majority and minority groups differentially mediate the consequences of intergroup contact . Whereas a one-group representation is of primary importance for whites, a same-team representation is the critical factor for people of color . Taken together, these findings indicate that, although racial and ethnic prejudice and discrimination have long traditions, these biases can potentially be addressed by harnessing the fundamental psychological processes that would typically promote bias and redirecting them to improve intergroup relations . We acknowledge that interventions based on these principles need to be considered carefully . It is important to be cognizant of how interventions might affect social categorization and to be sensitive to the differential effects for majority and minority groups . Nevertheless, understanding the multilevel nature of prejudice and discrimination is an essential step for finding solutions-which may need to be similarly multifaceted .

ACKNOWLEDGMENTS The research described and the preparation of this chapter were supported by NIMH Grant MH 48721 . We are also grateful to the editors and the reviewers for their helpful comments on an earlier version of this chapter .

REFERENCES Abelson, R . P ., D . R. Kinder, M. D . Peters, and S . T . Fiske . 1982 . "Affective and Semantic Components in Political Person Perception ." Journal of Personality and Social Psychology 42: 619-630 . Adorno, T . W ., E . Frenkel-Brunswik, D. J . Levinson, and R . N. Sanford . 1950. The Authoritarian Personality . New York : Harper. Allport, G . W . 1954 . The Nature of Prejudice . Cambridge, MA : Addison-Wesley.



Group Identity and Intergroup Relations

29

Altman, I ., and D . A . Taylor . 1973 . Social Penetration : The Development of Interpersonal Relationships . New York : Holt, Rinehart, and Winston . Archer, R . L., and J . H . Berg . 1978 . "Disclosure Reciprocity and its Limits : A Reactance Analysis ." Journal of Experimental Social Psychology 14 : 527-540 . Aronson, E ., and S . Patnoe . 1997 . The Jigsaw Classroom. New York : Longman . Bachman, B . A . 1993. "An Intergroup Model of Organizational Mergers ." Unpublished PhD Dissertation, University of Delaware, Newark, DE . Banker, B . S ., and S . L . Gaertner . 1998 . "Achieving Stepfamily Harmony : An Intergroup Relations Approach ." Journal of Family Psychology 12 : 310-325 . Batson, C . D . 1991 . The Altruism Question : Toward a Social-psychological Answer. Hillsdale, NJ : Erlbaum . Bennett, C., and A . M . Okinaka. 1990. "Factors Related to Persistence among Asian, Black, Hispanic and White Undergraduates at a Predominantly White University : Comparisons Between First and Fourth Year Cohorts ." The Urban Review 22 : 33- 60 . Berry, J . W . 1984 . "Cultural Relations in Plural Societies ." Pp. 11-27 in Groups in contact: The Psychology of Desegregation, edited by N . Miller and M . B . Brewer . Orlando, FL: Academic Press. Berry, J . W ., Y . H. Poortinga, M . H. Segall, and P. R . Dasen. 1992 . Cross-cultural Psychology : Research and Applications . Cambridge : Cambridge University Press . Bettencourt, B . A ., M . B . Brewer, M . R. Croak, and N. Miller. 1992. "Cooperation and the Reduction of Intergroup Bias: the Roles of Reward Structure and Social Orientation ." Journal of Experimental Social Psychology 28 : 301-319 . Blauner, R . 1972. Race Oppression in America . New York : Harper and Row . Blumer, H . 1958a . "Race Prejudice as a Sense of Group Position ." Pacific Sociological Review 1 : 3-7 . Blumer, H . 1958b . "Recent Research on Race Relations in the United States of America ." International Social Science Bulletin 10 : 403-477 . Blumer, H. 1965a . "Industrialization and Race Relations ." Pp . 228-229 in Industrialization and race relations : A symposium, edited by G . Hunter . New York: Oxford University Press . Blumer, H . 1965b . "The Future of the Color Line ." Pp . 322-336 in The South in continuity and change, edited by J. C . McKinney and E . T . Thompson. Durham, NC : Seeman. Bobo, L . 1988 . "Group Conflict, Prejudice, and the Paradox of Contemporary Racial Attitudes ." Pp . 85-114 in Eliminating Racism : Profiles in Controversy, edited by P . A. Katz and D . A . Taylor . New York: Plenum . Bobo, L . 1999 . "Prejudice as Group Position: Micro-foundations of a Sociological Approach to Racism and Race Relations." Journal of Social Issues 55(3) . Bobo, L ., and V. L . Huchings . 1996 . "Perceptions of Racial Group Competition : Extending Blumer's Theory of Group Position to a Multiracial Context ." American Sociological Review 61 : 951-972 . Bonacich, E . 1972 . "A Theory of Ethnic Antagonism ." American Sociological Review 77 : 547-559 . Brewer, M . B . 1979 . "Ingroup Bias in the Minimal Intergroup Situation : A Cognitive Motivational Analysis ." Psychological Bulletin 86 : 307-324 . Brewer, M . B. 1988 ." A Dual Process Model of Impression Formation." Pp. 1-36 in Advances in Social Cognition : Vol. I: A Dual Process Model of Impression Formation, edited by T . S . Srull and R . S . Wyer . Hillsdale, NJ : Erlbaum . Brewer, M . B ., and N . Miller . 1984 . "Beyond the Contact Hypothesis : Theoretical Perspectives on Desegregation ." Pp . 281-302 in Groups in Contact : The Psychology of Desegregation, edited by N. Miller and M . B . Brewer. Orlando, FL: Academic Press . Brewer, M. B ., and S . Schneider. 1990 . "Social Identity and Social Dilemmas : A Double-edged Sword ." Pp . 169-184 in Social Identity Theory : Constructive and Critical Advances, edited by D . Abrams and M . Hogg . London : Harvester Wheatsheaf.



30

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

Brown, R. J. 1984. "The Effects of Intergroup Similarity and Cooperative vs . Competitive Orientation on Intergroup Discrimination ." British Journal of Social Psychology 21 : 21-33 . Brown, R. J ., and G. Wade . 1987 . "Superordinate Goals and Intergroup Behavior : The Effect of Role Ambiguity and Status on Intergroup Attitudes and Task Performance ." European Journal of Social Psychology 17 : 131-142 . Carter, D . J., and R . Wilson . 1997 . Minorities in Higher Education. Washington, DC : American Council on Education . Cook, S . W . 1984. "Cooperative Interaction in Multiethnic Contexts ." Pp . 291-302 in Groups in Contact: The Psychology of Desegregation, edited by N . Miller and M. B . Brewer. Orlando, FL: Academic Press . Davis, J . D . 1976 . "Self-disclosure in an Acquaintance Exercise : Responsibility for Level of Intimacy ." Journal of Personality and Social Psychology 33 : 787-792 . Derlega, V . J ., and A . L . Chaikin . 1976 . "Norms Affecting Self-disclosure in Men and Women ." Journal of Consulting and Clinical Psychology 44 : 376-380. Derlega, V . J., S . Metts, S . Petronio, and S . T. Margulis . 1993 . Self-disclosure. Newbury Park, CA : Sage . Deschamps, J. C ., and R . Brown . 1983 . "Superordinate Goals and Intergroup Conflict ." British Journal of Social Psychology 22 : 189-195 . Deutsch, M . 1973 . The Resolution of Social Conflict . New Haven, CT : Yale University Press . Dindia, K ., and M . Allen . 1992 . "Sex Differences in Self-disclosure: A Meta-analysis ." Psychological Bulletin 112 : 106-124 . Dovidio, J . F. 1984 . "Helping Behavior and Altruism : An Empirical and Conceptual Overview ." Pp. 361-427 in Advances in Experimental Social Psychology, Vol . 17, edited by L. Berkowitz . New York: Academic Press . Dovidio, J. F ., J . Brigham, B . T . Johnson, and S . L . Gaertner. 1996 . "Stereotyping, Prejudice and Discrimination." Pp. 276-319 in Foundations of Stereotypes and Stereotyping, edited by N . Macrae, C. Stangor and M. Hewstone. New York: Guilford. Dovidio, J . F., and S . L. Gaertner. 1986 . "Prejudice, Discrimination, and Racism : Historical Trends and Contemporary Approaches ." Pp . 1-34 in Prejudice, Discrimination and Racism, edited by J. F . Dovidio, and S. L. Gaertner . Orlando, FL: Academic Press . Dovidio, J. F., and S . L . Gaertner . 1993 . "Stereotypes and Evaluative Intergroup Bias ." Pp . 167-193 in Affect, Cognition, and Stereotyping: Interactive Processes in Intergroup Perception, edited by D . M . Mackie and D . L . Hamilton. Orlando, FL: Academic Press . Dovidio, J . F ., and S . L . Gaertner . 1996. "Affirmative Action, Unintentional Racial Biases, and Intergroup Relations ." Journal of Social Issues 52(4) : 51-76. Dovidio, J. F., and S . L. Gaertner. 1998 . "On the Nature of Contemporary Prejudice : The Causes, Consequences, and Challenges of Aversive Racism ." Pp. 3-32 in Confronting Racism : The Problem and the Response, edited by J . Eberhardt and S . T . Fiske. Newbury Park, CA : Sage . Dovidio, J . F., S . L. Gaertner, and A. Validzic . 1998 . "Intergroup Bias : Status, Differentiation, and a Common In-group Identity ." Journal of Personality and Social Psychology 75 : 109-120. Dovidio, J . F ., S . L. Gaertner, A . Validzic, K. Matoka, B . Johnson, and S . Frazier . 1997 . "Extending the Benefits of Re-categorization : Evaluations, Self-disclosure and Helping ." Journal of Experimental Social Psychology 33 : 401-420 . Dovidio, J . F., and W. N. Morris. 1975 . "The Effects of Stress and Commonality of Fate on Helping Behavior ." Journal of Personality and Social Psychology 31 : 145-149. Duckitt, J . 1992 . "Psychology and Prejudice : A Historical Analysis and Integrative Framework." American Psychologist 47 : 1182-1193 . Esses, V ., G . Haddock, and M . Zanna. 1993 . "Values, Stereotypes, and Emotions as Determinants of Intergroup Attitudes ." Pp . 137-166 in Affect, Cognition and Stereotyping : Interactive Processes in Group Perception, edited by D . Mackie and D . Hamilton . San Diego, CA: Academic Press .



Group Identity and Intergroup Relations

31

Ferdman, B . M . 1995 . "Cultural Identity and Diversity in Organizations : Bridging the Gap Between Group Differences and Individual Uniqueness." Pp . 37-61 in Diversity in Organizations: New Perspectives for a Changing Workplace, edited by M . M . Chemers, S . Oskamp, and M . A. Costanzo . Thousand Oks, CA : Sage . Ferguson, C. K ., and H . H. Kelley . 1964 ." Significant Factors in Over-evaluation of Own Groups' Products." Journal of Abnormal and Social Psychology 69 : 223-228 . Fiske, S . T ., and S . L . Neuberg . 1990. "A Continuum Model of Impression Formation : From Category Based to Individuating Processes as a Function of Information, Motivation, and Attention ." Pp. 1-74 in Advances in experimental social psychology, Vol . 23, edited by M . P . Zanna . San Diego, CA : Academic Press . Flippen, A . R ., H . A. Hornstein, W. E . Siegal, and E . A . Weitzman. 1996. "A Comparison of Similarity and Interdependence as Triggers for Ingroup Formation ." Personality and Social Psychology Bulletin 22 : 882-893 . Gaertner, S . L ., and J. F . Dovidio. 1986a . "The Aversive Form of Racism ." Pp . 61-89 in Prejudice, Discrimination, and Racism, edited by J . F. Dovidio and S . L . Gaertner. Orlando, FL : Academic Press. Gaertner, S . L ., and J . F. Dovidio . 1986b . "Prejudice, Discrimination, and Racism : Problems, Progress and Promise ." Pp . 315-332 in Prejudice, Discrimination, and Racism, edited by J . F. Dovidio and S . L . Gaertner . Orlando, FL : Academic Press . Gaertner, S . L ., J. F . Dovidio, P. A. Anastasio, B . A . Bachman, and M . C . Rust . 1993 . "The Common Ingroup Identity Model : Recategorization and the Reduction of Intergroup Bias ." Pp. 1-26 in European Review of Social Psychology, Vol . 4, edited by W . Stroebe and M . Hewstone . New York: John Wiley and Sons . Gaertner, S . L ., J . F . Dovidio, B . Banker, M . Houlette, K. Johnson, and E. McGlynn . 2000. "Reducing Intergroup Conflict: From Superordinate Goals to Decategorization, Recategorization, and Mutual Differentiation . Group Dynamics 4 : 98-114 . Gaertner, S . L ., J . F . Dovidio, M . C . Rust, J . Nier, B . Banker, C . M. Ward, G. R . Mottola, and M . Roulette . 1999 . "Reducing Intergroup Bias : Elements of Intergroup Cooperation ." Journal of Personality and Social Psychology 76 : 388-402 . Gaertner, S . L., J. A . Mann, J . F . Dovidio, A . J . Murrell, and M . Pomare. 1990 . "How Does Cooperation Reduce Intergroup Bias?" Journal of Personality and Social Psychology 59 : 692-704 . Gaertner, S . L ., J . Mann, A . Murrell, and J . F. Dovidio . 1989 . "Reducing Intergroup Bias : The Benefits of Recategorization ." Journal of Personality and Social Psychology 57 : 239-249 . Gaertner, S . L ., M . C . Rust, J. F. Dovidio, B . A. Bachman, and P. A . Anastasio . 1994. "The Contact Hypothesis : The Role of a Common Ingroup Identity on Reducing Intergroup Bias ." Small Groups Research 25(2): 224-249 . Gaertner, S . L ., M . C . Rust, J. F. Dovidio, B . A. Bachman, and P. A. Anastasio . 1996 . "The Contact Hypothesis : the Role of a Common Ingroup Identity on Reducing Intergroup Bias among Majority and Minority Group Members ." Pp . 230-360 in What's Social about Social Cognition?, edited by J. L . Nye and A. M . Brower. Newbury Park, CA : Sage . Gouldner, A . 1960. "The Norm of Reciprocity : A Preliminary Statement ." American Sociological Review 25 : 161-178 . Green, C . W., A . M . Adams, and C . W. Turner . 1988 . "Development and Validation of the School Interracial Climate Scale ." American Journal of Community Psychology 16: 241-259. Hewstone, M . 1990 . "The 'Ultimate Attribution Error'? A Review of the Literature on Intergroup Attributions ." European Journal of Social Psychology 20 : 311-335 Hewstone, M . 1996 . "Contact and Categorization : Social Psychological Interventions to Change Intergroup Relations." Pp . 323-368 in Foundations of Stereotypes and Stereotyping, edited by N. Macrae, M . Hewstone, and C . Stangor. New York: Guilford .



32

JOHN F . DOVIDIO, SAMUEL L. GAERTNER, and GLADYS KAFATI

Hewstone, M ., and R. J . Brown . 1986 . "Contact Is Not Enough : An Intergroup Perspective on the 'Contact Hypothesis' ." Pp . 1-44 in Contact and Conflict in Intergroup Encounters, edited by M . Hewstone and R . Brown . Oxford : Basil Blackwell . Higgins, E. T ., and J . A . Bargh . 1987 . "Social Cognition and Social Perception ." Annual Review of Psychology 38 : 369-425 . Hochschild, J . L . 1995 . Facing up to the American Dream : Race, Class, and the Soul of the Nation . Princeton, NJ: Princeton University Press . Horenczyk, G . 1966 . "Migrant Identities in Conflict : Acculturation Attitudes and Perceived Acculturation Ideologies ." Pp. 241-250 in Changing European identities : Social Psychological Analyses of Social Change, edited by G . Breakwell and E . Lyons . Oxford : Butterworth-Heinemann . Hornstein, H . A. 1976 . Cruelty and Kindness: A New Look at Aggression And Altruism . Englewood Cliffs, NJ : Prentice Hall . Hornstein, H . A ., H . N. Masor, K . Sole, M . Heilman. 1971 . "Effects of Sentiment and Completion of a Helping Act on Observer Helping : A Case for Socially-Mediated Ziergarnik Effects ." Journal of Personality and Social Psychology 17 : 107-112 . Hornsey, M. J., and M . A . Hogg . 2000 . "Sub-group Relations : Two Experiments Comparing Subgroup Differentiation and Common Ingroup Identity Models of Prejudice Reduction ." Personality and Social Psychology Bulletin (in press) . Howard, J . M ., and M. Rothbart. 1980. "Social Categorization for In-group and Out-group Behavior ." Journal of Personality and Social Psychology 38 : 301-310 . Huo, Y. J ., H. J . Smith, T . R . Tyler, and A. E . Lind. 1996 . "Superordinate Identification, Subgroup Identification, and Justice Concerns : Is Separatism the Problem. Is Assimilation the Answer?" Psychological Science 7 : 40- 45 . Islam, M. R., and M. Hewstone . 1993. "Dimensions of Contact as Predictors of Intergroup Anxiety, Perceived Outgroup Variability and Outgroup Attitude : An Integrative Model ." Personality and Social Psychology Bulletin 19 : 700-7 10. Jetten, J ., R . Spears, and A . S . R . Manstead . 1997 . "Distinctiveness Threat and Prototypicality : Combined Effects on Intergroup Discrimination and Collective Self-esteem ." European Journal of Social Psychology 27 : 635-658. Johnson, D. W., and F . P . Johnson. 1975 . Joining Together: Group Theory and Group Skills . Englewood Cliffs, NJ : Prentice-Hall . Johnson, D. W ., F . P. Johnson, and G . Maruyama . 1983 . "Interdependence and Interpersonal Attraction among Heterogeneous and Homogeneous Individuals : A Theoretical Formulation and a Meta-analysis of the Research ." Review of Educational Research 52 : 5-54. Jones, J . M . 1997 . Prejudice and Racism (2nd ed .). New York: McGraw-Hill . Kafati, G . 1999. "Assessing the Intergroup Climate at Colgate University ." Unpublished manuscript, Department of Psychology, Colgate University, Hamilton, NY . Karpinski, A ., and W. Von Hippel. 1996 . "The Role of the Linguistic Intergroup Bias in Expectancy Maintenance ." Social Cognition 14 : 141-163 . Kawakami, K ., and K . L . Dion . 1993 . "The Impact of Salient Self-identities on Relative Deprivation and Action Intentions ." European Journal of Social Psychology 23 : 525-540. Kawakami, K ., and K . L . Dion. 1995 . "Social Identity and Affect as Determinants of Collective Action." Theory and Psychology 5 : 551-577. Kihlstrom, J . F., N . Cantor, J . S . Albright, B . R. Chew, S . B . Klein, and P. M . Niedenthal . 1988 . "Information Processing and the Study of the Self." Pp. 145-180 in Advances in Experimental Social Psychology, Vol. 21, edited by L . Berkowitz. Orlando, FL: Academic Press . Kovel, J . 1970 . White Racism: A Psychohistory . New York: Pantheon . Kramer, R. M ., and M. B . Brewer. 1984. "Effects of Group Identity on Resource Utilization in a Simulated Commons Dilemma ." Journal of Personality and Social Psychology 46 : 1044-1057 . LaFromboise, T., H . L . K. Coleman, and J. Gerton. 1993 . "Psychological Impact of Biculturalism: Evidence and Theory ." Psychological Bulletin 114 : 395-412 .



Group Identity and Intergroup

Relations

33

LaPiere, R. T . 1934. "Attitudes vs . Actions ." Social Forces 13 : 230-237 . Maas, A ., R . Ceccarelli, and S . Rudin . 1996 . "Linguistic Intergroup Bias : Evidence for In-group-protective Motivation ." Journal of Personality and Social Psychology 71 : 512-526. Maass, A., D . Salvi, L . Arcuri, and G . R. Semin . 1989 . "Language Use in Intergroup Contexts : The Linguistic Intergroup Bias ." Journal of Personality and Social Psychology 57 : 981-993 . Marcus-Newhall, A ., N. Miller, R . Holtz, and M . B . Brewer . 1993 . "Cross-cutting Category Membership with Role Assignment : A Means of Reducing Intergroup Bias ." British Journal of Social Psychology 322 : 125-146. Miller, N ., M . B . Brewer, and K . Edwards . 1985 . "Cooperative Interaction in Desegregated Settings : A Laboratory Analog ." Journal of Social Issues 41(3) : 63-75 . Mottola, G . 1996. "The Effects of Relative Group Status on Expectations of Merger Success ." PhD Dissertation, University of Delaware . Newark, DE . Mullen, B ., and C . Copper. 1994. "The Relation Between Group Cohesiveness and Performance : An Integration ." Psychological Bulletin 115 : 210-227 . Niemann, Y. F ., and J . F . Dovidio . 1998 . "Relationship of Solo Status, Academic Rank, and Perceived Distinctiveness to Job Satisfaction of Racial/ethnic Minorities . " Journal of Applied Psychology 83 : 55-71 . Osgood, C . E . 1962 . An Alternative to War or Surrender . Urbana, IL : University of Illinois Press. Park, B ., and M . Rothbart . 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 . 1997 . "Generalized Intergroup Contact Effects on Prejudice ." Personality and Social Psychology Bulletin 23 : 173-185 . Pettigrew, T. F . 1998 . "Intergroup Contact Theory." Annual Review of Psychology 49: 65-85 . Piliavin, J . A ., J . F . Dovidio, S . L . Gaertner, and R . D . Clark, III. 1981 . Emergency Intervention . New York : Academic Press . Rodriguez-Scheel, J . 1980 . "An Investigation of the Components of Social Identity for a Detroit Sample. " Unpublished manuscript, Occidental College, Psychology Department, Los Angeles, CA . Schofield, J . W. 1986. "Causes and Consequences of the Colorblind Perspective ." Pp . 231-253 in Prejudice, Discrimination and Racism, edited by J . F. Dovidio and S . L. Gaertner . Orlando, FL: Academic Press . Schroeder, D . A ., L . A . Penner, J . F . Dovidio, and J . A. Piliavin . 1995 . Psychology of Helping and Altruism: Problems and Puzzles. New York : McGraw-Hill . Schuman, H., C. Steeh, L . Bobo, and M . Krysan . 1997 . Racial Attitudes in America: Trends and Interpretations (2nd ed .) . Cambridge, MA : Harvard University Press. Sears, D . O . 1988 . "Symbolic Racism ." Pp . 53-84 in Eliminating racism : Profiles in controversy, edited by P. A . Katz and D . A . Taylor . New York : Plenum Press . Semin, G. R ., and K. Fiedler. 1988. "The Cognitive Functions of Linguistic Categories in Describing Persons : Social Cognition and Language ." Journal of Personality and Social Psychology 54 : 558-568 . Shaffer, D . R ., L . J . Pegalis, and D . G . Bazzini . 1996 . "When Boy Meets Girl (Revisited) : Gender, Gender-Role Orientation, and Prospect of Future Interaction on Self-Disclosure Among Sameand Opposite-Sex Acquaintances ." Personality and Social Psychology Bulletin 22 : 495-506. Sherif, M ., O. J . Harvey ; B . J . White, W . R . Hood, and C . W. Sherif. 1961 . Intergroup Conflict and Cooperation. The Robbers Cave Experiment. Norman, OK: University of Oklahoma Book Exchange . Sherif, M ., and C . W . Sherif. 1969. Social Psychology. New York : Harper and Row . Smith, H . J ., and T . R. Tyler. 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.



34

JOHN F . DOVIDIO, SAMUEL L . GAERTNER, and GLADYS KAFATI

Snider, K ., and J . F. Dovidio . 1996 . A Survey of the Racial Climate at Indiana State University . Institutional Research and Testing, Indiana State University, Terra Haute, IN . Sole, K., J . Marton, and H . A . Hornstein. 1975 . "Opinion Similarity and Helping : Three Field Experiments Investigating the Bases of Promotive Tension ." Journal of Experimental Social Psychology 11 : 1-13 . Stangor, C ., and G . B . Sechrist . 1998 . "Conceptualizing the Determinants of Academic Choice and Task Performance Across Social Groups ." Pp . 105-124 in Prejudice: the Target's Perspective, edited by J . K. Swim and C . Stangor . San Diego, CA : Academic Press . Steele, C . M . 1997 . "A Threat in the Air: How Stereotypes Shape Intellectual Identity and Performance." American Psychologist 52: 613-629. Stephan, W . G ., and C . W . Stephan . 2000. "An Integrated Threat Theory of Prejudice ." in Reducing Prejudice: Claremont Symposium on Applied Psychology, edited by S . Oskamp . Hillsdale, NJ : Erlbaum . Stephan, W . G ., O . Ybarra, C . Martinez, J. Shwarzwald, and Tur-Kaspa. 1998 . "Prejudice Toward Immigrants in Spain and Israel : an Integrated Threat Analysis Theory ." Journal of Cross-Cultural Psychology 29 : 559-576 . Struch, N ., and S . H. Schwartz . 1989 . "Intergroup Aggression : Its Predictors and Distinctness from Ingroup Bias ." Journal of Personality and Social Psychology 56: 364-373 . Tajfel, H. 1969. "Cognitive Aspects of Prejudice ." Journal of Social Issues 25(4) : 79-97 . Tajfel, H ., M . G . Billig, R . F . Bundy, and C . Flament . 1971 . "Social Categorisation and Intergroup Behavior ." European Journal of Social Psychology 1 : 149-177. Tajfel, H ., and J . C . Turner . 1979 . "An Integrative Theory of Intergroup Conflict ." Pp . 33-48 in The Social Psychology of Intergroup Relations, edited by W. G. Austin and S . Worchel. Monterey, CA : Brooks/Cole . Turner, J . C . 1981 . "The Experimental Social Psychology of Intergroup Behavior ." Pp . 66-101 in Intergroup Behavior, edited by J . C . Turner and H . Giles . Chicago, IL : University of Chicago Press. Turner, J. C . 1985 . "Social Categorization and the Self-concept : a Social Cognitive Theory of Group Behavior ." Pp . 77-122 in Advances in Group Processes, Vol . 2, edited by E . J . Lawler. Greenwich, CT : JAI Press . Turner, J. C ., M. A . Hogg, P . J . Oakes, S . D . Reicher, and M. S . Wetherell . 1987 . Rediscovering the Social Group : A Self-categorization Theory . Oxford, UK : Basil Blackwell . Urban, L. M., and N . Miller . 1998 . "A Theoretical Analysis of Crossed Categorization Effects : A Meta-analysis ." Journal of Personality and Social Psychology 74 : 894-908 . van Knippenberg, A. 1984. "Intergroup Differences in Group Perceptions." Pp . 560-578 in The Social Dimension : European Developments in Social Psychology, Vol . 2, edited by H . Tajfel . Cambridge, UK : Cambridge University Press. van Oudenhoven, J . P., K. S . Prins, and B . Buunk. 1998 . "Attitudes of Minority and Majority Members Towards Adaptation of Immigrants ." European Journal of Social Psychology 28 : 995-1013 . Verkuyten, M ., and L. Hagendoorn. 1998 . "Prejudice and Self-categorization : The Variable Role of Authoritarianism and In-group Stereotypes ." Personality and Social Psychology Bulletin 24: 99-110. Ward, C. M ., J . Nier, M . C . Rust, S . L . Gaertner, and T . Carpenter. 1995 . "A Game They Came to Watch : A Study of Compliance and Team Membership ." Poster presented at the Annual Eastern Psychological Association Convention, Boston, MA . Wilder, D. A . 1981 . "Perceiving Persons as a Group : Categorization and Intergroup Relations ." Pp . 213-257 in Cognitive Processes in Stereotyping and Intergroup Behavior, edited by D . L . Hamilton. Hillsdale, NJ : Erlbaum . Wilder, D . A. 1984. "Predictions of Belief Homogeneity and Similarity Following Social Categorization ." British Journal of Social Psychology 23 : 323-333 .



Group Identity and Intergroup Relations

35

Wilder, D . A. 1986. "Social Categorization : Implications for Creation and Reduction of Intergroup Bias ." Pp . 291-355 in Advances in experimental social psychology, Vol . 19, edited by L . Berkowitz . Orlando, FL : Academic Press . Wilson, W . J . 1978. The Declining Significance of Race . Chicago : University of Chicago Press .

THE AFFECTIVE DYNAMICS OF STEREOTYPING AND INTERGROUP RELATIONS

Neil J . MacKinnon and Jeffrey W . Bowlby

ABSTRACT Most of the research on stereotyping and intergroup relations in social psychology has been conducted within the social cognition paradigm of psychological social psychology . In this paper we identify three deficits of the social cognition approach to stereotyping and intergroup relations that have hindered its developing a satisfactory explanatory model : first, an overemphasis on cognition and a concomitant neglect of affect ; second, the confounding of affect and cognition by recent attempts to redress this imbalance ; and, third, the tendency to treat the relations among cognitive, affective, and behavioral aspects of intergroup relations as a one-way causal process, rather than as a system of reciprocal effects and cybernetic control . We show how affect control theory, with its twofold emphasis on affect and control, and its clear distinction between cognition and affect, provides a comprehensive, explanatory model for analyzing stereotyping and intergroup relations . To illustrate this claim, we report three studies based on intergroup attitudes for Canadian regional identities . The first applies the in-group bias/ethnocentrism and out-group

Advances in Group Processes, Volume 17, pages 37-76 . Copyright © 2000 by JAI Press Inc . All rights of reproduction in any form reserved . ISBN : 0-7623-0651-3 37



38

NEIL J . MacKINNON and JEFFREY W . BOWLBY homogeneity hypotheses from social cognition research on intergroup perceptions to affect control theory data on intergroup attitudes . The second employs the attribution equations of affect control theory to demonstrate how stereotypic traits can be generated from intergroup attitudes for social identities . The third applies the impression formation equations of the theory to analyze the affective outcomes of intergroup relations between Canadians with different regional identities . Our findings reveal qualified support for the application of the in-group bias and out-group homogeneity hypotheses to affect control theory data on intergroup attitudes . In addition, the generation of stereotypic traits from intergroup attitudes for regional identities, and the analysis of affective outcomes of intergroup relations, yield plausible results . Following a discussion of the limitations of our research, we show how social cognition theory and affect control theory are complementary approaches to the study of stereotyping and intergroup relations .

Originally coined by the journalist, Walter Lippman (1922), to refer to the simple, fixed, and often erroneous conceptions people have about members of social groups, stereotype remains a vibrant concept in social psychology 65 years after Katz and Braly (1933, 1935) applied it to the perception of racial groups . While sociologists and sociological social psychologists have often invoked the concept as an assumed antecedent of prejudice and discrimination, they have generally left theoretical and empirical work on stereotyping itself to psychological social psychologists (for a recent exception, see Ford and Tonander 1998) . Most of this work has been conducted within the social cognition paradigm, inspired by Tafjel's (1969, 1970) discovery that prejudice and discrimination can result from the mere categorization of individuals in laboratory settings . In the three decades following Tafjel's pioneering work, social cognition research on stereotyping has produced a massive literature (for concise summaries, see Hamilton and Mackie 1993a; Stangor, Sullivan, and Ford 1991 ; for reviews, see Hamilton, Sherman, and Ruvolo 1990 ; Messick and Mackie 1989 ; for an overview of social cognition theory from a sociological perspective, see Howard 1995). Impressive as this research has been in delineating the content and ascription of stereotypes and identifying some of the underlying cognitive mechanisms and conditions of stereotyping, it has failed to produce an adequate explanatory model of intergroup processes relating stereotypes to intergroup attitudes and behavior . For instance, one textbook laments a decade after Tafjel's (1969, 1970) pioneering work that "theories explaining the psychological process of stereotyping have not been developed or tested as have theories pertaining to the content of stereotypes . . . [and] the selection of target groups" (Albrecht, Thomas, and Chadwick 1980, p . 255) . And 10 years after this sobering assessment, Stangor, Sullivan, and Ford (1991, p. 360) conclude that "despite the voluminous research concerning patterns of social stereotype ascription, there has surprisingly been very little empirical research concerning the relationship between stereotyping and prejudice or between stereotyping and discrimination ."



Stereotyping and Intergroup Relations

39

Perhaps the past three decades can best be described as a period of normal science (Kuhn 1962), focusing on the implications of Tafjel's (1969, 1970) exemplary research on social categorization . Late into the 1990s research continues to address such conventional issues as differences in perceptions between groups engaged in mutual stereotyping (e .g ., Allen 1996 ; Judd and Park 1995), the perceived variability of in-groups versus out-groups (the out-group homogeneity effect) (e .g ., Judd and Park 1995 ; Lee and Ottati 1995), the favorability of in-groups versus out-groups (the in-group bias or ethnocentrism effect) (e .g., Allen 1996 ; Judd and Park 1995), and the accuracy of intergroup perceptions (e .g ., Ryan 1996) . The failure of social cognition theory to develop a satisfactory, explanatory model of intergroup processes stems in large part from its one-sided emphasis on cognition and its concomitant neglect of affect, along with the confounding of cognition and affect by recent attempts to redress this imbalance . As elaborated below, we attribute this confounding of cognition and affect to two related factors : first, a conceptualization of intergroup attitudes as summary cognitive judgments of group favorability, consisting of cognitive, affective, and behavioral components ; and, second, the use of predictive models to determine the relative contributions of these components to intergroup attitudes . For example, Esses, Haddock, and Zanna (1993) and Stangor and colleagues (1991) regressed intergroup attitudes (a single-item measure of group favorability) on personal and/or consensual stereotypes (the traits ascribed to members of social groups) and group-elicited emotions (the feelings evoked by members of social groups) . Both studies found that, net of group-elicited emotions, stereotypes play a negligible role in predicting intergroup attitudes . What are we to conclude from these implausible findings? To begin with, they are hardly surprising . The fact that both the attitude criterion and the emotions predictors are affective phenomena disadvantages stereotypic traits as predictors of intergroup attitudes . Contrary to the view that intergroup attitudes can be reduced to cognitive judgments of group favorability, we maintain that affect is involved in all such judgments . As argued below, affect is essential to attitude and cannot be excluded by definition without destroying the construct as conventionally understood. Moreover, if interpreted literally, these findings suggest that stereotypes are unimportant in explaining intergroup attitudes . Strictly speaking, however, predictive models do not explain anything, certainly not the causal relationships of stereotype-based cognitions and group-elicited emotions to intergroup attitudes, nor their causal connections to intergroup behavior itself . Besides its traditional neglect of affect, and the confounding of cognition and affect by recent attempts to rectify this deficiency, the failure of social cognition theory to develop an adequate explanatory model of intergroup processes can be attributed to a third factor : the tendency to view the relations among the cognitive, affective, and behavioral components of intergroup processes as a one-way causal process, rather than an integrative system of cybernetic feedback and control . This



40

NEIL J . MacKINNON and JEFFREY W . BOWLBY

is exemplified by the either-or selection of cognition-affect-behavior or affect-cognition-behavior models in the intergroup processes literature . More recent theoretical formulations, however, promise to correct this deficit . In the words of two prominent social cognition researchers, "viewing the intergroup encounter as involving constantly interacting cognitive and affective processes reveals the arbitrariness of any distinction between emotion-to cognition and cognition-to-emotion models of behavior . Instead, it becomes evident that intergroup behavior reflects the joint influence of both kinds of processes" (Hamilton and Mackie 1993b, p . 373) . This paper illustrates the application of affect control theory (Heise 1979; Smith-Lovin and Heise 1988 ; MacKinnon and Heise 1993 ; MacKinnon 1994) to intergroup processes . We begin by expanding the discussion of stereotypes, intergroup attitudes, and intergroup behavior with which we introduced this paper . Then, following a concise statement of affect control theory, we show how its conceptualization of attitude and its twofold emphasis on affect and control address the issues raised in the preceding discussion, and how the theory provides an explanatory model of intergroup processes based on reciprocal relations among stereotypes, intergroup attitudes, and intergroup behavior . Against the background of these conceptual and theoretical discussions, we illustrate the application of affect control theory to intergroup relations with attitudinal data on Canadian regional identities . These data were collected in 1995 from comparable samples of Alberta and Ontario university students, representing two regions of the country characterized by a history of interregional disparity and resentment. The research reported here comprises three distinct studies . Study 1 compares the intergroup attitudes of Alberta and Ontario samples for Canadian regional identities, exploring the classic in-group bias/ethnocentrism hypothesis and the out-group homogeneity hypothesis derived from social cognition research on stereotyping . Study 2 employs the attribution equations of affect control theory to generate stereotypic traits from intergroup attitudes for regional identities . Study 3 applies the theory's impression formation equations to explore the affective outcomes of intergroup relations between Albertans and Ontarions . In the general discussion of these three studies we address the limitations of our research and the additional issues it raises, including the use of mathematical models to study intergroup relations . We conclude this paper by showing how social cognition theory and affect control theory are complementary approaches to analyzing intergroup processes . This discussion includes a brief comparison of affect control theory with social identity theory (Tafjel 1974, 1978, 1982 ; Tajfel and Turner 1979 ; Turner 1982 ; Hogg 1992,1993 ; Hogg and Abrams 1988 ; Hogg, Terry, and White 1995), a later development in social cognition theory that advances a more adequate, explanatory model of intergroup relations than can be found in conventional social cognition research on stereotyping .



Stereotyping and Intergroup Relations

41

STEREOTYPES, INTERGROUP ATTITUDES, AND INTERGROUP BEHAVIOR As commonly understood, stereotypes are rigid, oversimplified, and inaccurate generalizations of the traits and behavior of members of social groups . These "category-based expectations" (Jones and Gillis 1976) engender "snap judgments" (Schneider et al . 1979)-simple, unreflective, and immediate cognitive appraisals about members of a particular group . People tend to engage in more reflective and complex cognitive processing only when the appearance or behavior of group members violates category-based expectations . Thus, stereotypes are more than group concepts in the Katz and Braly (1933, 1935) tradition ; they are cognitive structures that organize and direct the processing of social information ; and, like all cognitive structures, they enable people to deal with the complexity of their social environment efficiently . In this sense, stereotypes function as prototypes (Cantor and Mischel 1979), mental representations or images of typical members of a group that serve as fuzzy categories in cognitive processing. Or, at a more general and complex level, stereotypes can be viewed as generic schemas, abstract cognitive structures that organize experience at both the perceptual and cognitive-interpretative levels of psychological processing (Mandler 1984, pp . 55-60) . More specifically, because stereotypes conespond to social identities based upon membership in or ascription to social groups, they can be treated as identity schemas (MacKinnon 1994 ; Stryker and Serpe 1994) . Although the term, role schemas, is more commonly employed in the social cognition literature, we prefer the term, identity schemas to emphasize that role behavior is predicated upon the identification with social positions, a basic assumption of both identity theory (Stryker 1968, 1980 ; McCall and Simmons 1978) and social identity theory (see references above) . Terminological preferences aside, identity or role schemas refer to "organized knowledge about the expected behaviors of occupants of particular social positions" (Howard 1995, p . 93), those pertaining, not only to instrumental activities (e .g ., familial, occupational), but also to social groups (e .g ., ethnic, racial) . From this perspective, "role [identity] schemas can be considered the [cognitive] basis of stereotyping and intergroup prejudice and discrimination" (p . 93) . In contrast to the general consensus in the intergroup relations literature on the meaning and function of stereotypes, there is considerable disagreement on the definition and role of intergroup attitudes . This reflects the dissension in attitude theory and research in general (Hill 1981) . According to conventional definition, an attitude consists of "positive or negative affect toward or evaluation of some object" (Schuman 1995, p. 68, following Thurstone 1928, 1967 ; cf. Edwards 1957 and Fishbein and Ajzen 1975) . Proponents of this definition distinguish clearly between attitude-"essentially a subjective judgment . . . of like or dislike . . . good or bad . . .favorable or unfavorable" and belief-"a probability judgment that links



42

NEIL J . MacKINNON and JEFFREY W . BOWLBY

some object or concept to some attribute" (Fishbein 1978, p . 378) . While the strength of an attitude is measured by the extremity, intensity, or felt importance of an affectively based evaluative judgment (Schuman 1995), the strength of a belief "is defined by the person's subjective probability that the object-attribute relationship exists (or is true)" (Fishbein 1978, p . 378) ." In short, affectively based evaluations are the essence of attitudes ; cognitive judgments, the essence of beliefs . The causal connection between the two constructs is captured by Fishbein and Ajzen's (1975) model of attitudes, which proposes that beliefs evoke attitudes, which, in turn, engender behavioral intention and response . We return to the distinction between attitudes and beliefs, or affect and cognition, in our discussion of affect control theory . Although some authors consider the identification of attitude with affect or evaluation as the conventional meaning of the construct among social psychologists (e .g ., Schuman 1995), others assert that the most popular definition of attitude today is a multicomponent one (e .g ., Delamater 1992 ; Hamilton and Mackie 1993b ; Zanna and Rempel 1988) based on Rosenberg and Hovland's (1960) definition of attitude as a relatively stable organization of cognitive, affective, and behavioral components . According to this definition, an attitude encompasses, not only what we feel about an object, but also what we think about it, and how we are inclined to behave toward it . Applying this definition to intergroup attitudes, stereotypes constitute the cognitive component ; ethnocentrism and prejudice, the affective component ; and discrimination in social relations, the behavioral component . Zanna and Rempel (1988) propose a substantial revision of Hovland and Rosenberg's definition, redefining attitude "as the categorization of a stimulus object along an evaluative dimension based upon, or generated from, three general classes of information : (1) cognitive information, (2) affective/emotional information, and/or (3) information concerning past behaviors or behavioral intentions" (p . 319) . There are three important characteristics of this definition of attitude that distinguish it from the multicomponent definition popularized by Hovland and Rosenberg (1960) . First, it considers attitudes as simple summary evaluations of an object or event along a single dimension such as good-bad or favorable-unfavorable. Second, because such summary evaluations are viewed strictly as "items of knowledge" (Zanna and Rempel 1988, p . 330), attitudes are reduced to cognitive judgments . In Zanna and Rempel's view, "evaluation and affect or emotion are not the same concept . . .emotions may be a component, or even the sole basis, of an evaluation, but the two need not occur together" (p . 320) . Third, as independently variable "sources" or "antecedents" of attitudes, cognition, affect, and behavior are not simply separate components of attitudes or different modes of response to attitude-objects or events, "they constitute very different ways in which the attitude is formed and subsequently experienced" (p . 320) .



Stereotyping and Intergroup Relations

43

The studies of Esses and colleagues (1993) and Stangor and colleagues (1991), discussed in the introduction to this paper, take Zanna and Rempel's (1988) definition of attitude as their starting point . Regressing single-item measures of group favorability on group-elicited emotions and stereotypic traits, both studies found that the latter are of negligible import in predicting intergroup attitudes . In our earlier discussion we attributed this counterintuitive finding to the fact that traits are disadvantaged as predictors of intergroup attitudes because both the attitude criterion and the emotions predictors are affective in nature . To their credit, Stangor and colleagues acknowledge that the use of a criterion measure that may be "particularly sensitive to the affective component of prejudice . . . might have provided an unfair advantage for the emotion variables" (1991, p . 371) . Moreover, both groups of researchers interpret their findings cautiously, disclaiming any intention "to minimize the role of stereotypes in intergroup relations" (Stangor et al . 1991, pp . 376-377) or "to suggest that it is not important to understand the cognitive component of prejudice" (Esses et al . 1993, p . 158 n15) . Despite these disclaimers, the fact that both studies found that stereotyped-based cognitions play little if any role in predicting intergroup attitudes begs the question, "how then do stereotypes feed into this process?" (Hamilton and Mackie 1993b, p . 375) . This question cannot be resolved by predictive models, such as those estimated by the Esses and Stangor studies, because such single-equation models consider only the direct effects of predictors . Identification of indirect effects requires the estimation of a multiple equation causal model (a path analysis model) in conjunction with an analysis of effects (Pedhazur 1997) . To their credit, Esses and colleagues acknowledge that their research "has not examined the causal connections between intergroup attitudes and their affective, cognitive, and behavioral components" (1993, p . 160) . Contemplating the possibility that "stereotypes play a more indirect role in determining intergroup attitudes," they speculate that "the characteristics. . . we attribute to members of a group may influence how we feel about these people and thus our evaluation of the group" (p . 152) . This suggests that a model containing the following specifications should have been estimated : (1) Stereotypic Traits --- P 21 ---4(2) Group-Elicited Emotions --- P 32 ---I(3) Intergroup Attitudes (1) Stereotypic Traits p31 *(3) Intergroup Attitudes The total effect of stereotypic traits on intergroup attitudes consists of its direct effect (p31 ) plus its indirect effect via group-elicited emotions (p21P32) . The total effect of group-elicited emotions on intergroup attitudes, on the other hand, is coextensive with its direct effect (p32) . To assess the relative effects of stereotypic traits and group-elicited emotions on intergroup attitudes, one must compare the total effects (direct plus indirect) of each independent variable . In contrast, an



44

NEIL J . MacKINNON and JEFFREY W . BOWLBY

assessment of relative effects employing a predictive rather than a causal model is restricted to a comparison of direct effects, leading to the seemingly anomalous findings of Esses and colleagues (1993) and Stangor and colleagues (1991) . In summary, studies of intergroup processes employing Zanna and Rempel's (1988) definition of attitude as a purely cognitive evaluation based on cognitive, affective, and behavioral components, in conjunction with the use of predictive models to determine the relative contributions of these components to intergroup attitudes, create a number of problems . First, the Zanna-Rempel definition is predicated upon the assumption that attitudes can be reduced to cognitive judgments pasteurized of affect . However, as Schuman (1995, p . 85n1) has observed, "it is difficult to imagine an evaluation of interest to sociologists [and, presumably, many psychologists as well] that is totally devoid of affect ." Moreover, Stangor and colleagues (1991) implicitly question this assumption by acknowledging the possible sensitivity of their attitude measure to the emotions predictor ; and Zanna and Rempel (1988) themselves acknowledge the difficulty of creating affectively free evaluations in experimental settings . Second, because predictive models do not address the causal relations among predictors, they mask any indirect effects these predictors may have on the criterion . This appears to be the case for the stereotypes predictor in the Esses and Stangor studies . Third, predictive models do not do justice to Zanna and Rempel's (1988) definition of attitude, which emphasizes that cognition, affect, and behavior are not simply constituent components, but rather "sources" or "antecedents," of attitudes . While Zanna and Rempel (1988) suggest the use of predictive models to determine the extent to which the relative effects of these three sources of attitudes change over time, they advocate causal experimental research to study their relative effects on attitude formation and change . Finally, neither the Zanna-Rempel (1988) model of attitude nor its application to intergroup attitudes by the Stangor and Esses studies incorporates the distinction between generalized or global affect and the specific emotions . As we shall see in the following section, this distinction is a critical one in affect control theory.

AFFECT CONTROL THEORY Affect control theory formalizes Mead's (1934) model of mind as an internal process of cybernetic feedback and control, and his conceptualization of social interaction as an ongoing process of mutually adjusted response among interactants (MacKinnon 1994) . Moreover, the theory accepts Mead's premise that language creates shared objects of consciousness through social categorization, and that these social cognitions are the basis of intersubjectivity and coordination in social interaction . Contrary to Mead's view that the affective associations of social cognitions are individual or idiosyncratic rather than social, affect control theory views these sentiments as more or less shared by members of the same culture or



Stereotyping and Intergroup Relations

45

subculture . These cultural sentiments, rather than the social cognitions that evoke them, are the raw materials for affect control theory ; and, in contrast to Mead's cognitive social psychology, affective rather than cognitive processing is its focus . Because "affective associations to social categorization generally are recognized as `attitudes"' (Heise 1979, p . 179n1), affect control theory draws heavily upon attitude theory and research for insights into affective processing . Operationally, the theory measures cultural sentiments by capitalizing on the established EPA (evaluation, potency, activity) structure of the semantic differential (Osgood, Suci, and Tannenbaum 1957 ; Osgood, May, and Miron 1975), which Osgood has interpreted as capturing the connotative aspect of meaning, "the `feeling tones' of concepts as part of their total meaning" (1969, p . 195) . Evaluation and potency, respectively, have been identified with the sociological dimensions of status and power (Kemper and Collins 1990) ; and, depending upon context, activity has been associated with the emotional energies of social interactants (Collins 1990), the expressiveness of identity-based roles (Heise 1988, p . 6), or the agency of participants in social interaction . We describe the specific scales employed by affect control theory to measure evaluation, potency, and activity in the method section of this paper . The EPA structure of connotative, affective meaning enables one to get a quantitative handle on the indeterminable content and dimensional complexity of denotative, cognitive meaning (Heise 1979, 1988 ; MacKinnon and Langford 1994) . Because they serve as generalized attitude scales (Osgood et al . 1957, pp . 195-198), EPA ratings also allow one to measure all kinds of social phenomenasocial identities, interpersonal behaviors, social settings, social characteristics, personality traits, and emotions-on a single, common metric . Moreover, EPA rating scales provide a mathematically coherent metric (Heise 1979, p . 50), enabling one to transform one type of phenomenon into another within the same semantic space . This is accomplished in affect control theory by applying the various mathematical models of the theory . For example, given the EPA profiles for the situated identities of social interactants, one set of equations generates their culturally expected behaviors . Another set of equations simulates the reidentification processes (labeling and trait attributions) that take place when the actions of participants belie culturally expected behavior . A third set of equations generates the emotions produced by interactional events, and so on . At the heart of these various models lie the principles of affective reaction and affect control. The first principle proposes that people react affectively to interactional events, experiencing transient feelings for the actor, behavior, and object person in each event . The second principle proposes that people construct and interpret events to confirm cultural sentiments for these event components . The discrepancy between established cultural sentiments and the transient feelings produced by events is called affective deflection in affect control theory . Restating the affect control principle in these terms, people construct and interpret interactional events to minimize affective deflection . Thus, besides advancing a model of



46

NEIL J . MacKINNON and JEFFREY W. BOWLBY

mind based on language, affect control theory incorporates Mead's conceptualization of mind as a cybernetic control system . In contrast to Mead, however, the theory applies the principle of control to affective (attitudinal) rather than cognitive (information) processing . The overall operation of the affect control model can be summarized in six steps (adapted with revisions from Heise 1979, p . 3) . (For a formal propositional statement and verbal exposition of the theory, see MacKinnon 1994 ; for a more technical, mathematical account, see Heise 1988 .) 1.

2.

3.

4.

5. 6.

Each person present in a situation (as actor, object person, or observer) engages in cognitive work to define the situation at hand, employing social categories (identities) to identify people present. Because social identities are associated with characteristic acts (e .g ., counsel and medicate for doctor ; listen and obey for patient), additional cognitive work enables one to narrow down the potential actions of each person . At the same time, the social categories identified by all this cognitive activity evoke cultural sentiments for the imputed identities and likely behaviors of people present . Events that are judged to be comprehensible within the cognitive framework established by the definition of the situation and additional cognitive work are tentatively selected for recognition . Events that come closest to confirming cultural sentiments for their constituent identities and behaviors are more likely to be selected for final recognition than those that do not . The recognition of events generates transient impressions or feelings about participants that may confirm or disconfirm cultural sentiments for their imputed identities and perceived behaviors . Discrepancies between transient feelings and established cultural sentiments (i .e ., affective deflection) engender the conceptualization of new events that would bring transient feelings back into line with cultural sentiments . If self is the actor in such conceptualized events, a behavioral intention or disposition is the outcome of this process ; if another participant is the actor, an expectation for that person's behavior . The realization of a behavioral intention or expectation through action creates a new event, looping the process back to (2) . If cultural sentiments for the identities of participants cannot be confirmed through restorative action (5), the process loops back to (1) instead, and a person engages in a redefinition of the situation on the basis of the most recent event . This higher-order feedback helps to stabilize the system . In view of the greater observability and/or finality of behavior, a redefinition of the situation generally focuses on reidentifying participants (MacKinnon 1994, p . 24) . This is accomplished by imputing new identities (labeling) or modifying the original identities through the attribution of explanatory traits .



Stereotyping and Intergroup Relations

47

This summary outline makes explicit three essential features of affect control theory . First, the theory specifies the relations among cognition, attitudes, and behavior, not as a simple causal chain, but rather as a control system in which people actively engage in maintaining their experience of social reality through corrective action or through reconceptualization of what has taken place . Second, the theory views the relation between cognitive and affective processes as complementary and interdependent phases of human consciousness . That is, the theory supposes that present cognitions evoke affective reaction and control which, in turn, stimulate new cognitions-the conceptualization of restorative events (see Step 4), for instance, or the reconceptualization of those that have been previously recognized (see Step 6) . Third, the theory acknowledges that attitude-behavior dynamics are "subordinate to a definition of the situation-that is, to a person's categorizations of people and objects in a scene" (Heise 1979, p . 2) . The theory does not predict the initial definition of the situation (Step 1), only its modification (see Step 6) . The original definition of the situation is provided by the researcher conducting affect control simulations of social interaction . In short, affect control theory begins with the affective associations or attitudes evoked by social categories, and models the attitude-behavior dynamics that follow . Because social categorization is the focus of social cognition theory, while attitude-behavior dynamics is the focus of affect control theory, this suggests a division of labor and a point of complementarity between the two theories . We return to this point in our general discussion . Absent from the summary outline of affect control theory just provided is the theory's model of emotions . In this regard, affect control theory distinguishes between the global affects of cultural sentiments, transient feelings, and deflection and the specific emotions experienced by participants in social interaction . Derivations from the empirically based attribution equations of the theory reveal that the latter are a consequence of two factors : first, the amount of affective deflection produced by events ; and, second, the extent to which events confirm or disconfirm the situated identities of participants (MacKinnon and Heise 1993 ; MacKinnon 1994 ; Smith-Lovin 1990) . The three studies with which we illustrate the application of affect control theory to stereotyping and intergroup relations can be located in the above outline of the theory . Study 1 employs only the cultural sentiments data (EPA measures of Canadian regional identities) (see Step 1) to test hypotheses about in-group versus out-group attitudes . Study 2 applies the attribution equations (see Step 6) to generate the characteristic traits of regional stereotypes . Study 3 employs the impression formation equations of the theory to generate the transient feelings (Step 3) and affective deflection (Step 4) produced by a set of a priori events, in order to study the affective outcomes of interaction between Canadians with different regional identities . It is important to stress, however, that simulations of intergroup relations can be conducted employing the full affect control model, including the affect control or impression management equations generating



48

NEIL] . MacKINNON and JEFFREY W . BOWLBY

interpersonal behaviors (Steps 4 and 5), and the equations predicting the specific emotions experienced by participants . While we have conducted many such analyses, considerations of space preclude reporting them here . With this brief statement of affect control theory in hand, we return to our previous discussion of stereotypes and intergroup attitudes to show how the theory addresses the issues raised there. While affect control theory accepts that stereotypes are identity schemas, it does not deal with the content of these cognitive structures directly, in the sense of measuring the traits attributed to members of social groups . In contrast to social cognition research, affect control theory takes the affective associations evoked by social identities (their EPA profiles) as its starting point . At the same time, the plausibility of affect control predictions suggests that EPA measures of social identities code much of the cognitive information embodied in identity schemas, including stereotypic beliefs about the characteristic traits and behavior of group members . For example, Study 2 shows how the stereotypic traits of Canadian regional identities can be recovered from their EPA profiles by applying the attribution equations of the theory . While the divergence between affect control and social cognition theories with respect to the concept of stereotype is more complementary than irreconcilable, it is a different matter when we turn to intergroup attitudes . In contrast to Hovland and Rosenberg's (1960) definition of attitude that includes cognition, affect, and behavior, affect control theory restricts the construct to affective meaning or response . Therefore, in contrast to Zanna and Rempel's (1988) definition of attitude, employed as a conceptual linchpin in the studies of intergroup attitudes by Esses and colleagues (1993) and Stangor and colleagues (1991), the theory does not accept that evaluations can be reduced to cognitive judgments devoid of affect. For affect control theory, affect is an essential property of attitude and cannot be eliminated by definitional fiat without destroying the construct as conventionally understood. Affect control theory recognizes that, ontologically, cognition and affect are inextricable modes of psychological experience, a matter of relative preponderance than mutual exclusiveness . Epistemologically, however, the theory recognizes the distinction between cognition and affect is an analytically useful one"particularly at the extremes of `cold' cognitions (where affective arousal is of low intensity) and `hot' cognitions (where affective arousal is more pronounced) ; or, alternatively, at the extremes of affective experience largely unmediated by cognition and that conjoined with a high level of cognitive awareness and appraisal" (MacKinnon 1994, p . 45) . From this perspective, beliefs are more cognitive than affective ; and attitudes, more affective than cognitive . Moreover, some attitudes are hotter than others-that is, of greater intensity in either direction on the bipolar scales employed to measure them . For example, while a bigot evokes intense negative feelings (-2 .34) on the evaluation scale for Canadian male respondents, and a charity worker, intense positive feelings (2 .48), other identities evoke affective associations that are closer to neutral on the evaluation scale (e .g .,



Stereotyping and Intergroup Relations

49

.39 for a customer) . (For an extensive discussion of the relation between cognition and affect as debated in the psychological literature, and the implications of this issue for affect control theory, see MacKinnon 1994, pp . 41-50 .) Thus far, we have discussed attitude as the affective response to objects in terms of positive or negative evaluation (good-bad, favorable-unfavorable, and so on) . However, following the EPA structure of affective meaning established by Osgood and associates (Osgood 1969 ; Osgood, May, and Miron 1975), affect control theory extends the domain of attitude to feelings of potency and activity as well . Whether viewed in terms of E or EPA, restricting the construct of attitude to affective response analytically separates intergroup attitudes (ethnocentrism and prejudice) from their presumed antecedents (stereotype-based cognitions) and behavioral consequences (social discrimination) . As acknowledged by Zanna and Rempel, this one-component definition of attitude has the advantage of making the consistency between attitudes and behavior "a theoretical or even empirical question rather than a definitional necessity" (1988, p . 316) . Besides rejecting Zanna and Rempel's reduction of attitudes to cognitive judgments devoid of affect, affect control theory does not share the view of these authors that emotions are a direct source of attitudes . Instead, as discussed above, the theory treats emotions as a product of the more global affective dynamics of identity processes in social interaction . On the other hand, the theory's emphasis on the role of action in attitude-formation and change (Heise 1979, p . 38) is consistent with Zanna and Rempel's identification of past behavior and behavioral intentions as important sources of attitudes . As discussed above, affect control theory views the relation between attitudes and behavior as a reciprocal process of cybernetic feedback and control, rather than as one-way causal process . While this model of attitude resonates with the model proposed by Fishbein and Ajzen (1975), there is an important difference in emphasis . In the Fishbein and Ajzen model, the feedback from behavior to attitude is more incidental than designed . In the affect control model, the discrepancy between fundamental attitudes and the transient attitudes produced by events "generate behaviors designed to create new feelings that confirm fundamental attitudes" (Heise 1979, p . 39) . This model of attitudes sheds some light on the empirically tenuous link between attitudes and behavior that continues to plague attitude theory in general (Hill 1981 ; Schuman 1995) and theories of intergroup processes in particular (Mackie and Hamilton 1993b ; Seeman 1981 ; Zanna and Rempel 1988) . According to the affect control model of attitudes, "behavior is not generated directly from either fundamental or transient attitudes but rather from the discrepancy between the two ." Hence, "an attitude as such does not ordinarily generate any typical behavior" (Heise 1979, pp. 38-39) . In summary, affect control theory provides a powerful model for exploring the relationships among stereotypes, intergroup attitudes, and intergroup relations, an explanatory model of intergroup processes based on attitude-behavior dynamics and affective processing . From the perspective of affect control theory, ste-



50

NEIL J . MacKINNON and JEFFREY W . BOWLBY

reotypes are identity schemas deriving from membership in or ascription to social groups ; intergroup attitudes are the cultural sentiments evoked by group identities ; and stereotyping itself exemplifies the identification processes that underlie all social interaction . Thus, the theory models stereotyping and intergroup relations in the same way that it models identification processes and social interaction in general . When people act in accord with their group identities, affective deflection is minimal and interaction between members of social groups proceeds in stereotypic fashion . When the actions of group members deviate from stereotype-based expectations, the motivation to minimize affective deflection mobilizes restorative actions to repair the violation of the "expressive order" (Goffman 1967) defined by intergroup attitudes . In the event that the deflection produced by disruptive events cannot be resolved through restorative action, stereotypes themselves may become modified through the attribution of explanatory traits, or discarded altogether through the imputation of entirely new identities .

METHOD The general methodology of affect control research is available from a number of published sources (e .g ., Heise 1978 ; Smith-Lovin and Heise 1988), enabling us to restrict our discussion to details specific to the research reported here . In this section we describe the sample, instrument, and measurement procedures employed in this research . We discuss other aspects of research design, including data analysis procedures, in the specific sections dealing with each of our three studies . Sample Data on cultural sentiments for regional identities were collected as part of a larger study conducted in 1995, employing convenience samples of undergraduate students enrolled in introductory sociology methods courses at an Ontario and an Alberta university . Each of the two samples consisted of approximately 30 males and 30 females born and/or raised in Canada . Instrument Our survey instrument was introduced by a covering page, providing instructions to respondents and asking them to provide information on their gender and the country in which they were born and/or raised . The body of the survey consisted of 11 pages of word stimuli pertaining to social identities and interpersonal behaviors . Each word stimulus was followed by the evaluation, potency, and activity semantic differential scales described in the next section .



Stereotyping and Intergroup Relations

51

Measurement We measured the affective meaning of regional identities and interpersonal acts with the evaluation, potency, and activity scales developed in past affect control research . The evaluation scale is anchored by "bad, awful" to "good, nice" ; the potency scale, by "small, weak, powerless" to "big, strong, powerful" ; and the activity scale, by "slow, old, quiet" to "fast, young, noisy ." Calibrated from -4 to +4, each scale captures a wide range of negative to positive valence, from infinitely "bad, awful" to infinitely "good, nice," for example . Actual values generally fall between ±3, with a ±2 considered a large (positive or negative) value . Following established procedures in affect control research, estimates of cultural sentiments were obtained by aggregating individual scores on EPA scales . As discussed above, affective deflection (the dependent variable in Study 3) is conceptualized as the discrepancy between the fundamental cultural sentiments for social identities and interpersonal behaviors outside the context of any event and the transient impressions or feelings for these components created by their combination in ABO (Actor-Behavior-Object) events . Thus, deflection can be viewed as a kind of global affective response to interpersonal events, employing fundamental cultural sentiments as a point of reference . Operationally, deflection is measured as the sum of the squared discrepancies between fundamental sentiments and transient impressions for the ABO (Actor-Behavior-Object) components of events on the EPA (evaluation, potency, and activity) dimensions of affective meaning : Deflection = E [(A e - Ae' )2+ (Ap - Ap')2 + (A a - Aa ) 2 + (B e - B e' )2 + (BP - Bp' )2 + (B a - Bay )2+ We - Oe ')2 + (op -O p ')2 +(Oa -Oa ) 2 ]

Unprimed terms in this expression denote fundamental cultural sentiments ; primed terms, transient impressions created by events . Subscripts (e, p, a) denote evaluation, potency, and activity . Squaring discrepancies avoids the cumbersome algebra of absolute differences and provides greater weight to larger discrepancies than to trivial ones . The fundamental sentiments in the deflection measure are simply the mean EPA values of identities and behaviors measured on the semantic differential scales described above . The transient impressions or feelings created by events analyzed in Study 3 were generated from the empirically based impression formation equations of affect control theory, employing estimates of parameters from earlier Canadian research (MacKinnon 1985) . Because these equations are far too complex to discuss here, we refer the interested reader to Heise (1988) and Smith-Lovin (1988) for technical detail .



52

NEIL J . MacKINNON and JEFFREY W . BOWLBY

STUDY 1

The first study explores two classic hypotheses from social cognition research on stereotyping, employing Alberta and Ontario data on cultural sentiments for Canadian regional identities . According to the in-group bias or ethnocentrism hypothesis, people tend to evaluate their own group more favorably than out-groups . According to the out-group homogeneity hypothesis, people tend to perceive members of an out-group as more stereotypic and less diverse than members of that group see themselves (Judd and Park 1995) . This study differs from past research on these hypotheses in two important ways . First, we apply the ethnocentrism and out-group homogeneity hypotheses to intergroup attitudes (cultural sentiments) rather than to intergroup perceptions (attributed traits), the usual focus of social cognition research on these hypotheses (e.g ., Judd and Park 1995) . Second, we explore whether the ethnocentrism effect generalizes from evaluation (the dimension of attitude emphasized by social cognition research) to the potency and activity dimensions of affective response . On the other hand, by studying the attitudes of both Albertans and Ontarions, we maintain the intergroup perspective of more recent work on stereotyping by social cognition researchers (e .g ., Allen 1996 ; Judd and Park 1995) . Findings In-group Bias/Ethnocentrism Hypothesis

Evidence for this hypothesis within each sample is provided by comparing mean attitudes for regional identities down columns in Table 1 . For example, the mean evaluation of male Alberta respondents for an Albertan (2 .01) is higher than its evaluation of all other regional identities-Maritimer (1 .74), Ontarion ( .42), and Quebecer (- .30) . This pattern prevails for female Alberta respondents and for Ontario respondents of both genders, with one minor exception-Ontario respondents evaluate a Maritimer as a bit nicer than an Ontarion . We do not report tests of significance for within-sample comparisons of means because such comparisons probably violate the assumption of statistically independent evaluations (Blalock 1979, p . 141) . The ethnocentrism hypothesis is tested across samples by comparing the mean evaluations of Alberta and Ontario respondents for their respective regional identities, Albertan and Ontarion . Although the two samples are not random, we employ tests of significance here as a heuristic device to discriminate between the trivial and the possibly important . As revealed by across-row comparisons in Table 1, Alberta respondents evaluate an Albertan more highly than Ontario respondents do (2 .01 versus .76 for male respondents ; 1 .84 versus .98 for female respondents) ; and Ontario respondents evaluate an Ontarion more highly than Alberta respondents (1 . 16 versus .42 for male respondents ; 1 .12 versus .66 for



•

•~o

10 P, lD V n N M

•

O

O

r u M

M O M N N 10 00 h

C}

0 R N

O

m L N -D

O 0 n M N N N 0-c

Q

•

E • C/') •

N ;! M n N 0~ 0C

Q

n .- O M Z3 O ~ iD N r

Q

00o~v v C v M C-

• •

00 N to M N ' M M r ' N N

C ro

N

U

O

U N C

0-c O cO cO 7 V ~ V'

C

0

O 2 0C in N N

a>

V

0

ro

11

N N n O ~o co

Q1

-Q

N O N M l0 M 00 r

Q

Q

x x M V ~O 0~ M - r C O

C O

0 W

0

O

W

O y C O

co 0-~ N cO O~ M r O

ro

rq

aN

V N O O n V M N

0-c 10 00 v .O 00 cc co V Q

C

O C N E O N Y N

•

.p Q

'0 0

C V Q

Q~E

C

0

ooa

._E C N Q) Q :2 oa

53

O z



54

NEIL J . MacKINNON and JEFFREY W. BOWLBY

female respondents) . While the latter comparison falls short of statistical significance, the overall pattern of differences supports the in-group bias/ethnocentrism hypothesis for the evaluation dimension of affective meaning . Extending the ethnocentrism hypothesis to the potency dimension of affective meaning, both male and female Ontario respondents view an Ontarion as more powerful than all other regional identities . However, contrary to the hypothesis, male Alberta respondents view an Albertan as less powerful than an Ontarion, and female Alberta respondents view an Albertan as less powerful than either an Ontarion or a Quebecer. Applying the hypothesis to across-sample comparisons, Alberta respondents view an Albertan as more powerful than Ontario respondents do (1 .12 versus .49 for male respondents ; .62 versus .29 for females), albeit the latter difference does not achieve statistical significance . While these across-sample differences support the ethnocentrism hypothesis for Albertan as the in-group target identity, across-sample comparisons for Ontarion as the in-group target identity does not . In fact, Alberta respondents rate an Ontarion as more powerful than Ontario respondents rate themselves (1 .67 versus 1 .16 for males ; 1 .82 versus 1 .15 for females) . Taken together, these findings provide only limited support for extending the ethnocentrism hypothesis to the potency dimension of affective response . Within-sample (down-column) comparisons for the activity dimension of affective response also reveal only limited support for the ethnocentrism hypothesis . Consistent with the hypothesis, Ontario respondents of either gender see themselves as more lively than Albertans, Maritimers, or Quebecers, and male Alberta respondents also view themselves as more lively than Canadians from other regions . However, contrary to the hypothesis, female Alberta respondents see themselves as less lively than either an Ontarion or a Quebecer . Moving to across-sample comparisons, Albertans view themselves as significantly more lively than Ontario respondents do (1 .20 versus .11 for males ; .72 versus - .23 for females) . Although these comparisons support the hypothesis for Albertan as the in-group target identity, those for Ontarion as the in-group target identity are small, statistically insignificant, and inconsistent in direction across gender . The Out-Group Homogeneity Hypothesis

Following the statement of this hypothesis by Judd and Park (1995)-in terms of intergroup perceptions or mutual stereotyping, we explore the hypothesis with across-sample comparisons of intergroup attitudes for Albertan and Ontarion as out-group target identities . We use standard deviations to measure the homogeneity of group attitudes, smaller values indicating greater homogeneity . We do not employ tests of significance for differences of standard deviations across samples, even as a heuristic device, because such (two-sample F) tests are notoriously unrobust in the presence of even slight departures from normality (Moore and McCabe 1993 ; Pearson and Please 1975) . Therefore, differences in standard devi-



v

0 rtl C O

C O

0 O O

~O N

a

U Q

fTS C

O OA

L O y w n C

n 00 C~ Ol 0 O~ N

U Q N .Q Q

aJ

ro G O

co -It ~o ~Y I~ -~ In

v

N C N M M O M

Q

0

cz Cn O c

W N N O O~ O N

cu

c a U

L

N O

C

L

C

O ~ n O

C O

C O N

Q1

v C

C ro

G (li

W

m

ro

O N M I. n V N C

C Q)

N

Q

Q

t

Q m O

O L V O

. n N 0 O O

O ~

C O ro W

G O ro ro W

m Q!

W Ln ~O N ,I: N M O~

Q

O

u a)

M -Q N N N N

Q

U N C ro

c

E .0 .ro

Q

C

ci

c Boa

c 0

55

y

C E .o ro

U

Q~oa



56

NEIL) . MacKINNON and JEFFREY W . BOWLBY

ations between samples must be interpreted strictly as sample-bound, descriptive data which we employ to explore rather than test the homogeneity hypothesis . As reported in Table 2, Ontario respondents view an Albertan as more homogenous on evaluation than Alberta respondents themselves (1 .14 versus 1 .48 for males ; .87 versus 1 .21 for females . Ontario respondents also view an Albertan as more homogenous on potency than Alberta respondents ( .98 versus 1 .49 for males ; .70 versus 1 .36 for females), and more homogenous on activity as well (1 .00 versus 1 .41 for males; .97 versus 1 .31 for females) . Thus, for the Ontario sample as the in-group and Albertan as the out-group target identity, these findings support the out-group homogeneity hypothesis across all three dimensions of intergroup attitudes . In contrast, our findings reveal a reversal of the out-group homogeneity effect for the Alberta sample as the in-group and Ontarion as the out-group target identity . That is, Alberta respondents view an Ontarion as less homogenous than Ontario respondents themselves on all three dimensions of affective meaning . For evaluation, the relevant comparisons yield 1 .36 for Alberta male respondents versus 1 .04 for Ontario male respondents (1 .26 versus 1 .07 for females) ; for potency, 1 .43 versus 1 .17 (males) and 1 .25 versus 1 .17 (females) ; and for activity, 1 .64 versus 1 .06 (males) and 1 .09 versus .99 (females) . Discussion This study applied the ethnocentrism and out-group homogeneity hypotheses from social cognition research on stereotyping to Alberta and Ontario attitudes for Canadian regional identities . While both within- and across-sample comparisons support the ethnocentrism hypothesis for evaluation, the results are mixed when the hypothesis is extended to the potency and activity dimensions of intergroup attitudes . Within-sample comparisons on potency support the hypothesis for only the Ontario sample . In across-sample comparisons on potency, the ethnocentrism effect was found for Albertan as the in-group target identity, but a reversal of the effect was found for Ontarion as the in-group target identity . With the exception of Alberta females, within-sample comparisons support the ethnocentrism hypothesis for the activity dimension of intergroup attitudes . On the other hand, across-sample comparisons on activity reveal the ethnocentrism effect only for Albertan as the in-group target identity . Our findings support the out-group homogeneity hypothesis for the evaluation, potency, and activity dimensions of intergroup attitudes, but for only the Ontario sample as in-group and Albertan as the out-group identity . In fact, for all three dimensions of intergroup attitudes, we found a reversal of the out-group homogeneity effect for the Alberta sample as in-group and Ontarion as the out-group identity . The group-specific nature of the ethnocentrism and out-group homogeneity effects revealed by this study resonates with the findings of Judd and Park (1995) .



Stereotyping and Intergroup Relations

57

Eliminating rival hypotheses through a sequence of carefully designed studies and sophisticated analyses, these researchers accumulated support for a differential socialization explanation of their finding that African-American university students manifested ethnocentrism and in-group homogeneity effects while white American students did not . Specifically, Judd and Park (1995) argue that young black Americans have been socialized to believe that ethnic group membership is a meaningful social category and social identity, to be emphasized in the ongoing struggle for social equality, while young white Americans have been socialized to avoid making distinctions on the basis of ethnic group membership . While the intensity of interregional relations in Canada pales by comparison to the intensity of racial conflict in the United States, the differential socialization hypothesis proposed by Judd and Park (1995) may help to explain some of our findings-specifically, the greater ethnocentrism effect for the Alberta sample on evaluation and the manifestation of this effect on potency and activity for the Alberta sample only . Historically, Western Canada has been an economically disadvantaged region relative to Central Canada ; and although this disparity has narrowed in recent decades, intergroup attitudes may lag behind economic change . If true, we would expect the intergroup attitudes of young Albertans to manifest greater in-group favorability than those of young Ontarions, the kind of defensive ethnocentrism described by Judd and Park (1995) . While this line of reasoning helps to explain why the ethnocentrism effect is stronger for the Alberta sample, it does not shed light on why a reversal of the out-group homogeneity effect was found for Albertans (historically, the economically disadvantaged in-group) with Ontarion as the out-group target identity ; nor, by the same token, does it explain why the out-group homogeneity effect was manifested only for Ontarions (historically, the economically advantaged group) with Albertan as the out-group target identity . However, while many studies have revealed that in-group members perceive out-group members as more homogeneous and less variable than themselves, other studies reveal that this is far from a universal tendency (see Lee and Ottati 1995 for references) . Moreover, the form in which we have stated and tested the out-group homogeneity hypothesis in this study employs the in-group homogeneity of the target group itself as a basis for comparison. Therefore, the in-group homogeneity of target groups becomes relevant to explaining our findings with respect to the out-group homogeneity hypothesis . In this regard, Lee and Ottati demonstrate experimentally how cognitive and motivational factors affect in-group homogeneity, suggesting that "instead of being static, perceptions of in-group variation appear to fluctuate across time and circumstances" (p . 610) . The issues here are complex and we can only speculate as to why Ontarions exhibit greater in-group homogeneity for their group identity than Albertans do for theirs (see Table 2) . However, since the purpose of this paper is to illustrate the application of affect control theory to stereotyping and intergroup relations, rather than to debate the relative merits of any



58

NEIL J . MacKINNON and JEFFREY W. BOWLBY

particular explanation, we move on to the next study without additional commentary .

STUDY 2 As discussed in our description of affect control theory, the measurement of all kinds of social phenomena on EPA scales enables one to transform the affective meaning of one type of phenomenon into another within the same semantic space . This study illustrates how stereotypic traits can be generated from social identities by employing the attribution equations of the theory . In affect control theory, traits are treated as descriptive attributions of a person with a particular social identity . The procedure of recovering characteristic or stereotypic traits from social identities involves simulating a social event that perfectly confirms a particular social identity, then applying the attribution equations to generate a numerical EPA profile for the kinds of traits that might be attributed to a person with this confirmed identity . With this numerical EPA profile in hand, the cultural data base for identity modifiers is searched for traits that optimally fit this test profile . This procedure is based on the notion that when events perfectly confirm a person's social identity, the traits attributed to that person become a function of the cultural sentiments for that identity (see MacKinnon 1994, p . 128 for a discussion of this rationale applied to characteristic or stereotypic emotions) . For example, when an event (say the delivery of a flawless, enthusiastically received lecture) confirms a professor's identity, observers of this event might consider that professor to be studious, experienced, dignified, tactful, practical, responsible, reasonable, and mature (predictions from Canadian equations and cultural sentiments) . Table 3 reports the characteristic or stereotypic traits generated by the attribution equations of affect control theory for the four regional identities employed in Study 1 . Each list of traits is preceded by the numerical EPA profile generated by the model and trait retrievals are ordered according to their fit to this test profile . For example, the numerical EPA profile (2.8 1 .9 .7) suggests that the characteristic traits attributed to an Albertan by male Alberta respondents would be extremely positive, quite powerful, and slightly lively in nature friendly, loved, employed, confident, competent, talented, and reliable. Those attributed to an Albertan by male Ontario respondents (1 .2 1 .1 .1) are much lower on evaluation and notably less on potency and activity as well-conscientious, modest, tactful, sophisticated, poised, accommodating, and practical . The difference in stereotypic traits attributed to an Albertan by male Alberta and Ontario respondents stem from differences in their cultural sentiments for this social identity . Because male Alberta respondents view themselves as much nicer and more powerful and lively than male Ontarion respondents do (see Table 1), they make self-attributions that are more positive, powerful, and lively .



Stereotyping and Intergroup Relations Table 3 .

59

Stereotypic Traits of Regional Identities for Alberta and Ontario Samples Males

Regional Identity

Albertan

Ontarion

Maritimer

Quebecer

Alberta

Females Ontario

Alberta

Ontario

(2 .8 1 .9 .7)

(1 .2 1 .1 .1)

(2 .31 .4 .4)

(1 .2 1 .0 - .1)

friendly loved employed confident competent talented reliable

conscientious modest tactful sophisticated poised accomodating practical

easygoing helpful worthy trusting appreciative cooperative intellectual

poised virtuous cautious contented modest sophisticated warm

(1 .5 1 .8 .6)

(2 .01 .7 .7)

(1 .82 .0 .7)

(1 .9 1 .6 .5)

witty wealthy intellectual rich persuasive perceptive successful

original bright efficient self-discipl talented confident wealthy

confident clever self-discipl perceptive rich wealthy independent

intellectual worthy self-discipl clever respectful charming creative

(1 .3 .5 - .6)

(1 .4 .7- .2)

(1 .4 .8- .7)

(1 .3 .7- .4)

wholesome noble humble religious discrete cautious

accomodating earnest modest noble agreeable

wholesome accomodating modest earthy humble reverent sentimental

modest earthy accomodating agreeable humble cautious poised

( .71 .1 .8)

( .7 .9 .7)

(1 .41 .5 .7)

(.7 .9 .6)

modern well-bred provocative demanding talkative suave idealistic

suave lucky modern clean-cut demanding white normal

well-bred charming married witty creative rich earnest

sensuous idealistic masculine sophisticated suave normal candid

The numerical EPA profile for the characteristic traits of an Ontarion is more similar for male respondents from each province, differing only in the higher evaluation of traits (2 .0 versus 1 .5) attributed by Ontario males to themselves . Interestingly, considering the historical status of Ontario as the most prosperous Canadian province, Alberta respondents view an Ontarion as wealthy, rich, and successful . In contrast, the self-attributions of Ontario respondents emphasize talent, intelligence, and efficiency-original, bright, intellectual, clever, efficient, self-disciplined, and talented, depending on the gender of respondent . The EPA profile generated by the attributions equations for a Maritimer is quite similar across both province and gender, denoting traits that are positive, weak,



60

NEIL J . MacKINNON and JEFFREY W. BOWLBY

and unlively in nature-wholesome, noble, modest, humble, accommodating, agreeable, earnest, discrete, cautious, religious, reverent, and sentimental. Again, these traits are consistent with the positive, weak, and unlively cultural sentiments for a Maritimer reported in Study I (see Table 1) . With the exception of Alberta females, the EPA profile for the stereotypic traits of a Quebecer is also quite similar across samples, resulting in trait retrievals that are slightly good, powerful, and lively-modern, suave, well-bred, and the like . In the case of Alberta females, whose EPA profile is much higher on evaluation and a bit higher on potency, retrievals also include such traits as charming, witty, and creative. In view of the resurgence of Quebec nationalism in the last two decades, and the publicly expressed frustration by other Canadians to understand the aspirations of Quebecers, the most interesting trait retrievals are those that describe a Quebecer as provocative, demanding, and idealistic. Again, these trait retrievals reflect the cultural sentiments for the regional identity at hand . As reported in Study 1 (see Table 1), a Quebecer is viewed as slightly negative to slightly positive on evaluation, and as low to slightly positive on potency and activity, depending on the gender and province of the respondent . Discussion

This study demonstrates how the stereotypic traits attributed to Canadians from different regions of the country can be generated from the cultural sentiments for regional identities by applying the attribution equations of affect control theory . Since cultural sentiments for regional identities differ by region, the stereotypic traits attributed to these identities differ across region as well . The differences in trait attributions between Alberta and Ontario samples revealed by our findings are plausible enough, judging by our knowledge of Canadian culture . However, because stereotypic traits were not measured directly in this study, our findings should be considered hypotheses for more conventional research in the social cognition tradition . We return to this point in our general discussion of findings .

STUDY 3 This study explores the affective outcomes of intergroup relations between Albertans and Ontarions from the cultural perspective of each group . Design and Analysis

The units of analysis in this study are interpersonal events . The dependent variable is affective deflection, described in the method section of this paper. We regress the affective deflection produced by events on the nature of the act (its evaluation, potency, and activity), the type of event (Ontarion > Albertan and



v -

v rou v

ra v

o c

Q b o >

£ >

a) N ro

a 22

E

v

v

p A E u y L

y

O_

C

-N

R1 E

c i

3

o

aN

L

L

C

a

o p

v a) v z

v

v v

;9 E ro v -o .-

E- 'ro

0 0E W

0 3 v

O

E u

ao . c '

Y O

O N V' 0 r I

Ln c N W ~ I

W M r

N V rn cc M I r I

v °

v

co °

v

v

O '-

M N

O cC O N

V I

C cO O Cn cC I

•

I + +

+ I +

+ +

w co

W W

61

cd L

v m v

N

v av v = v = ro ro v =m v = ro v = ro yv u ro E ra E ra E ro E ra E a E ro E ro E E a? E- E .v E- E 01 E- E- E-

I I +

ft

•E

L

o v

•pN-n y0 E `~ O rp L v _ a 0 v u O- 2 f_ v O U° 2 v E -O 7 C Y a p C o

• •

c

y v °

v

v u `-o

73

0 c

° Q E '~ ao avo c v ro w ° v vE a °c C a - bO

°v N a w N cn 0 E v ° o > o v v u ° o c ro

a

-'

v a

a s

4 C

v

a

m ° v

a v ._ ro yo v v -a a C - .N a u ai `°

ro

WD bu> >. w 3O ..Q Y v 3 N N D

o o u

m

v4

C v a Cv ce C ro a) 4O L Q W



62

NEIL J . MacKINNON and JEFFREY W . BOWLBY

Albertan > Ontarion), and interactions between these two sets of predictors . The analysis is replicated across Alberta and Ontario samples to determine the effect of cultural perspective on the affective outcomes of intergroup events . In one analysis we employ the cultural sentiments of the Alberta sample for an Albertan and an Ontarion ; in the other those of the Ontario sample for the two regional identities . In the dummy variable analysis for the Alberta sample, we employ Albertan > Albertan events as the reference category ; in the analysis for the Ontario sample, Ontarion > Ontarion events . This enables one to compare the effect of out-group to in-group interaction within each analysis, To ensure maximum variation in behavior for the events analyzed in this study, we searched our Canadian database for behaviors having EPA profiles closest to the eight possible configurations of low (-) and high (+) EPA values listed in Figure 1 . The empirical EPA profiles for behaviors available in our database that come closest to these ideal EPA profiles are reported in Figure 1, along with representative behaviors . In some cases empirical EPA profiles only approximate ideal EPA configurations because of the unavailability of matching behaviors . For instance, there are no behaviors that are simultaneously high on evaluation and low on potency . As a consequence, empirical EPA profiles are close to neutral on potency for (+ - +) and (+ --) configurations ; and, because behavior is inherently lively, a number of other empirical EPA profiles are close to neutral on activity . Behaviors were selected from those listed in Figure 1 to construct the events analyzed in this study . A cross-classification of these eight types of behaviors by the three types of events described above yields 24 ABO (actor-behavior-object person) events to analyze for each gender, or 48 events in all when pooled across gender . These 48 events were analyzed separately for the Ontario and Alberta samples, yielding parallel equations describing the affective outcomes of intergroup relations from the cultural perspective of each group . Employing ordinary least squares regression, we performed a dummy variable analysis for each sample to estimate the effects of event-type (D1 : Ontarion > Albertan events ; D2 : Albertan > Ontarion events), the evaluation, potency, and activity of the behavior in each event (BE, BP, and BA), and the interactions between event-type and the nature of the act (D1BE, D1BP, D1BA and D2BE, D2BP, D2BA) . As explained above, in-group interaction (Ontarion > Ontarion events in the Ontario equation ; Albertan > Albertan events in the Alberta equation) was defined as the reference category, thus omitted from the equation for each sample . In preliminary analysis, we included a dummy variable for the gender of respondent and tested its main and interaction effects with event-type and behavior evaluation, potency, and activity . Because neither main nor interaction effects were found, gender was deleted from the regression analysis . Thus, the general equation for the dummy variable analysis reported in this study is as follows :



O OO OO OO ON OO OC ~0 O M N O CL O 0 0 0 0 0 0 7 C~M OO N;-o0N OC

E

U > wQ) LoM m > U

'O t M N t0 0~ r t N ~O a1 Ct CO of WN MN ~0t t~OW Oo0 ON'D N NN t '°N -N`° I I ~ OM ro CW7ro`v co IO u)t c in t cc rn N M u1 u1W Co u)t 10 N NM~7n C4 . N r-N Q N N N

C C U > le Q) WC CL NO L

tIn NLnN~o l0 l0 N O1 M 10 M h N r, 10 u1 inC0O tItN ~oM 0.-C~M NN CN ~n I I N

0 O 0 II d Ma

fl II C 0 7 0Q r. II C OO O

c3 .Nco0 0^ M OO CL 0O 0O 0O 0O 0O Mn co 0 r-

C O Q U • N •

GO

N . N 1 Ln M un N 0 ~ON~O O O t M t O IM C-4 IDLn O O N M C~ M M t N ~-- O

W O • Oc ~c M u7 u1 l0 u) n l0 V) Q-'D•O O O O O M t O M t 7

t NCcotr.000 NO t ..O u1 t ~O t MLnM r.M mO NO MO mn nM NN u1 t I r-i I I M • •

c a• -Q v •vv .~~ a _ S

c w o Q m C C N N C c cc) co co 0 C) Q

63



64

NEIL J . MacKINNON and JEFFREY W. BOWLBY

Deflection' = a + b1D1 + b 2 D2 + b 3 BE + b4BP + b 5 BA + b6D1BE + b 7 D1BP + b 8 DIBA + b 9 D2BE + b 10D2BP + b1 1 D2BA For predicting the affective deflection produced by a particular type of event (D 1 : Ontarion > Albertan or D2 : Albertan > Ontario), this equation reduces to the constant (a), the main effect of event-type (either D1 or D2) ; the evaluation, potency, and activity of the behavior (BE, BP, and BA) ; and three terms (either D1BE, D1 BP, D1BA or D2BE, D2BP, D2BA) representing the interaction of event-type with the nature of the act. Findings Overall Equation

As reported in Table 4, the overall equation explains 97 percent of the variance in deflection for the Alberta sample (F = 96 .983 ; p = .000) and an equivalent amount of variance for the Ontario sample (F = 98 .741 ; p = .000) . Behavioral Dynamics

In this section we deal with the main effects of behavior evaluation (BE), potency (BP), and activity (BA), net of event-type (Dl or D2) and interactions between event-type and the nature of the act (D1BE, D1BP . . . D2BA) . For both samples, behavior evaluation, potency, and activity manifest sizable and significant effects on deflection (-2 .567, p = .000 ; .656, p = .000 ; - .406, p = .020, respectively, for the Alberta equation ; -1 .348, p < .001 ; .354, p < .001 ; - .377, p < .001 for the Ontario equation) . The direction of these effects in both equations indicates that nice and lively behaviors between Albertans and Ontarions reduce affective deflection while powerful actions increase it . While the pattern of effects is similar across equations, the effects of behavior evaluation and potency are notably larger in the Alberta equation . The main effects of behavior evaluation, potency, and activity are modified by a number of interactions with event-type-D1BE and D1BP for the Alberta equation, D2BE for both equations, and so on . While one can interpret these interactions in terms of event-type moderating the main effects of behavior, the principal concern of this study is with the effect of event-type (Ontarion > Albertan versus Albertan > Ontarion) on the affective deflection produced by intergroup relations . Therefore, we delay our discussion of interaction effects until the following section, where we interpret them in terms of the nature of the act (BE, BP, and BA) moderating the amount of deflection produced by event-type (D1 and D2) .



Stereotyping and Intergroup Relations

65

Effects of Event-Type on Deflection The main effects of BE, BP, and BA have been discussed in the preceding section . Because the pattern of these effects is similar in the general equations for Alberta and Ontarion samples, we ignore them here in order to simplify our presentation of findings regarding the effects of event-type on affective deflection . Dl : Ontarion > Albertan Events Alberta equation : Deflection' = 5 .571 - 1 .264D1 + 1 .246DIBE - .367DIBP Ontario equation : Deflection' = 3 .454 - .454D1 Recalling that the reference categories for the Alberta and Ontario equations are, respectively, Albertan > Albertan and Ontarion > Ontarion events, the negative coefficient for D1 in both equations indicates that out-group interaction involving Ontarion >Albertan events produces less deflection than in-group interaction . While the main effect of D1 events is about three times larger in the Alberta equation (-1 .264 ; p = .000) than in the Ontario equation (- .454 ; p = .000), it is moderated by behavior evaluation (BE) and potency (BP) in the Alberta equation . The sign of the D1BE effect in the Alberta equation suggests that the nicer the actions of an Albertan toward an Ontarion, the larger the deflection produced by D1 (Ontarion > Albertan) events . Considering the size of this effect (1 .246 ; p = .000), positively evaluated actions by an Ontarion could easily wash out the reduction in deflection (-1 .264) produced by the main effect of D1 events in the Alberta equation . Recalling our discussion of Table 1, both Albertan and Ontarion are positively evaluated identities for Alberta respondents, albeit they evaluate an Albertan as much nicer than an Ontarion . The impression formation equations of affect control theory reveal that unexpectedly good actions produce transient impressions that enhance the fundamental status of both actor and recipient . Therefore, unexpectedly nice actions in Ontarion > Albertan events produce sizeable deflections because such events create transient impressions that exceed the fundamental positive status of both participants . The large positive effect of D1BE in the Alberta equation simply captures this property of the evaluation dynamics of affect control theory . The potency of the act also moderates the main effect of D1 (Ontarion > Albertan) events in the Alberta equation . In this case, the negative coefficient for the D1BP term (- .367 ; p = .10) reveals that powerful actions by an Ontarion enhance the reduction in deflection produced by D1 (Ontarion > Albertan) events . As reported in Table 1, Alberta respondents view Ontarions as more powerful than themselves ; and according to the impression formation equations of affect control theory, powerful actions maintain the fundamental potency of powerful actors and the relative impotency of weaker recipients . Thus, deflection should be reduced



66

NEIL J . MacKINNON and JEFFREY W. BOWLBY

when an Ontarion implements powerful actions in Ontarion > Albertan events . The negative coefficient for D1 BP in the Alberta equation simply reflects this aspect of the potency dynamics of affect control theory . In sum, the amount of affective deflection produced by Ontarion > Albertan events for Alberta respondents depends, not only on their cultural sentiments for the two regional identities, but also on what an Ontarion actually does to an Albertan . By the same token, the absence of interactions between event-type and behavior in the Ontario equation reveals that the amount of deflection produced by Ontarion > Albertan events for Ontario respondents is influenced solely by their cultural sentiments for an Ontarion and an Albertan . D2 : Albertan > Ontarion Events Alberta equation: Deflection' = 5 .571 -1 .752D2 + .933D2BE Ontario equation : Deflection' = 3 .454 - .620D2 + .397D2BE + .339D2BP Recalling, again, that the reference category in each equation represents in-group interaction, the negative coefficient for D2 in both equations reveals that deflection decreases as one moves from in-group to out-group interaction involving Albertan > Ontarion events . The size of this effect in the Alberta equation (-1 .752 ; p = .000) is about three times larger than in the Ontario equation (- .620 ; p < .001) . However, the main effect of D2 events in both equations is moderated by the nature of the act-behavior evaluation (BE) in both equations and, additionally, behavior potency (BP) in the Ontario equation . The large positive effect of D2BE in the Alberta ( .933 ; p = .000) and Ontario ( .397 ; p < .001) equations counteracts the reduction in deflection attributed to the main effect of D2 events in each equation (-1 .752 and - .620, respectively) . This suggests that the nicer the actions of an Albertan, the greater the deflection produced by Alberta > Ontario events for both Alberta and Ontario samples . The main effect of D2 (Albertan > Ontarion) events in the Ontario equation is also moderated by its interaction with the potency of an Albertan's actions (BP) . The positive coefficient for D2BP ( .339 ; p = .017) indicates that increasingly powerful actions by an Albertan toward an Ontarion washes out the deflection-reducing main effect of D2 events (- .620 ; p = 000) . Discussion As reported above in Table 4, the general equation for predicting the global affective outcomes of intergroup relations between Albertans and Ontarions explains a substantial and significant amount of variance in the dependent variable . The large and significant main effects for behavioral terms in this equation reveal that the amount of deflection produced by intergroup events depends on the nature of the act-what interactants actually do to one another, regardless of the



Stereotyping and Intergroup Relations

67

type of event . However, the interactions between behavior and event-type in the general equation indicate that the effect of behavior on deflection is influenced by who implements the behavior and who is the recipient . The main effects of event-type in the general equation represent the impact of intergroup relations (relative to within-group relations) on the deflection produced by events . In the remainder of this section, we discuss the mechanisms through which event-type influences the affective outcomes of intergroup relations . We conclude our discussion by addressing several seemingly anomalous findings from this study . As defined above, affective deflection is based on the discrepancy between the cultural sentiments for identities and the transient impressions or feelings created by events . The impression formation equations generating transient impressions in affect control theory are asymmetrical . As a consequence, the transient impressions for the actor in ABO events are different from those for the recipient . Because the transient impressions of an Ontarion created by D 1 (Ontarion > Albertan) events are quite different from those produced by D2 (Albertan > Ontarion) events, for instance, the amount of affective deflection produced by those two kinds of events will also be different . This explains the differences between equations for predicting D1 (Ontarion > Albertan) and D2 (Albertan > Ontarion) events within each sample . Differences in equations across samples for predicting the same type of event-D 1 (Ontarion > Albertan) events, for instance-can be attributed to differences between samples in cultural sentiments for the regional identities, Albertan and Ontarion . Recalling our discussion of Table 1, Ontario respondents evaluate an Ontarion higher than an Albertan ; and, depending on their gender, Alberta respondents evaluate an Albertan as three to five times nicer than an Ontarion . The cultural sentiments of Ontario and Alberta samples for regional identities also differ on the potency and activity dimensions of affective meaning . For example, Alberta respondents rate an Ontarion as more powerful than an Albertan . These differences in cultural sentiments across samples result in sample-specific equations for predicting the affective response to the same kind of event . Our analysis of the affective outcomes of intergroup relations has revealed several apparently anomalous findings-for instance, that out-group interaction between Albertans and Ontarions creates less deflection than in-group interaction . This is reflected in the negative effects of D1 (Ontarion > Albertan) and D2 (Albertan > Ontarion) events in both the Alberta and Ontario equations compared to the reference categories in these equations (Albertan > Albertan and Ontarion > Ontarion events, respectively) . This counterintuitive finding results from the fact that we have "forced" a full range of intergroup behaviors in the events analyzed in this study, including extremely bad and powerful actions (see Figure 1) . Because such actions are even more uncharacteristic of social relations within than between groups, deflection declines as we move from in-group to out-group interaction .



68

NEIL J . MacKINNON and JEFFREY W. BOWLBY

Another finding that seems anomalous at first blush concerns the increase in deflection resulting from the interaction between event-type and behavior evaluation-the 131 BE term in the Alberta equation for D1 events and the D2BE term in both Alberta and Ontario equations for D2 events . The DIBE term suggests that deflection increases in Ontario > Alberta events with the goodness of an Ontarion's actions ; and the D2BE term, that deflection increases in Albertan > Ontarion events with the goodness of an Albertan's actions . This remains a strange finding only if one fails to consider that deflection is a squared quantity, hence not resulting exclusively from negatively evaluated actions . While an examination of our data matrix reveals that bad actions create much greater deflection than good ones, positively evaluated actions that exceed those expected by virtue of the fundamental evaluation of an Albertan or Ontarion also produce considerable deflection . This kind of deflection, however, enhances the status of interactants . That is, while extremely nice interpersonal actions between an Albertan and an Ontarion create deflection because they are culturally unexpected, their consequence is to increase rather than diminish the status of each participant . The positive effects of DIBE and D2BE in our equations simply reflect the status-enhancing deflection produced by such events .

GENERAL DISCUSSION This paper has reported three studies to illustrate the application of affect control theory to stereotyping and intergroup relations . The first explored the in-group bias/ethnocentrism hypothesis and the out-group homogeneity hypothesis from social cognition research on stereotyping, employing data on intergroup attitudes (cultural sentiments) rather than intergroup perceptions (ascribed traits) . The second study applied the attribution equations of affect control theory to generate stereotypic traits from the cultural sentiments for regional identities . The third applied the theory's impression formation equations to explore the affective outcomes of intergroup relations . Findings from these three studies have been discussed extensively above . Here we address the more general issues raised by this kind of research . The first issue concerns the external validity or generalizability of our findings . For instance, we did not directly measure the affective reactions of Ontario and Alberta respondents to the events analyzed in Study 3 . Instead, we constructed a measure (affective deflection) based on the discrepancy between the transient feelings produced by events and the culturally established sentiments for the regional identities and actions involved in them . While the cultural sentiments of Alberta and Ontario respondents were directly measured in our research, transient feelings were generated from the impression formation equations of affect control theory estimated from previous Canadian research . This calls into question the external validity of our findings-whether we can generalize from estimates of



Stereotyping and Intergroup Relations

69

affective deflection based on equations established in previous research to the affective reactions that might actually be experienced by Alberta and Ontario respondents . However, the impression formation equations of affect control theory have been estimated empirically from large-scale, carefully designed research on the affective reaction of people to hundreds of interpersonal events . Moreover, predictions from affect control models based on these empirical equations have been supported by experimental research (Wiggins and Heise 1988 ; Robinson and Smith-Lovin 1992 ; Robinson, Smith-Lovin, and Tsoudis 1994), survey research (Heise and MacKinnon 1988 ; Heise and Calhan 1995), and qualitative research (Smith-Lovin and Douglas 1992 ; Francis 1997) . Therefore, despite the fact that we did not measure the affective reactions of Alberta and Ontario respondents directly, we feel confident in what the findings from Study 3 reveal about the affective outcomes of intergroup relations between Albertans and Ontarions . The same kind of question concerning external validity also applies to our findings from Study 2, which employed the attribution equations of affect control theory to generate the stereotypic traits of regional identities . However, these equations are simple derivations from the empirically based amalgamation equations of the theory, which describe how traits and other modifiers combine with social identities to create impressions of composite identities (Averett and Heise 1988) . Therefore, we also feel confident in what the findings from Study 2 suggest about the stereotypic traits of Canadian regional identities . There remains, however, a second and more mundane kind of external validity to consider, one that applies to all three studies reported in this paper-the extent to which we can generalize findings based upon non-probability, convenience samples of university students to the larger populations of Ontario and Alberta residents . In this case, the question of external validity pertains to the representativeness of our samples, rather than the validity and integrity of our research procedures . Here, we rest our case on the assumption that the kind of cultural data employed in our research-EPA measures of social identities-is uniform enough within cultures to be accessible from convenience samples of university students . From this perspective, sample respondents serve as "informants," providing information on the intergroup attitudes of Alberta and Ontario cultures . This raises the related question as to whether samples of about 30 males and 30 females from each culture are sufficiently large to provide reliable information . Here, we rely on the mathematical model developed and tested by Romney, Weller, and Batchhelder (1986), which estimates that stable results for the aggregation of high concordance cultural data (which we presume ours to be) can be expected from samples of as few as a half dozen informants .



70

NEIL J . MacKINNON and JEFFREY W . BOWLBY

CONCLUSION In the introduction to this paper we identified three deficits of social cognition research on stereotyping that have hindered its developing an explanatory model of intergroup relations, articulating the causal connections among stereotypes, intergroup attitudes, and intergroup behavior . These deficits include a traditional neglect of affect, a components approach to intergroup attitudes that inspires predictive rather than explanatory models, and the lack of a control perspective incorporating cognitive and affective processes as aspects of a larger system . We also described how affect control theory provides a powerful explanatory model of intergroup relations by not sharing these limitations of social cognition research on stereotyping . At the same time, however, one must acknowledge the strengths of social cognition theory and research that complement the limitations of affect control theory . For instance, affect control theory does not measure stereotypic traits of social groups directly ; instead, the theory generates them from attitudinal data on group identities (Study 2) . Therefore, it is important to validate predictions from affect control theory with direct research on the stereotypic traits actually attributed to social groups . This remains the forte of social cognition research, which has developed sophisticated measures and procedures to study the content, ascription, and accuracy of intergroup perceptions (e .g ., Esses et al . 1993 ; Judd and Park 1995 ; Ryan 1996 ; Stangor et al . 1991) . Moreover, social cognition researchers have expanded the cognitive content of stereotypes to include, not only ascribed traits, but also symbolic beliefs (those pertaining to whether groups uphold or violate cherished values and norms), and have demonstrated the importance of such beliefs in predicting intergroup attitudes (Esses et al . 1993) . In contrast, there is no direct equivalent to the construct of symbolic beliefs in affect control theory . Finally, while the attitudinal data of affect control theory enables one to explore such intergroup phenomena as ethnocentrism and homogeneity effects (Study 1), social cognition research on intergroup perceptions enriches our understanding of these phenomena . In short, social cognition theory's focus on cognition and cognitive processes and affect control theory's focus on affect and affective processes complement one another in the study of intergroup relations . We have located this paper in the context of social cognition theory because the bulk of research on stereotyping and intergroup relations in social psychology has been conducted within this general paradigm . In order to balance our comparison of affect control and social cognition approaches to this area of research, we conclude this paper with a brief discussion of social identity theory (see above references), a more or less separate development within social cognition theory itself . Like affect control theory, social identity theory advances an explanatory model that is lacking in general social cognition research on stereotyping and intergroup relations .



Stereotyping and Intergroup Relations

71

Social identity theory proposes that broad social categories or attributes (gender, ethnic origin, and so on) describing groups to which one belongs, or to which one feels one belongs, are represented in one's mind as social identities . Thus conceived, social identities become important components of self-concept, motivating group comparisons and behavior that favor the in-group and hence the self . As a consequence, in-group and out-group perception become stereotypic and intergroup behavior acquires competitive and discriminatory properties . Social identity theory emphasizes the generative nature of the "sociocognitive mechanisms" (self-categorization and the self-enhancement motive) underlying identity processes, and views the salience and content of social identities as dynamically responsive to situational context (Hogg et al . 1995) . A later extension of social identity theory, self-categorization theory (Turner 1985 ; Turner, Hogg, Oakes, Reicher, and Wetherell 1987) draws heavily from cognitive psychology to elaborate self-categorization and identity processes . Like social identity theory, affect control theory deals with identities based upon broad social categories or attributes . Affect control research has accumulated EPA data for many such identities, and the existence of comparable data for various cultures (American, Canadian, German, Japanese) enables one to analyze intergroup relations from a cross-cultural perspective . As discussed earlier in this paper, the theory also advances a social psychological model of identity processes . Like the model advanced by social identity theory, the affect control model is an inherently dynamic and generative one, according to which the affective meaning of identities shifts with the identities of others in an event, the actions implemented by each interactant, and the setting in which the event takes place . Finally, like social identity theory, affect control theory supposes the operation of a self-enhancement motive in identity processes, locating it in a social psychological theory of motivation based upon the principle of affect control and the incentive of identity-confirmation (MacKinnon 1994, pp . 50-59) . For both theories, the motivation to enhance self-esteem through the confirmation of valued social identities is the well-spring of stereotyping, ethnocentrism, and discrimination in intergroup relations . Despite these similarities, there are important conceptual differences between the two theories that stem from their origins in different traditions . While social identity theory developed from the cognitive tradition of individual psychology, affect control theory traces its roots to the social psychological tradition of Mead and symbolic interactionism . This difference in origin has important repercussions for how each theory views the micro-foundations of group processes and intergroup relations . For social identity/self-categorization theory, the depersonalization of self-the shift in identity from unique individual to group member resulting from the categorization of self in terms of group prototypes-is the fundamental process underlying group and intergroup aspects of behavior such as stereotyping and ethnocentrism . As a consequence, "social identity theory and self-categorization models . . . do not construct group processes from interpersonal



72

NEIL J . MacKINNON and JEFFREY W . BOWLBY

processes . The process of self-categorization depersonalizes perception, feelings, and actions in terms of the contextually relevant self-defining in-group prototypes" (Hogg et al . 1995, p . 262) . Establishing group processes on cognitive structures (group or otherwise), rather than interpersonal processes, brings social identity theory perilously close to "the jagged rocks of psychological reductionism" (p . 264) . In contrast, affect control theory constructs group processes and intergroup relations from interpersonal events, the units of analysis in symbolic interactionism thought and the ultimate empirical referents of social phenomena (Wallace 1983, p . 167) . Moreover, the theory does not associate group identification with the depersonalization of group members, but rather with the internalization of cultural sentiments for the groups with which individuals identify . Thus, by establishing group processes on interpersonal events and the internalization of culture, affect control theory avoids the psychological reductionism implicit in social identity theory, maintaining its integrity as a genuine social psychological theory . In conclusion, social identity theory, like social cognition theory in general, has focused almost exclusively on cognition and cognitive processes . Despite its emphasis on self-enhancement motivation, social identity theory has generally ignored the role of affect, affective processing, and emotions in identity processes (for one attempt to integrate social identity theory with appraisal and constructionist theories of emotion, see Smith 1993) . This deficit of social identity theory (and social cognition theory in general) creates an opening for affect control theory to make an important contribution to our understanding of group processes and intergroup relations .

ACKNOWLEDGMENTS The research reported in this paper has been supported by the Social Sciences and Humanities Research Council of Canada (Grant #410-94-0087) . We wish to express our gratitude to John Goyder, David Heise, Herman Smith, and two anonymous reviewers for comments on earlier drafts of this paper .

REFERENCES Albrecht, S . L ., D . L . Thomas, and B . A. Chadwick. 1980 . Social Psychology . Englewood Cliffs, NJ : Prentice-Hall . Allen, B . P . 1996. "African American and European Americans' Mutual Attributions : Adjective Generation Technique (AGT) Stereotyping ." Journal of Applied Social Psychology 26 : 884-912. Averett, C., and D . R . Heise . 1988. "Modified Social Identities : Amalgamations, Attributions, and Emotions ." Pp. 103-132 in Analyzing Social Interaction: Advances in Affect Control Theory, edited by L . Smith-Lovin and D. R . Heise . New York : Gordon and Breach (reprint from a special Issue of the Journal of Mathematical Sociology 13) . Blalock, H . M . 1979 . Social Statistics (2nd ed .) . New York : McGraw-Hill .



Stereotyping and Intergroup Relations

73

Cantor, N ., and W .Mischel . 1979 . "Prototypes in Person Perception ." In Advances in Experimental Social Psychology, Vol . 12, edited by L. Berkowitz . New York: Academic Press . Collins, R . 1990 . "Stratification, Emotional Energy, and the Transient Emotions ." Pp . 27-57 in Research Agendas in the Sociology of Emotions, edited by T . D . Kemper. Albany, NY : State University of New York Press . DeLamater, J . D . 1992 . "Attitudes ." Pp . 117-124 in Encyclopedia of Sociology, edited by E . Borgatta. New York: MacMillan . Edwards, A . L. 1957 . Techniques of Attitude Scale Construction . New York: Appleton-Century-Crofts . Esses, V. M ., G . Haddock, and M . P . Zanna. 1993 . "Values, Stereotypes, and Emotions as Determinants of Intergroup Attitudes ." Pp . 137-166 in Affect, Cognition, and Stereotyping : Interactive Processes in Group Perception, edited by D . M . Mackie and D . L. Hamilton. San Diego : Academic Press . Fishbein, M . 1978 . "Attitudes and Behavioral Prediction : An Overview ." Pp. 377-389 in Major Social Issues : A Multidisciplinary View, edited by J .M . Yinger and S .J . Cutler . New York : Free Press . Fishbein, M., and I . Ajzen . 1975 . Belief Attitude, Intention, and Behavior : an Introduction to Theory and Research . Reading, MA: Addison-Wesley . Ford, T. E ., and G . R. Tonander . 1998 . "The Role of Differentiation Between Groups and Social Identity in Stereotype Formation." Social Psychology Quarterly 61 : 372-384. Francis, L . E . 1997 . "Ideology and Interpersonal Emotion Management : Redefining Identity in Two Support Groups ." Social Psychology Quarterly 60 : 153-171 . Goffman, E. 1967. Interaction Ritual : Essays on Face-to-Face Behavior. Chicago : Aldine . Hamilton, D. L., and D . M . Mackie . 1993a. "Cognitive and Affective Processes in Intergroup Perception : The Developing Interface ." Pp . 1-11 in Affect, Cognition, and Stereotyping : Interactive Processes in Group Perception, edited by D . M . Mackie and D . L. Hamilton. San Diego : Academic Press . Hamilton, D . L ., and D . M . Mackie. 1993b . "Affect, Cognition, and Stereotyping : Concluding Comments ." Pp. 371-383 in Affect, Cognition, and Stereotyping : Interactive Processes in Group Perception, edited by D . M. Mackie and D . L . Hamilton . San Diego : Academic Press. Hamilton, D. L., S . J . Sherman, and C . M . Ruvolo. 1990. "Stereotype-Based Expectancies : Effects on Information Processing and Social Behavior ." Journal of Social Issues 46 : 35-60 . Heise, D. R. 1978 . Computer-Assisted Analysis of Social Action : Use of Program INTERACT and Survey UNC75 . Chapel Hill, NC : Institute for Research in Social Science . University of North Carolina. Heise, D . R . 1979 . Understanding Events : Affect and the Construction of Social Action . New York : Cambridge University Press . Heise, D. R . 1988 . "Affect Control Theory : Concepts and Model ." Pp . 1-33 in Analyzing Social Interaction : Advances in Affect Control Theory, edited by L . Smith-Lovin and D . R . Heise . New York : Gordon and Breach (reprint of a special issue of the Journal of Mathematical Sociology 13) . Heise, D . R ., and C . Calhan . "Emotion Norms in Interpersonal Events." Social Psychology Quarterly 58 : 223-240 . Heise, D. R ., and N . J . MacKinnon . 1988. "Affective Bases of Likelihood Judgments ." Pp. 133-151 in Analyzing Social Interaction : Advances in Affect Control Theory, edited by L . Smith-Lovin and D. R. Heise . New York: Gordon and Breach (reprint of a special issue of the Journal of Mathematical Sociology 13) . Hill, R. J . 1981 . "Attitudes and Behavior ." Pp . 347-377 in Social Psychology : Sociological Perspectives, edited by M . Rosenberg and R . H. Turner. New York : Basic Books . Hogg, M . A. 1992. The Social Psychology of Group Cohesiveness : From Atraction to Social Identity . London : Harvester/Wheatsheaf .



74

NEIL J . MacKINNON and JEFFREY W . BOWLBY

Hogg, M . A . 1993 . "Group Cohesiveness : A Critical Review and Some New Directions ." European Review of Social Psychology 4 : 85-111 . Hogg, M . A ., and D . Abrams . 1988 . Social Identifications : A Social Psychology of Intergroup Relations and Group Processes . London: Routledge. Hogg, M . A ., D. J. Terry, and K . M . White. 1995 . "A Tale of Two Theories : A Critical Comparison of Identity Theory With Social Identity Theory ." Social Psychology Quarterly 58 : 255-269 . Howard, J . A. 1995 . "Social Cognition ." Pp . 90-117 in Sociological Perspectives on Social Psychology, edited by K . S . Cook, G. A . Fine, and J. S . House . Boston : Allyn and Bacon. Jones, E . and D . McGillis . 1976 . "Correspondent Inferences and the Attribution Cube : A Comparative Reappraisal." Pp. 389-420 in New Directions in Attribution Research, Vol . 1, edited by J . Harvey, W . Ickes, and R . Kidd . Hillsdale, NJ : Lawrence Erlbaum Associates . Judd, C. M ., and B . Park. 1995 . "Stereotypes and Ethnocentrism : Diverging Interethnic Perceptions of African American and White American Youth ." Journal of Personality and Social Psychology 69 :460-481 . Katz, D., and K . Braly. 1933 . "Racial Stereotypes of One Hundred College Students ." Journal of Abnormal and Social Psychology 28 : 280-290 . Katz, D ., and K. Braly . 1935 . "Racial Prejudice and Racial Stereotypes ." Journal of Abnormal and Social Psychology 30 : 175-193 . Kemper, T . D ., and R . Collins . 1990 . "Dimensions of Microinteraction ." American Journal of Sociology 96: 32-68 . Kuhn, T . 1962 . The Structure of Scientific Revolutions . Chicago : University of Chicago Press . Lee, Y-T., and V . Ottati . 1995 . "Perceived In-group Homogeneity as a Function of Group Membership Salience and Stereotype Threat ." Personality and Social Psychology Bulletin 21 : 610-619. Lippman, W . 1922 . Public Opinion. New York : Harcourt Brace Jovanovich . Messick, D . M., and D. M . Mackie . 1989 . "Intergroup Relations ." Annual Review of Psychology 40 : 45-81 . MacKinnon, N . J . 1985 . Affective Dynamics and Role Analysis . Final Report, Social Sciences and Humanities Research council of Canada Project 410-81-0089 . Guelph, Ontario : Department of Sociology and Anthropology, University of Guelph . MacKinnon, N . J . 1994. Symbolic Interactionism as Affect Control. Albany, NY : State University of New York Press . MacKinnon, N . J ., and D . R. Heise. 1993 . "Affect Control Theory : Delineation and History ." Pp. 64-103 in Theoretical Research Programs : Studies in the Growth of Theory, edited by J. Berger and M. Zelditch, Jr . Stanford, CA : Stanford University Press . MacKinnon, N . J., and T. Langford. 1994 . "The Meaning of Occupational Prestige Scores : A Social Psychological Analysis and Interpretation ." The Sociological Quarterly 35 : 215-245 . Mandler, G . 1984 . Mind and Body : Psychology of Emotion and Stress . New York: Norton . McCall, G. J., and J. L. Simmons. 1978 . Identities and Interactions (2d ed.) . New York : Free Press . Mead, G . H . 1934 . Mind, Self and Society, edited by C. W. Morris . Chicago: University of Chicago Press. Moore, D . S ., and G . P. McCabe. 1993 . Introduction to the Practice of Statistics (2nd ed.) . New York: Freeman . Osgood, C . E. 1969 . "On the Whys and Wherefores of E, P, and A ." Journal of Personality and Social Psychology 12 : 194-199 . Osgood, C . E., W . H . May, and M . S . Miron . 1975 . Cross-Cultural Universals of Affective Meaning . Urbana, IL : University of Illinois Press . Osgood, C . E ., G. C. Suci, and P. H . Tannenbaum . 1957. The Measurement of Meaning . Urbana : University of Illinois Press . Pearson, E . S ., and N. W. Please . 1975 . "Relation Between the Shape of Population Distribution and the Robustness of Four Simple Test Statistics ." Biometrika 62 : 223-241 .



Stereotyping and Intergroup Relations

75

Pedhazur, E . J. 1997 . Multiple Regression in Behavioral Research (3rd ed .). Fort Worth, TX : Harcourt-Brace . Robinson, D . T., and L . Smith-Lovin . 1992 . "Selective Interaction and Identity ." Social Psychology Quarterly 55 : 13-28 . Robinson, D. T ., L. Smith-Lovin, and O. Tsoudis. 1994 . "Heinous Crime or Unfortunate Accident? The Effects of Remorse on Responses to Mock Criminal Confessions ." Social Forces 73 : 175-190. Romney, A. K., S . C . Weller, and W . H . Batchelder . 1986 . "Culture as Consensus : A Theory of Culture and Informant Accuracy ." American Anthropologist 88 : 313-338 . Rosenberg, M . J., and C . I . Hovland . 1960 . "Cognitive, Affective, and Behavioral Components of Attitudes." Pp . 1-14 in Attitude Organization and Change, edited by C . I . Hovland and M. J. Rosenberg . New Haven, CT : Yale University Press . Ryan, C . S . 1996. "Accuracy of Black and White College Students' In-Group and Out-Group Stereotypes ." Personality and Social Psychology Bulletin 22 : 1114-1127 . Schneider, D ., A . Hastort, and P . Ellsworth. 1979 . Person Perception. Reading, MA: Addison-Wesley. Schuman, H . 1995 . "Attitudes, Beliefs, and Behavior ." Pp . 68-89 in Sociological Perspectives on Social Psychology, edited by K. S . Cook, G. A . Fine, and J . S . House . Boston : Allyn and Bacon . Seeman, M . 1981 . "Intergroup Relations." Pp . 378-410 in Social Psychology: Sociological Perspectives, edited by M. Rosenberg and R . H . Turner . New York : Basic Books . Smith, E . R . 1993 . "Social Identity and Social Emotions : Toward New Conceptualizations of Prejudice ." Pp . 297-315 in Affect, Cognition, and Stereotyping : Interactive Precesses in Group Perception, edited by D . M . Mackie and D . L . Hamilton . San Diego : Academic Press . Smith-Lovin, L . 1988 . "Impressions from Events ." Pp . 35-70 in Analyzing Social Interaction : Advances in Affect Control Theory, edited by L . Smith-Lovin and D . R. Heise. New York : Gordon and Breach (reprint of a special issue of the Journal of Mathematical Sociology 13) . Smith-Lovin, L. 1990 . "Emotion as the Confirmation and Disconfirmation of Identity : An Affect Control Model ." Pp . 238-270 in Research Agendas in the Sociology of Emotions, edited by T . D . Kemper . Albany : State University of New York Press . Smith-Lovin, L ., and W. Douglas . 1992 . "An Affect Control Analysis of Two Religious Subcultures ." Pp. 217-248 in Social Perspectives on Emotion, Vol . 1, edited by V . Gecas and D . Franks . Greenwich, CT : JAI Press . Smith-Lovin, L ., and D . R . Heise, eds . 1988 . Analyzing Social Interaction : Advances in Affect Control Theory . New York: Gordon and Breach (reprint of a special issue of the Journal of Mathematical Sociology 13). Stangor, C ., L . A . Sullivan, and T. E . Ford . 1991 . "Affective and Cognitive Determinants of Prejudice." Social Cognition 9: 359-380 . Stryker, S . 1968 . "Identity Salience and Role Performance : The Relevance of Symbolic Interaction Theory for Family Research ." Journal of Marriage and the Family 30 : 558-564 . Stryker, S . 1980. Symbolic Interaction : A Social Structural Version . Menlo Park, CA : Benjamin/Cummings . Stryker, S ., and R . T . Serpe . 1994 . "Identity Salience and Psychological Centrality : Equivalence, Overlapping, or Complementary Concepts?" Social Psychology Quarterly 57 : 16-35 . Tafjel, H . 1969 . "Cognitive Aspects of Prejudice ." Journal of Social Issues 25 : 79-98 . Tafjel, H . 1970 . "Experiments in Intergroup Discrimination." Scientific American 223 : 96-102 . Tafjel, H . 1974 . "Social Identity and Intergroup Behaviour ." Social Science Information 13 : 65-93 . Tafjel, H ., ed. 1978 . Differentiation Between Social Groups . London : Academic Press . Tafjel, H ., ed . 1982 . Social Identity and Intergroup Relations . Cambridge, UK : Cambridge University Press .



76

NEIL J . MacKINNON and JEFFREY W . BOWLBY

Tafjel, H ., and J. C . Turner. 1979 . "An Integrative Theory of Intergroup Conflict ." Pp . 33-47 in The Social Psychology of Intergroup Relations, edited by W . G . Austin and S . Worchel . Monterey, CA : Brooks-Cole . Thurstone, L .L . 1928 . "Attitudes Can Be Measured ." American Journal of Sociology 33 : 529-554 . Thurstone, L .L . 1967 . "The Measurement of Social Attitudes ." Pp . 14-25 in Readings in Attitude Theory and Measurement, edited by M . Fishbein . New York: Wiley . Turner, J . C . 1982 . "Towards a Cognitive Redefinition of the Social Group ." Pp . 15-40 in Social Identity and Intergroup Relations, edited by H. Tafjel . Cambridge, UK: Cambridge University Press. Turner, J. C . 1985. "Social Categorization and the Self Concept : A Social Cognitive Theory of Group Behavior." Pp. 77-122 in Advances in Group Processes : Theory and Research, Vol 2, edited by E . J. Lawler. Greenwich, CT : JAI . Turner, J . C., M. A . Hogg, P . J . Oakes, S . D . Reicher, and M . S . Wetherell . 1987. Rediscovering the Social Group: A Self-Categorization Theory . Oxford : Blackwell . Wallace, W. 1983. Principles of Scientific Sociology . New York: Aldine . Wiggins, B ., and D . R . Heise. 1988 . "Expectations, Intentions, and Behavior: Some Tests of Affect Control Theory." Pp . 153-169 in Analyzing Social Interaction: Advances in Affect Control Theory, edited by L . Smith-Lovin and D . R . Heise . New York : Gordon and Breach (reprint of a special issue of the Journal of Mathematical Sociology 13) . Zanna, M . P., and J . K. Rempel. 1988 . "Attitudes : A New Look at an Old Concept ." Pp . 315-334 in The Social Psychology of Knowledge, edited by D . Bar-Tal and A. W. Kruglanski . Cambridge : Cambridge University Press .

THE FORMATION OF STATUS BELIEFS IMPROVING STATUS CONSTRUCTION THEORY

Cecilia L. Ridgeway

ABSTRACT Widely shared status beliefs that characterize people in one social category as more esteemed and competent than people in another category play an important role in the organization of inequality . Status construction theory argues that the terms on which people interact across a social difference boundary can cause shared status beliefs to form and spread widely throughout a population . This chapter reviews the evidence for status construction theory and develops it further by offering an explicit theoretical account of the mechanism by which interactional contexts induce participants to form status beliefs even when those beliefs disadvantage the participant. I derive testable hypotheses from this account . I show as well how this account can be represented in the graph-theoretic terms of expectation states theory, allowing further testable predictions to be derived.

Social status is a fundamental dimension of social inequality in human societies along with social power and material wealth (Weber 1968) . Status involves relations based on social evaluations of prestige, honor, or esteem . Given its recog-

Advances in Group Processes, Volume 17, pages 77-102 . Copyright © 2000 by JAI Press Inc . All rights of reproduction in any form reserved . ISBN : 0-7623-0651-3 77



78

CECILIA L . RIDGEWAY

nized importance in the organization of inequality, it is odd that contemporary sociologists approach social status in a disjointed manner . Macro sociologists conceptualize status as a relationship between social groups and often study it in terms of the relative prestige of occupational groups . Micro sociologists, on the other hand, approach status as a social hierarchy of influence and esteem within a group, often called a power and prestige order . These disparate approaches to status can be brought into dialogue with one another by a focus on the linking concept of status beliefs. Status beliefs are widely shared cultural beliefs that people in one social category are more esteemed and competent than people in another social category . Status beliefs can be perceptions of what "most others" believe or what actors themselves believe . As their definition suggests, status beliefs are beliefs about the evaluative relationship between social categories or groups . Yet, as decades of research with expectations states theory has shown, they are also the primary determinant of the social influence hierarchies actors form when they interact in regard to a shared goal (Berger et al . 1977 ; Webster and Foschi 1988 ; Berger and Zelditch 1998) . A broader investigation of status beliefs-how they form and change as well as what their effects are-is the key to integrating sociological approaches to status and developing a deeper understanding of the role of status dynamics in systems of social stratification and inequality . Status construction theory is a recent approach that describes one set of processes by which status beliefs about social differences might develop and become widely shared among a population (Ridgeway 1991, 1997 ; Ridgeway and Balkwell 1997 ; Ridgeway, Boyle, Kuipers, and Robinson 1998 ; Webster and Hysom 1998) . The theory claims that interactional contexts are not only an arena in which status beliefs are at play ; they are also important contexts for the development and spread of status beliefs . This chapter attempts to increase our understanding of the nature and development of status beliefs in society by elaborating and improving status construction theory's account of the formation of status beliefs in interaction .

FRAMING THE PROBLEM Status construction theory argues that when people on opposite sides of a recognized social difference boundary regularly interact in regard to shared goals, the terms on which they interact and the influence hierarchies that emerge can induce the participants to form shared status beliefs about their difference . People carry these beliefs to other encounters with different others and, by acting on them there, induce some of these others to take on the beliefs as well . This creates a diffusion process that, under some circumstances, will create roughly consensual (i .e ., widely shared) beliefs that people from one of the difference categories are more esteemed and competent than are those from the other difference category .



The Formation of Status Beliefs

79

When a status belief is roughly consensual in a society, it has become part of the society's shared culture and is part of the taken-for-granted social framework of beliefs that actors bring to interaction and use to organize their behavior together . Investigations of status construction theory suggest that interactional contexts where people from different social categories come together may indeed be a potent source of status beliefs . Experiments confirm that when influence hierarchies repeatedly develop between actors with a salient social difference, the actors form corresponding status beliefs about their difference, as status construction theory predicts (Ridgeway et al . 1998 ; Ridgeway and Glasgow 1996) . Other experiments show that actors can indeed spread status beliefs to others by treating those others according to the belief (Ridgeway and Glasgow 1996 ; Ridgeway and Erickson 1998) . Finally, computer simulations support the logical sturdiness of theory's argument that, if interactional processes create and spread status beliefs, the diffusion process will cause these beliefs to become roughly consensual in society under many circumstances (Ridgeway and Balkwell 1997 ; Mark, Smith-Lovin, and Ridgeway 1999) . How can we explain the power of interaction to create status beliefs? It is well known that the creation of any salient difference between people is enough to cause people to favor their own group over the other (Brewer and Kramer 1985 ; Dovidio and Gaertner 1993 ; Messick and Mackie 1989) . With status beliefs, however, all groups agree, or at least concede, that one group is socially recognized as better and more competent than the others are . Thus it is the beliefs of those who are disadvantaged by status beliefs that distinguish status beliefs from the in-group bias created by difference alone . What is it about an interactional context that could induce an actor to form a status belief that disadvantages that actor? Status construction theory offers an account of belief formation in interaction that can be summarized as follows (Ridgeway 1991) . When people who differ in standing in the power and prestige order of an encounter also differ in an easily recognized, distinguishing attribute, there is a likelihood that they will associate their perceived differences in competence and status in the setting with the distinguishing attribute and form a status belief about that attribute . This assumes that the actors do not already have shared status beliefs about the attribute . Note that this account of belief formation is essentially descriptive . It does not explain the processes through which an association between a nominal difference and standing in a power and prestige order induces people to generalize to status beliefs about whole categories of actors, even when these beliefs are against their personal interests . As an initial account of belief formation through interaction, this descriptive formulation has worked well . It has stimulated empirical investigations of the formation of status beliefs in interaction and led to the further development of status construction theory . Given what has been learned, however, it is now clear that this initial descriptive account of belief formation has limitations that need to be addressed . First, it does not explain exactly how an association between standing



80

CECILIA L . RIDGEWAY

in a power and prestige order and a distinguishing attribute leads to status beliefs among participants . The informal, general terms in which the process is described in the theory are open to diverse interpretations . Second, the formulation fails particularly to make clear why the associational process it describes would be sufficient to induce actors to take on status beliefs that disadvantage them . Finally, status construction theory uses expectation states theory to formulate key parts of its argument about the development of power and prestige orders in interactional contexts . Yet the core of status construction theory, its description of the formation of status beliefs, is not expressed in terms that are easily related to expectation states theory . Expectation states theory is the best-developed and empirically documented theory of status dynamics at the interactional level . The ability of status construction theory to improve our understanding of status beliefs and their role in linking micro and macro status processes will be greater if the theory is formulated in terms that can be articulated with the dominant micro-level account. The overall goal of this chapter is to improve our understanding of the emergence of status beliefs in society by developing status construction theory's account of belief formation . I will examine in detail the relationship between the local realities created by power and prestige orders in interactional contexts and the nature of status beliefs . Based on this examination, I will suggest a formal belief formation assumption for status construction theory that offers an explicit theoretical account of the mechanism by which interaction induces actors to form status beliefs. The goal is to theorize this mechanism in a manner that accounts for the acceptance of status beliefs by those they disadvantage as well as advantage . A specification of the mechanism by which interaction creates status belief must also take into account recent theoretical and empirical developments in status construction theory . I will then show how the belief formation assumption can be explicitly articulated with expectation states theory . I begin by reviewing recent developments in status construction theory and the issues they raise for understanding how interaction creates status beliefs . Then I turn to the task of formulating an improved and formally explicit theoretical account for this process .

DEVELOPMENTS IN STATUS CONSTRUCTION THEORY Creating Status Beliefs From Resource Differences In its initial formulation, status construction theory focused on the way a correlation between the distribution in exchangeable resources and a nominal (i .e ., unordered) distinction (say, As and Bs) among people in a society could be transformed by interactional processes into widely held beliefs that members of the resource-advantaged category are more esteemed and competent than members of the resource-disadvantaged category (Ridgeway 1991) . Given this focus, the the-



The Formation of Status Beliefs

81

ory highlighted encounters between people who differed in both resources and the AB distinction (called "doubly dissimilar" encounters) as the central contexts that drive the formation of status beliefs about the distinction . Resource advantages tend to create expectations for competence in the situation, as expectation states research has shown (Cook 1975 ; Harrod 1980 : Stewart and Moore 1992) . These expectations shape behavior in ways that result in a power and prestige order favoring the better paid . In doubly dissimilar encounters, actors may associate their standing in the power and prestige order, created by their resource differences, with their corresponding AB differences and form status beliefs about that distinction . Due to the correlation in society between resources and the A/B distinction, more doubly dissimilar encounters will foster status beliefs favoring the resource-advantaged category (say, As) than the resource-disadvantaged category (Bs) . As actors spread their beliefs formed in doubly dissimilar encounters to other encounters with people who differ on the AB distinction, the numerical predominance of status beliefs favoring As should gradually overwhelm those favoring Bs, resulting in widely held status beliefs favoring As . An experimental test of this formulation has shown that doubly dissimilar encounters do indeed induce their participants to form status beliefs favoring the resource-advantaged nominal category . After two encounters with partners who differed from them in a nominal distinction, were paid differently, and attained influence in the encounters corresponding to their pay level, subjects developed beliefs that "most people" consider the typical person in the resource-advantaged category to be more respected, competent, and leaderlike, but not as cooperative and considerate as the typical member of the disadvantaged category (Ridgeway et al . 1998) . ! Significantly, even subjects in the disadvantaged category formed beliefs that those in the other group are considered higher status and more competent than people in their own group . Two subsequent experiments, which I will describe shortly, further demonstrated that actors could transmit status beliefs to others by acting on those beliefs in interaction (Ridgeway and Glasgow 1996 ; Ridgeway and Erickson 1998) . Thus, status beliefs created in doubly dissimilar encounters can be spread more broadly through interaction . In addition to this experimental evidence, simulations of the proposed diffusion process provide logical support for the theory's prediction that status beliefs favoring the resource-advantaged category will become roughly consensual among the population (Ridgeway and Balkwell 1997) . Therefore, the evidence thus far supports (although does not fully prove) the initial predictions derived from status construction theory's assumptions about status belief formation and transmission in interaction . These predictions can be summarized as follows . Given a correlation between the distribution of resources and a nominal distinction in a society, interactional processes will be sufficient to create widely shared status beliefs about that distinction . Those beliefs will favor the resource-advantaged nominal category .



82

CECILIA L . RIDGEWAY

A More General Condition for the Development of Status Beliefs

Despite status construction theory's initial focus on resource differences, however, its model of belief formation actually implies a more general set of circumstances by which status beliefs could develop . If status beliefs are fostered by a systematic association between standing in power and prestige orders and a nominal distinction, then any factor that systematically biases the development of influence hierarchies in encounters between nominally different actors will create status beliefs favoring the nominal category advantaged by that factor . This second derivation is status construction theory's most general implication (Ridgeway and Glasgow 1996 ; Webster and Hysom 1998) . Webster and Hysom employ this derivation to expand status construction theory to show how distinguishing characteristics, like sexual orientation, that are not immediately visible but that carry a moral evaluation in society can become status-valued characteristics . Webster and Hysom (1998) propose as well that not only resources with exchange value, but also those with status value (e.g ., a larger office) could act as the biasing factors that create status beliefs . Differential access to valued information between As and Bs, or control over technology are other examples of biasing factors that could potentially create status beliefs about an A/B distinction . For this second, more general derivation to hold, people must form status beliefs from the simple association in encounters of standing in the power and prestige order and a nominal distinction without obvious accompanying differences in resources or other factors . In other words, the experience of interactional hierarchy itself associated with a nominal distinction must be sufficient to create status beliefs if interactional processes are to act as a general translator of biasing factors into status beliefs . In an experiment that tests this hypothesis, Ridgeway and Glasgow (1996) had confederates treat a nominally distinct subject in a deferential or assertive, nondeferential manner during a cooperative decision-making task . As expectations states theory predicts, this created patterns of behavioral exchange between the subject and confederate that resulted in the deferential confederate becoming less influential than the subject and the assertive confederate becoming more influential than the subject . After two experiences with an influence hierarchy associated with a nominal distinction, subjects formed beliefs that "most people" consider the typical member of the influence-advantaged nominal category to be higher status and more competent, but less cooperative and considerate, than the average person in the influence-disadvantaged nominal category . Again, subjects who were influence disadvantaged themselves formed status beliefs that disadvantaged their own nominal category . Thus interactional hierarchies alone, without supporting resource differences, were sufficient to create status beliefs . This suggests that status construction theory's second and most general derivation is plausible .



The Formation of Status Beliefs

83

Implications for the Spread of Status Beliefs

Ridgeway and Glasgow's (1996) experiment also indicates that if actors who have formed a status belief in one encounter then treat others according to it in subsequent encounters, those others are likely to acquire the belief themselves as status construction theory assumes . The confederates in that experiment acted as though they thought the nominal distinction between them and the subjects carried status value . By so acting, they created a power and prestige order between them and the subjects in self-fulfilling manner and from that, the subjects inferred the status beliefs themselves . Thus this experiment shows that status beliefs can be spread through behavior in interaction, as status construction theory argues . A subsequent experiment suggests that interaction may spread status beliefs quite broadly . Ridgeway and Erickson (1998) had a third member of a group observe the status valued (i .e ., deferential or nondeferential) treatment of someone who was nominally the same as they were by someone else who was nominally different. These bystander participants also formed status-corresponding status beliefs about the nominal distinction . Thus, status beliefs can be spread through observation in interaction as well as personal treatment. As a result, interaction may be a powerful teacher of status beliefs, greatly speeding the diffusion process, as simulations suggest (Ridgeway and Balkwell 1997) . Creating Status Beliefs Through Chance Processes

If any factor that systematically biases the development of influence hierarchies between nominally distinct actors can foster status beliefs, could chance processes alone result in the development of status beliefs about the nominal distinction? In the initial statement of the theory, Ridgeway (1991) reasoned that, in encounters between nominally different actors where no biasing factor was present, the association between nominal category and power and prestige would be governed by chance . As a result, these encounters would produce equal numbers of status beliefs favoring one nominal category as favoring the other category . In the diffusion process, these contradicting beliefs would eventually wash each other out . This analysis, however, neglects the way chance processes can produce path dependent effects that act as biasing factors in favor of a given status belief. By chance alone, a series of coin flips can occasionally produce a string of "heads" or "tails ." Similarly, a series of initial encounters between As and Bs could by chance produce a substantial preponderance of status beliefs favoring one category (e .g ., As) rather than another . As actors spread these beliefs through future encounters, there is a chance that this initial preponderance will be magnified and gradually result in consensual status beliefs favoring the initially preponderant category . In effect, the chance string of early encounters creating beliefs favoring one category, not the other, acts as a type of biasing factor that leads to widely held status beliefs . In a formal analysis with supporting simulations, Mark



84

CECILIA L . RIDGEWAY

and colleagues show that an initial predominance of status beliefs favoring a particular nominal category, created by chance, would logically result in consensual status beliefs favoring that category under many circumstances (Mark 1999 ; Mark, Smith-Lovin, and Ridgeway 1999) . Mark and colleagues' analyses suggest that the belief formation processes described by status construction theory have an additional set of implications not initially anticipated by the theory . When actors regularly interact across a nominal difference boundary, there is a likelihood that they will form shared status beliefs about the nominal distinction through chance processes alone . When status beliefs are created by chance factors, however, the group that will be favored by the beliefs cannot be predicted in advance . This is in contrast to the situation where beliefs are created by a nonrandom biasing factor such as resources or technology. Thus, when status beliefs are created by chance processes, all categories of the nominal distinction are equally likely to become high status . Conclusions and Implications

For the purposes of this paper, what conclusions can we draw from these theoretical and empirical developments in status construction theory? First, the theory's initial focus on doubly dissimilar encounters as the crucial sites for the creation of status beliefs is too limited . All encounters between people who differ on a given nominal distinction are potentially sites for the creation of status beliefs . In cases where a systematic biasing factor is present, a subset of nominally different encounters will also be doubly dissimilar encounters and the latter will be important for determining the evaluative direction (i .e ., favoring As or Bs) of the status beliefs that develop. Second, a wide variety of factors other than resources can function as the biasing factor that creates status beliefs favoring a given category of a nominal distinction . These two points suggest that any specification of the mechanism by which interactional associations induce status beliefs must accommodate encounters where no biasing factors are salient in the situation as well as encounters where such factors are present . The specification must also allow the incorporation of a variety of biasing factors .

THE NATURE OF STATUS BELIEFS

To specify a mechanism by which interaction creates status beliefs we must first keep in mind the nature of status beliefs . In the terms of expectations states theory, status beliefs are the widely shared beliefs that render a recognized social distinction into a status characteristic . Status beliefs associate higher social evaluation and expectations for greater general and/or specific competence with one state of a characteristic (men, the better educated) than another (women, the lesser educated) (Berger et al . 1977) . At a minimum, then, interaction must associate both



The Formation of Status Beliefs

85

evaluative and competence differences with actor's nominal categories in order for actors to impute status beliefs about the nominal distinction . In this section I consider the means by which interactional contexts create an apparent correspondence between a nominal distinction and differences in both social evaluation and competence . In the next section I take up the further question of how a mere situational correspondence between competence and evaluation differences and a nominal distinction could induce people to form status beliefs that disadvantage them. It is clear how goal-oriented interaction juxtaposes presumed competence differences with the actors' nominal distinction . As expectations states research shows, actors in goal-oriented interaction form differentiated expectations for the usefulness or competence of the contributions from each actor compared to the others (Berger, Conner, and Fisek 1974 ; Fisek, Berger, and Norman 1991) . These differentiated performance expectations are the foundation of the power and prestige order (i .e ., influence hierarchy) that organizes interaction in the setting . In goal-oriented settings, then, interactional hierarchies embody presumptions about differences in competence in that setting . This expectation states theory argument is the basis for status construction theory's assumption that interaction can induce status beliefs (Ridgeway 1991) . In addition to embodying presumptions of competence, the behaviors by which actors actually enact relations of influence and esteem also act as signs or markers of each actor's evaluative standing, or status, in the setting . These observable power and prestige behaviors, as expectation states theory calls them, include opportunities to participate, participation rate, positive evaluations received, influence, and choice for leadership positions (see Webster and Hysom 1998, p . 354) . These behaviors cast the influential actor in the position of the primary performer, the center of evaluative attention in the setting, while less influential actors are cast in the position of reactors to the primary performer (Wagner and Berger 1997) . It is apparent, then, that the observable power and prestige behaviors by which interactional hierarchies are enacted have a dual aspect . They are driven by and represent actors' presumptions about their differences in competence in the situation . By the same token, they are also status markers that signify differences in the social evaluation of actors in the situation (Ridgeway and Berger 1986) . Not surprising, research indicates a reciprocal relationship between the display of observable power and prestige behaviors and presumptions of competence (Berger et al . 1974 ; Ridgeway, Berger, and Smith 1985 ; Fisek, Berger, and Norman 1991) . Because these interactional power and prestige behaviors are culturally associated with the enactment of competence differences, actors can display them to create the appearance of competence and gain influence in the first place (especially among social peers) . On the other hand, any other factors that affect the actors' appearance of competence, such as external status characteristics (e .g ., occupation, education, race, or gender), resource differences, or actual performance in



86

CECILIA L . RIDGEWAY

the setting will also determine the power and prestige behaviors that the actor can effectively display in the situation . The enactment of an interactional influence hierarchy, then, provides a display of both apparent competent differences and status markers indicating evaluative differences . Thus the association of position in an interaction hierarchy with a nominal distinction creates a juxtaposition between that distinction and competence and evaluative differences that could encourage the formation status beliefs . Observable power and prestige behaviors, however, are not the only means by which evaluative differences among actors might be displayed in an interactional (or other) setting . The social evaluation of actors can be indicated through a variety of other social signs and attributes as well . As Webster and Hysom (1998) note, evaluative standing can be indicated by the possession of valued goal objects (a nice house) or by differences in exchangeable resources such as pay differences in addition to behavioral gestures of deference and esteem . The diverse indicators of evaluative differences between actors, however, can all be understood as status markers (Ridgeway and Berger 1986) . A status marker is any possession, attribute, or behavior that is culturally associated with a position of high rather than low status in power and prestige orders . While there are a wide variety of status markers in any culture, the observable power and prestige behaviors by which actors actually enact relations of influence and esteem may function as core status markers . I base this suggestion on evidence that the enactment of a behavioral power and prestige order creates impressions of the actors that correspond closely to what research indicates is the core content of all status beliefs . Geis and colleagues (1984) and Gerber (1996) report that actors in the high status position of an observable power and prestige order are perceived as more instrumental and directive while those in the low status position are seen as more expressive and emotional . In strikingly parallel findings, Conway and colleagues (1996) show that the status distinctions associated with gender, occupation, and general evaluative standing in a society all involve similar impressions of those in high rather than low status groups . In each case, people in higher status groups are seen as more agentic and instrumentally competent while those in the lower status groups are perceived as more interpersonally responsive and communal . Recall that experiments testing status construction theory's arguments about belief formation in interaction also found that subjects developed beliefs that one nominal group was more leaderlike and competent but not as cooperative and considerate as the other nominal group (Ridgeway et al . 1998 ; Ridgeway and Glasgow 1996 ; Ridgeway and Erickson 1998) . Perhaps it is not surprising that the content of status beliefs so closely reflects the impressions actors create when they enact an interactional status order . Status beliefs are widely shared cultural schemas for organizing cooperative interaction across a social difference boundary, but on unequal terms (Ridgeway et al . 1998) . As shared schemas for organizing interaction, status beliefs typify interactional structures . The correspondence between status beliefs and the impression actors



The Formation of Status Beliefs

87

give in observable power and prestige structures underlines the importance of interactional processes for status beliefs . It adds credence to status construction theory's claim that interaction is not only an arena that is shaped by status beliefs, but one in which such beliefs can be created and changed . The above considerations suggest a more detailed and useful conceptualization of the process by which status beliefs are created in status construction theory . To begin with, it is useful to conceptualize the process as the establishment of a link between a nominal distinction and both differences in status markers and expected competence . The concept of status markers is broad enough to encompass the diverse interactional contexts that status construction theory argues can lead to status beliefs, including those involving resource differences and those with behavioral influence hierarchies alone . This conceptualization is also phrased in the terms of expectation states theory, facilitating its integration with that approach . This is the first step in specifying the mechanism by which interaction induces status beliefs . It is still not clear, however, why a correspondence between a nominal distinction and differential status markers and competence expectations should induce people to infer status beliefs that disadvantage them . Addressing this important problem is the crucial next step in the argument .

CONSENSUS, VALIDITY, AND THE ACCEPTANCE OF STATUS BELIEFS Ridgeway and colleagues (1998) argue that interactional power and prestige orders can create local realities where a nominal difference appears to the participants to be consensually evaluated before it actually is so on a broader scale . The local appearance of an evaluative consensus about a nominal distinction, I suggest, is the key to inducing those who would be disadvantaged by a status belief to take it on . The appearance of consensus also spurs actors to generalize beyond the current situation to a belief, like a status belief, that they expect to use to make sense of future interactions . When actors with one state of a nominal distinction have consistently higher status markers and expected competence in the situation than those with another state of the distinction, then there is a correspondence in the situation among states of the nominal distinction, status markers, and expected competence . When a power and prestige order enacts such a correspondence and this correspondence is not explicitly commented upon or resisted, the enacted correspondence appears to be socially accepted and consensual . The appearance of consensus makes the correspondence seem like an objective social reality in the situation that is accepted by others as a social fact. In effect, the appearance of consensus gives the correspondence among the nominal distinction, status markers, and competence expectations social validity in the eyes of participants (Ridgeway and Berger 1986 ; Weber 1968 ; Zelditch and



88

CECILIA L . RIDGEWAY

Floyd 1998 ; Zelditch and Walker 1984) . 2 Following Weber (1968, pp . 31-33), Zelditch and Floyd argue that something is valid in a situation if actors expect other actors to accept that thing in the situation. Thus the correspondence is socially valid in the situation when an actor, P, expects others to accept the correspondence in the situation . In a situation where such a correspondence is enacted without contest, then, participants enter into a local reality where it seems accepted as objectively valid that people in one nominal category are more competent and leaderlike than are people in another category . Given the apparently valid social truth that the situation presents to its participants, is it surprising that they take the next step and form a more general belief that As are typically more competent and status worthy than Bs? The social validity that the situation lends to the correspondence among the nominal distinction, status markers, and competence has two important effects that facilitate this belief formation process . First, since the correspondence appears to be a valid social fact, actors presume it would be accepted as a social fact by others beyond the situation as well . The sense that it would be accepted by others is an inducement to the actors to generalize the local correspondence into a belief about As and Bs that could be transferred to other actors and situations . In effect, the social validity of the correspondence in the local situation suggests that a status belief about As and Bs is an effective social schema for making sense of and acting toward others in future encounters because those others will also accept its social truth . Second, the sense that the correspondence would be accepted as a social fact by others as well confronts the actors who are disadvantaged by the correspondence with a hard reality . Whether the disadvantaged actors like it or not, others are likely to treat them as though the correspondence indicates a valid social truth about the actors' status worthiness and competence compared to those of the other nominal category . Confronted with this prospect, the disadvantaged actors have little choice but to accept the status belief themselves as an aspect of social reality with which they must contend . They may accept the disadvantaging status belief simply as an estimate of what "most others" believe . Or they may feel the pressure of social validity more deeply and accept the status belief as a matter of personal belief as well . Either way, however, the status belief will affect how they react to others in subsequent encounters . The creation of a sense of social validity about a link between a nominal distinction and status markers and competence, by this account, is the central mechanism by which interaction induces its participants to form status beliefs . The following belief formation assumption summarizes this argument . It provides a formal belief formation assumption for status construction theory that achieves two goals . It can account for the development of beliefs in situations where resource or other differences are present and where they are not . Furthermore, it specifies a mechanism by which interaction induces belief formation even among those the



The Formation of Status

Beliefs

89

beliefs disadvantage . This formal assumption replaces the informal principle of association between standing in a power and prestige order and belief formation stated earlier . Belief Formation Assumption :

Given a situation in which there is a socially valid correspondence between individuals' differing states of a nominal distinction and their differentiated status markers and expected competence, then there is a likelihood that actors in the situation will form corresponding status beliefs about the nominal distinction. Status beliefs are corresponding when the nominal state that is associated with more valued status markers and greater expected competence in the situation is also associated in the status belief with more positive social evaluation and greater competence compared to the other nominal state . To this formal assumption about belief formation, status construction theory adds an assumption from expectation states theory that actors transfer expectations about categories of actors formed in one setting to future encounters with actors that also differ in those categories (Markovsky, Smith, and Berger 1984 ; Pugh and Wahrman 1983) . Finally, status construction theory also assumes that subsequent encounters with others who differ on the attribute that confirm an actor's newly formed status belief strengthen that belief while inconsistent, disconfirming experiences undermine it . Together, these three assumptions constitute status construction theory's account of the formation and spread of status beliefs through interaction .

CHALLENGES TO CONSENSUS AND THE FORMATION OF STATUS BELIEFS In the above analysis the appearance of consensus creates social validity and social validity fosters the formation of status beliefs . Naturally, then, a question arises about the effect that a challenge to consensus in the situation would have on the likelihood that participants would form status beliefs . An examination of this question will clarify some implications of the belief formation assumption for status construction theory and suggest a means for testing hypotheses derived from the assumption . The reasoning below draws heavily on Zelditch and Floyd's (1998) insightful analysis of the relationships among consensus, dissensus, and the construction of social validity in a situation . If a participant speaks up to challenge an apparent correspondence in the situation between actors' nominal categories and their status markers and expected competence, what is the effect? The challenge might question the validity of the competence differences enacted between nominally different actors . Or the challenge might accept the competence and status marker differences but question



90

CECILIA L. RIDGEWAY

their link to the actors' nominal categories . Such a challenge might explicitly attribute the status and competence differences to some other difference between the actors or might suggest that the actors are exceptions to their nominal categories . However phrased, the challenge suggests that a link between the nominal distinction and evaluation and competence would not necessarily be accepted by everyone, creating uncertainty about its validity as a social schema. Zelditch and Floyd (1998) argue that the validity of something that has been challenged can be reaffirmed in the situation if the majority of the actors present appear to reject the challenge and support the thing that was challenged . An explicit action on the part of some participant to reject a challenge to the correspondence among actors' nominal categories and their status markers and expected competence would be necessary to reassert its validity . Most others present must appear to support this rejection by standing by the validity of the correspondence. If, on the other hand, no one speaks up to actively resist a challenge or if the majority in the situation do not support that resistance, then the social validity of the correspondence between the nominal distinction, status markers, and competence will crumble . Furthermore, even when the validity of the correspondence is reasserted in the situation, the fact that it was challenged places some limits on the range of actors that can be presumed to accept the correspondence, narrowing and weakening its social validity . Classic research on conformity suggests that a break from the appearance of unanimity in a group greatly reduces the pressure on those who privately disagree with the majority to go along with them (Asch 1955 ; Shaw, Rothschild, and Strickland 1957) . Thus any challenge to consensus about the correspondence between actors' nominal categories and their status markers and competence undermines the validity of that correspondence to some degree, especially in the eyes of actors who are not personally persuaded by the correspondence . Actors disadvantaged by the correspondence may be more likely to fall in the latter category . These considerations suggest hypotheses that may be derived from the formal belief formation assumption in order to test its assertion that social validity is a powerful ingredient in the formation of status beliefs . The assumption implies that actors will be less likely to form status beliefs about a nominal distinction in a situation where a correspondence among actors' nominal categories and their status markers and expected competence is challenged than in a situation where the correspondence appears consensual . In addition, in a situation in which a challenge to such a correspondence occurs, the greater the proportion of participants who act to support the challenge rather than the correspondence, the less the likelihood that the participants will form status beliefs about the nominal distinction . 3 Empirical verification of these hypotheses would support the assumption that the ability of interactional contexts to lend social validity to an association between nominal differences and status and competence is central to the process by which interaction creates status beliefs .



The Formation of Status

91

Beliefs

A GRAPH-THEORETIC REPRESENTATION OF BELIEF FORMATION With a formal specification of the mechanism by which interaction induces status beliefs in hand, the next step is to represent this mechanism in the terms of expectation states theory . Expectation states theory represents its formal arguments about the development of power and prestige orders in terms of theoretical graphs (Berger et al . 1977 ; see Webster and Whitmeyer 1999 for a recent explication) . My goal here is to adapt these theoretical graphs to represent status construction theory's assumptions about the formation of status beliefs in interaction .

Graphing Belief Formation Expectation states theory's graphs represent the way established status characteristics, reward levels, and behavioral patterns that are salient in a situation shape the expectations for task performance that a focal actor, P, holds for self compared to each other actor (0 1 . .0,) in the situation . When P and 0 differ on an existing status characteristic, such as educational attainment, that carries broad, diffuse status value in society and P has a more valued state of that characteristic (is better educated) than 0, expectation states theory graphs the situation as shown in the

P

D+

0

D

P

Na(+) (I'+/Y+) C*+

0

Nb(-) (F /Y) C*

C *+

T+

C*

T

a)

I'

T

Actors who differ on an existing diffuse status characteristic b). The formation of a status belief about a nominal distinction, N

Figure 1 . a) .



92

CECILIA L . RIDGEWAY

top of Figure 1 . This graph for an existing diffuse status characteristic provides a template for graphing the formation of a new status belief that creates a status characteristic . For clarity, let me explain Figure la. In it "possession" lines link P to a more valued state of the diffuse status characteristic (D+) and link 0 to a less valued state of that characteristic (D-) . Negative signed "dimensionality" lines link the two states of D indicating that they are contrasting states of the same characteristic . Widely shared status beliefs about D (say, education) link it to general expectations for competence (F), indicated by "relevance" lines connecting states of D to states of F . General expectations for greater competence from the better educated (F+) than the less educated (I'-) lead through relevance lines to expectations that P has more ability at the task at hand in this situation (C*+) while 0 has less of that ability (C*-) . Expectations about task ability lead to corresponding expectations for better task outcomes (T+) from P than from 0 (T-) . When actors form status beliefs in interaction, they impute differences between nominally distinct actors in some degree of general competence beyond that necessary for the specific situational task. In addition, these imputed general competence differences between actors are linked to the state of the nominal distinction that the actor possesses . As a first step in graphing this situation, the nominal distinction can be represented as "N" which divides actors into Nas and Nbs . When, in the formation of a status belief about N, actors link general competence to the nominal distinction, it transforms N into an evaluated distinction with a more and less valued state (i .e ., N+ and N- instead of just Na and Nb) . Therefore, the formation of a status belief about a nominal distinction, N, can be represented as an imputed line of relevance that links P and O's states of N and their presumed competence in the situation (C*-T) by means of an intervening assumption about their differences in some type of valued general competence . 4 In Figure lb this logic is used to graph the formation of a status belief about N . In this graph P is an Na and 0 is an Nb . P's presumed competence in the situation is higher (C*,-T,) than O's is (C*A-TA) . From their apparent differences in competence in the situation and their differences in N, P imputes (dashed relevance lines) corresponding differences in some valued general competence (I'/y) that P infers is linked to their states of N . When actors link general competence differences with a nominal distinction, they may assume, unless they encounter evidence to the contrary, that these competence differences apply broadly across most situational tasks and goals . In this case, expectation states theory would represent the imputed competence differences as a gamma (T'), for general competence . On the other hand, actors may impute competence differences that they assume apply to a specific range of tasks, represented as an upsilon (y) . The circumstances in which actors infer one type a competence difference rather than another is a question that is beyond the scope of the present chapter .



The Formation of Status Beliefs

93 GO

/T t + C*+ T2+



TZ

C*Ti -

0Z Figure 2 .

Formation of Status Beliefs After Two "Doubly Dissimilar" Encounters

From this foundation, we can develop graphs that represent the formation of status beliefs in two situations that have been investigated in research on status construction theory . In the first "doubly dissimilar" situation, actors differ in resources as well as the nominal distinction . After two encounters like this in which power and prestige orders developed that corresponded to resource (pay) differences, Ridgeway and colleagues (1998) found that actors formed status beliefs favoring the pay advantaged nominal category . Figure 2 represents an actor, P's, formation of a status belief about a nominal distinction, N, after two doubly dissimilar encounters . P is an Nb . 0 1 , who is an Na, is the actor P dealt with in the first encounter . All P's assumptions about 0 1 from the first encounter remain as a referent for P in the second encounter where P interacts with 02 who is also an Na . Thus lines representing P's and 0 1 's links to differential task success in the first encounter (i .e ., links to T 1 ) are present in the graph as well as lines linking P and 02 to expected task success in the second encounter (links to T2) . Since, in the situation studied by Ridgeway and colleagues (1998), the task was the same for both encounters, T 1 and T2 are linked to the same ability characteristic (C*) . Both 0 1 and 02 possessed higher levels of a valued goal object (GO+), pay, than did P (GO-) . Differences in pay imply different levels of social rewards (R+ and R-) between P and each 0 which turn imply differences in expected competence in each encounter (C*-T 1 and C*-T 2 ) . After two encounters where those who are better rewarded and appear more competent are Nas rather than Nbs, P infers a link (dashed lines) between states of N and general competence (17y), forming a status belief .5 In the second situation investigated by status construction theory, nominally different actors develop a power and prestige order on the basis of the behavioral inter-



94

CECILIA L . RIDGEWAY

/0 1

Na(+) --

Figure 3.

The Formation of Status Beliefs After Two Behaviorally Differentiated Encounters Between Nominally Different Actors

change patterns that emerge between them without supporting differences in resources . After two such encounters, Ridgeway and Glasgow (1996) found that subjects formed status beliefs favoring the nominal category that consistently had higher standing in the power and prestige orders of the encounters . A graph of belief formation in this situation is shown in Figure 3 . P, an Nb, has had two partners, 01 and 02, both of whom were Nas . Both partners acted confidently and assertively (b 1 + and b2+) in task discussions with P, casting P into reactive, less assertive behavior patterns (b l - and b2-) . As these assertive-reactive behavioral interchange patterns came to characterize the interaction between P and each 0, they evoked cultural images of status typified, leader-follower behavior (B+ and B-) . Status typified behaviors are culturally associated with differences in general problem solving ability (Y+ and Y-) which in turn suggests differences between the actors in competence at the task at hand . After two nominally different partners who seem more competent and act more leaderlike than P, P infers an evaluative link between the nominal characteristic and general competence, forming a status belief . Graphing Social Validity and Challenges

The belief formation assumption that I have proposed for status construction theory asserts that the social validity that an interaction lends to an association between actors' nominal differences and their status markers and expected competence is an important determinant of the likelihood that actors will form status beliefs from the experience . How can social validity and challenges to social validity be represented in graph theoretic terms?



The Formation of Status Beliefs

95

Graphs represent an interactional situation from the perspective of a given actor, P . In that sense, they do not directly represent collective attributes of the actors' situation, such as consensus or social validity . On the other hand, validity develops from the appearance of consensus that the situation presents to each of the actors . In the face of a challenge, it develops from the appearance of explicit majority support for something . Graphs can represent the appearance of consensus that a situation presents to an actor . They can do so through the relationship the graph portrays between an actor, P's, own first-order expectations for each person's relative performance in the situation and the second-order expectations communicated by another actor for those people's relative performances (Moore 1985 ; Troyer and Younts 1997 ; Webster and Whitmeyer 1999) . P's own expectation that "0 knows more about this task than I do" is a first-order expectation . Second order expectations are someone else's expectations about P and 0 that are communicated to P . Thus, P's assumption that "someone else thinks 0 knows more about this task than I do" is a second-order expectation . There is substantial evidence that actors presume consensus in an interactional situation unless they receive explicit information to the contrary (see Zelditch and Floyd 1998) . When actors form expectations for self and others based on events in interaction and people's salient attributes, they presume that these expectations are shared . Consequently, a graph of the situation from P's perspective that contains no explicit second order expectations from another represents a definition of the situation that is socially valid for P because P takes it to be consensusal . A similar graph that contains another's second order expectations that support P's own expectations for people in the situation actively confirms P's sense that this definition of the situation is consensual, strongly validating it for P (Berger et al . 1998) . Therefore, a graph that contains no second-order expectations that are inconsistent with P's first-order expectation represents a definition of the situation that P presumes to be consensual and socially valid . If the graph also represents a situation where nominally different actors are consistently linked to differentiated performance expectations and valued goal objects, there is a likelihood that P will form status beliefs in the situation . An actor who speaks up to challenge an apparent correspondence between actors' nominal categories and their status markers and expected competence can be represented as communicating second-order expectations about the nominally different actors that conflict with P's first order expectations . Figure 4 graphs one such situation . In it P, an Nb, observes 0 1 , an Na, and 0 2 , an Nb, establish a pattern of behavior between them in which 0 1 is influential and leaderlike while 02 defers and accepts 0 1 's influence. P forms differentiated competence expectations for 0 1 and 0 2 , creating a correspondence between actors' states of N and performance expectations that P presumes to be consensual and, therefore, socially valid . Another actor in the situation, 0 3 , speaks up to challenge the presumed consensus, questioning 0 1 's reasoning and asking to hear more from 0 203'S second order expectations for 0 1 and 0 2 are represented by wavy lines con-



96

CECILIA L . RIDGEWAY

Figure 4 . Second-Order Expectations Challenge The Validity of an Association Between a Nominal Distinction, N, and Expected Competence

fleeting 02 to task success and 01 to task failure . 6 Since the second-order expectations are inconsistent with the P's first-order expectations for nominally different actors, this graph represents a situation in which the likelihood that P will form status beliefs is reduced . Graph Theoretic Implications

As these analyses suggest, the degree of consensus or dissensus in a group can be portrayed by the degree of inconsistency between P's first-order expectations for actors in the situation and second-order expectations for those same actors that others make explicit in the situation . Given the belief formation assumption, a straightforward implication of the graph theoretic representations is the following. The greater the number and strength of second-order expectations for nominally different actors in a situation that conflict with P's first-order expectations for those actors, the less the likelihood that P will form status beliefs about the nominal distinction . It is also apparent that the overall consistency of a graph structure linking nominally different actors to differentiated task outcomes also has implications for the formation of status beliefs in that situation . By overall consistency, I mean the extent to which all lines linking nominally different actors to task outcomes link those actors to similarly evaluated outcomes . Lines representing first-order as well as second-order expectations are both included . Holding constant the strength of these lines (generally inversely proportional to their length), the greater the degree of consistency in a graph structure linking nominally different actors to evaluated outcomes, the greater the likelihood that P will form status beliefs about the nominal distinction . This more general implication of the graph theoretic representation of belief formation covers situations in which a consistent correspondence between actors' nominal differences and their status marker and expected competence differences never clearly develops as well as situations where such a correspondence is explicitly challenged .



The Formation of Status Beliefs

97

There is a third prediction about belief formation that can be derived from graph theoretic representations as well . Consider a graph that represents a completely consistent correspondence between actors' nominal differences and their links to evaluated task outcomes . The more elements linking nominally different actors to differentially evaluated task outcomes in such a graph, the more comprehensive it is (Berger et al . 1998) . These elements can be attributes of the actors, such as supporting resources or status characteristics, they can be referent others from previous encounters, or they can be additional others in the present situation . The point is that each of these elements provides another consistent link between actors' nominal differences and their expected competence . Therefore, the more comprehensive a graph structure is that links nominally different actors to evaluated outcomes in a consistent manner, the greater the likelihood that P will form status beliefs about the nominal distinction . Comprehensiveness affects belief formation partly by strengthening the expected competence differences between nominally different actors . However, since the impact on competence expectations of each new consistent element is progressively less, this strengthening effect attenuates as comprehensiveness increases . Comprehensiveness also affects belief formation by broadening the appearance of consensus about an evaluative difference between nominally distinct actors by portraying a wide array of social evidence that supports it. The reinforcement of validity through wider evidence may be comprehensiveness' most powerful effect on belief formation . These predictions derived from graph theoretical representations of the belief formation process suggest the value of transcribing status construction theory into terms that are compatible with expectation states theory .

BECOMING PART OF THE SOCIAL FRAMEWORK When status beliefs are widely held in a society, they become part of its taken-for-granted, roughly consensual culture . Expectation states theory argues that status beliefs which are established in the culture form part of the social framework of beliefs that actors bring to a situation of action and use to frame their behavior there (Berger, Wagner, and Zelditch 1982 ; Zelditch and Floyd 1998) . The social framework of beliefs is a cultural tool kit of taken-for-granted beliefs that actors use to make sense of interactional situations and the actors within them in order to orient their own behavior in the situation . Among other things, the social framework includes beliefs defining the socially significant categories of actors in the culture and their social meaning and status value in society (e .g ., beliefs about men and women, people of different ethnic groups, or different occupations) . Actors presume that beliefs in their social framework are consensual and act on them accordingly .



98

CECILIA L . RIDGEWAY

If certain interactional experiences are sufficient to induce status beliefs in their participants, how and when do these beliefs become stable elements in those actors' taken-for-granted social framework of beliefs? Due to their specific interactional experiences, individuals who acquire a status belief about a nominal distinction may come to presume that the belief is widely consensual before it actually is in the society . These individuals learn to take the belief for granted as part of their social framework of beliefs . In this way, individuals, due to the sequences of interactional experiences they have had, may "bootstrap" the process which the status beliefs become broadly consensual in society by presuming that it already is so . Status construction theory deals with this process by arguing that beliefs that actors form in one encounter are strengthened by future encounters that enact power and prestige orders between nominally different others that confirm the status beliefs (Ridgeway 1991 ; Ridgeway and Balkwell 1997) . Of course, by acting on the newly acquired status belief in future encounters with nominally different others, the individual substantially increases the likelihood that the status belief will, in fact, be supported in those encounters . Encounters with nominally different others that disconfirm the status belief undermine it, according to status construction theory, and reduce the likelihood that the individual will act on the belief in the next encounter . The extent to which the belief is undermined by a disconfirming encounter, however, depends on strength with which the belief is held . A belief which has been strongly established in prior experience will only be slightly weakened by a single disconfirming event . The belief formation assumption offered here suggests that the "strength" of status beliefs might better be conceived as the extent to which an actor presumes that belief to be consensually valid in a society . The broader the presumed consensuality of a status belief, the more the actor can rely on the belief to effectively frame any given interaction where it is relevant and salient (i .e ., encounters with those who differ on the status valued distinction) . Subsequent encounters with different tasks and actors that support a person's newly formed status belief confirm that the belief is not merely a locally valid construction of reality, but rather, a broadly valid interpretation of the nominal distinction that, it appears, would be accepted by almost anyone . It is an empirical question how many confirming experiences are necessary before a status belief appears sufficiently consensual and valid to an actor that it becomes a stable element in the actor's taken-for-granted social framework . I suspect, however, that the number is small, say three or four . The local reality of encounters and the social validity they convey are powerfully persuasive agents for individuals . I suspect that people generalize quickly from a few such persuasive experiences of local consensus to the presumption of a more general consensus . Indeed, without such a tendency to generalize quickly from local consensus, the presumption of general consensus about a status belief would be difficult to for actors to achieve . The presumption of general consensus about a social frame-



The Formation of Status Beliefs

99

work of beliefs, in turn, is fundamental to the emergence of social structure from routine interaction . It seems that people take on status beliefs from interaction because of the appearance of consensus in a local situation that makes an enacted correspondence between people's nominal differences and their status markers and competence appear to be a valid social fact. Subsequent encounters that confirm this status belief, even if they do so because of the self-fulfilling effects of an actor's own behavior, broaden the appearance of consensus about the status belief . As the appearance of consensus is validated across contexts and actors, the belief becomes part of the actor's accepted social framework . As more actors in the society take on the belief and make it part of their social framework, the chance of disconfirming experiences drops . At some point the status belief really does become roughly consensual in society in that almost everybody assumes that "most people" hold the belief and act on it. Thus, the persuasive power of the appearance of consensus in some local situations eventually creates a genuinely broad consensus in a society.

CONCLUSIONS Given the importance of status relations in society and the central role that widely held status beliefs play in organizing status relations, it is remarkable that sociologists have so little systematic evidence about the formation and change of status beliefs . Status construction theory posits a set of processes that are sufficient (although not necessary) to create status beliefs that become part of the taken-for-granted social framework of most actors in a population . As initially formulated, status construction theory offers some deceptively simple ideas about the formation, transfer, and strengthening and weakening of status beliefs in interaction and seeks to specify the social conditions under which they would lead to widely shared status beliefs about a nominal distinction . Empirical and theoretical developments now suggest that widely held status belief are indeed likely to develop when a nominal distinction becomes correlated in a population with an inequality in resources, as status construction theory originally posited . More generally, however, current research suggests that status beliefs are likely to develop when one nominal group is relatively advantaged compared to another in any factor, not just resources, that systematically biases the development of power and prestige when members from the two groups interact . Furthermore, chance processes alone may sometimes create status beliefs about a nominal distinction through path dependent processes . Such conclusions about the development and spread of status beliefs under different social conditions depend, however, on the nature of the processes by which interaction induces actors to form status beliefs in the first place and "teach" them to others through interaction . Thus, status construction theory turns on its assump-



1 00

CECILIA L . RIDGEWAY

tions about the processes by which interaction causes people to form status beliefs about nominal differences . To develop status construction theory into a useful account of the development and change of status beliefs, it is important to provide a clear theoretical specification of the process by which interaction creates status beliefs and suggest means for testing this specification . This has been my goal here . Analyzing the mechanism by which an interactional context induces status beliefs in its participants even when the belief disadvantages them highlights the importance of the appearance of consensus and social validity that the context provides for those who participate in it . I argue that it is the apparent social validity of a correspondence between participants' states of a nominal distinction and their differentiated status markers and expected competence that persuades them to infer a status belief about the distinction . On the basis of this argument I have proposed a formal belief formation assumption for status construction theory and derive testable hypotheses from it . I have shown as well how this assumption can be represented in the graph theoretic terms of expectation states theory, allowing further predictions to be derived . The next step is to conduct empirical tests of these hypotheses . To understand the role of status relations in social stratification and inequality, we must understand how structural conditions, social distinctions, and status beliefs co-evolve in society . Gaining a fuller theoretical and empirical understanding of the power of interaction to create status beliefs under different structural conditions is a first step in this project . From it, we may begin to assemble the tools as well to understand the conditions under which status beliefs change in a society .

NOTES 1 . In this experiment as well as those reported in Ridgeway and Glasgow (1996) and Ridgeway and Erickson (1998), a nominal distinction was created by giving subjects a test purportedly measuring "personal response style." Subjects were told that personal response style is a stable characteristic of individuals that divides the population into two roughly equal sized categories . "Results" of the test showed that subjects and partners differed in their personal response style category . 2. Zelditch and Walker (1984) refer to this form of validity as endorsement because it is based on apparent support by an actor's peers in the situation . 3 . As a simplification, this assumption holds constant any status differences among the participants. It is likely that the support of high status participants will have a greater impact than the support of lower status participants . Incorporating the impact of status differences among the participants is a project for another paper . 4. Expectation states graphs represent actors' competence expectations for one another rather than their power and prestige behaviors because power and prestige behaviors reflect competence expectations. As a consequence, an expectations states graph of the formation of status beliefs represents more explicitly the development of a link to competence expectations than to status markers and evaluation (although the latter is implied) . However, the graph does explicitly represent the formation of an evaluative dimension to the (formally) nominal distinction by indicating that Na and Nb become N+ and N- .



101

The Formation of Status Beliefs

5 . Webster and Hysom (1998) suggest that people in doubly dissimilar situations may also form status beliefs by forming referential beliefs about N . That is, they may form beliefs that Nas generally possess more valued social rewards than do Nbs in society . In a graph this would be represented by the imputation of a link between states of the nominal distinction and levels of goal objects (GO+ and GO-). This is a plausible suggestion that awaits empirical investigation . 6 . The lines are wavy because the strength of the impact of second-order expectations on P's own expectations varies with strength of P's expectations for the person who communicates the secondorder expectations (Fisek, Berger, and Norman 1995 ; Webster and Whitmeyer 1999). I will not deal with this aspect of second-order expectations here, however . 7 . Expectation states theory refers to this as the size of the performance expectation advantage in the situation that actors of one nominal category have over those of another category .

REFERENCES Asch, S . 1955 . "Opinions and Social Pressure ." Scientific American 193 : 31-35 . Berger, J ., T . L . Conner and M . H. Fisek. 1974 . Expectation States Theory: A Theoretical Research Program. Cambridge, MA : Winthrop. Berger, J ., M. H . Fisek, R. Z . Norman, and M . Zelditch, Jr . 1977 . Status Characteristics and Social Interaction. New York: Elsevier . Berger, J ., C . L. Ridgeway, M . H. Fisek, and R. Z . Norman . 1998 . "The Legitimation and Delegitimation of Power and Prestige Orders ." American Sociological Review 63 : 379-405 . Berger, J ., D . Wagner, and M . Zeldtich, Jr. 1982 . "A Working Strategy for Constructing Theories : State Organizing Processes ." Pp . 107-123 in Metatheorizing, edited by G. Ritzer . Newbury, CA : Sage . Berger, J ., and M . Zelditch, Jr . 1998 . Status, Power, and Legitimacy: Strategies and Theories. New Brunswick, NJ : Transaction Press . Brewer, M ., and R . Kramer. 1985 . "The Psychology of Intergroup Attitudes ." Annual Review of Psychology 36 : 219-243 . Conway, M . M ., T. Pizzamiglio, and L . Mount. 1996 . "Status, Communality, and Agency : Implications for Stereotypes of Gender and Other Groups ." Journal of Personality and Social Psychology 71 : 25-38 . Cook, K . S . 1975 . "Expectations, Evaluations, and Equity ." American Sociological Review 40 : 372-88 . Dovidio, J . F., and S . L . Gaertner. 1993 . "Stereotypes and Evaluative Intergroup Bias ." Pp . 167-194 in Affect, Cognition, and Stereotyping, edited by D. M . Mackie and D . L . Hamilton . New York : Academic Press . Harrod, W . J . 1980. "Expectations from Unequal Rewards ." Social Psychology Quarterly 43 : 126-30. Fisek, M . H ., J . Berger, and R . Norman . 1991 . "Participation in Heterogeneous and Homogeneous Groups : A Theoretical Integration ." American Journal of Sociology 97 : 114-42. Fisek, M . H ., J . Berger, and R . Norman . 1995 . "Evaluation and Formation of Expectations ." American Journal of Sociology 101 : 721-746. Gerber, G . 1996 . "Status in Same-gender and Mixed-Gender Police Dyads : Effects on Personality Attributions ." Social Psychology Quarterly 59 : 350-63 . Geis, F . L ., J. Jennings, and D . Corrado-Taylor . 1984 . "Sex vs . Status in Sex-Associated Stereotypes ." Sex Roles 11 : 771-85 . Mark, N. 1999 . "The Emergence of Status Inequality ." Paper presented at the annual meeting of the American Sociological Association, Chicago, August . Mark, N., L. Smith-Lovin, and C . L . Ridgeway. 1999. "The Emergence of Status Inequality ." Unpublished paper. Stanford University .



102

CECILIA L . RIDGEWAY

Markovsky, B., L . F. Smith, and J . Berger. 1984 . "Do Status Interventions Persist?" American Sociological Review 49: 373-82. Messick, D . M ., and D . M . Mackie . 1989 . "Intergroup Relations." Pp . 45-81 in Annual Review of Psychology, Vol. 40, edited by M. R . Rosenzweig and L . W. Porter. Palo Alto, CA : Annual Reviews Press. Moore, J . C . 1985 . "Role Enactment and Self Identity : An Expectations States Approach ." Pp . 262-316 in Status, Rewards, and Influence : How Expectations Organize Behavior, edited by J . Berger and M. Zelditch . San Francisco: Jossey-Bass. Pugh, M . D ., and R. Wahrman. 1983 . "Neutralizing Sexism in Mixed-Sex Groups : Do Women Have to Be Better than Men?" American Journal of Sociology 88 : 746-62. Ridgeway, C . L. 1991 . "The Social Construction of Status Value : Gender and other Nominal Characteristics ." Social Forces 70 : 367-86. Ridgeway, C . L . 1997 . "Where Do Status Value Beliefs Come From? New Developments ." Pp. 137-158 in Status, Network, and Structure, edited by J . Szmatka, J . Skvoretz, and J . Berger . Stanford, CA: Stanford University Press . Ridgeway, C . L., and J. W . Balkwell . 1997. "Group Processes and the Diffusion of Status-Value Beliefs ." Social Psychology Quarterly 60 : 14-31 . Ridgeway, C . L ., and J . Berger . 1986 . "Expectations, Legitimation, and Dominance Behavior in Task Groups ." American Sociological Review 51 : 603-17 . Ridgeway, C. L ., J . Berger, and L. Smith . 1985 . "Nonverbal Cues and Status : An Expectation States Approach ." American Journal of Sociology 90 : 955-78 . Ridgeway, C . L ., E. H . Boyle, K. Kuipers, and D . T . Robinson . 1998. "How Do Status Beliefs Develop? The Role of Resources and Interactional Experience ." American Sociological Review 63 : 331-50 . Ridgeway, C . L ., and K . Glasgow . 1996 . "Acquiring Status Beliefs from Behavior in Interaction ." Paper presented at the meeting of the American Sociological Association, New York, August . Ridgeway, C. L ., and K. G. Erickson . 1998. "Third Party Observers and the Spread of Status Beliefs in Interaction ." Paper presented at the International Sociological Association 14 World Congress of Sociology, Montreal, July . Shaw, M . E ., G . H . Rothschild, and J . F. Strickland . 1957 . "Decision Processes in Communication Nets ." Journal of Abnormal and Social Psychology 54: 323-330. Stewart, P . A ., and J . C . Moore, Jr . 1992. "Wage Disparities and performance Expectations ." Social Psychology Quarterly 55 : 70-85 . Troyer, L ., and C . W . Younts. 1997 . "Whose Expectations Matter? The Relative Power of First-Order and Second-Order Expectations in Determining Social Influence ." American Journal of Sociology 103 : 692-732 . Wagner, D ., and J. Berger . 1997. "Gender and Interpersonal task behaviors : Status Expectation Accounts ." Sociological Perspectives 40 : 1-32 . Weber, M . 1968. Economy and Society, edited by G. Roth and C . Wittich and translated by E . Fischoff et al . New York : Bedminster Press . Webster, M., and M . Foschi. 1988. Status Generalization : New Theory and Research . Stanford, CA : Stanford University . Webster, M ., and S . J. Hysom . 1998 . "Creating Status Characteristics." American Sociological Review 63 : 351-79. Webster, M ., and J. M . Whitmeyer. 1999 . "A Theory of Second-Order Expectations and Behavior." Social Psychology Quarterly 62 : 17-31 . Zelditch, M., Jr. and A . S . Floyd . 1998 . "Consensus, Dissensus, and Justification ." Pp. 339-68 in Status, Power, and Legitimacy : Strategies and Theories, edited by J . Berger and M. Zelditch . New Brunswick, NJ: Transaction Press. Zelditch, M ., Jr. and H. A . Walker. 1984 . "Legitimacy and the Stability of Authority ." Pp . 1-25 in Advances in Group Processes, vol . 1, edited by E . J . Lawler . Greenwich, CT: JAI Press .

THE IMPACT OF TASK AND CATEGORICAL CUES ON SOCIAL INFLUENCE FLUENCY AND ETHNIC ACCENT AS CUES TO COMPETENCE IN TASK GROUPS

Margaret Foddy and Phoebe Riches

ABSTRACT A theoretical extension of Berger, Webster, Ridgeway, and Rosenholtz's (1986) theory of status cues is proposed, which argues that task cues which contradict expectations based on categorical cues may reduce or overcome the effects of the latter on status processes . Possible mechanisms for these differential effects were suggested . Two studies examined the relative impact of verbal fluency and ethnic accent on perceptions of competence, and on acceptance of influence in a group task . In Study 1 tape-recorded passages read by Greek- and Anglo-Australian speakers were judged by independent groups on rating scales which tapped three dimensions : intelligence, competence/education, (referred to as status dimensions), and solidarity . Fluency of speaker affected raters' judgments of confidence and intelligence, but

Advances in Group Processes, Volume 17, pages 103-130 . Copyright © 2000 by JAI Press Inc. All rights of reproduction in any form reserved . ISBN : 0-7623-0651-3 103



104

MARGARET FODDY and PHOEBE RICHES

not solidarity ; ethnic identity of speaker affected only ratings of intelligence . Tape recordings of fluent and stilted versions of a conversation were used in Study 2, which employed a computer controlled joint decision task . Subjects were required to resolve prearranged disagreements with a partner whose ethnicity (Anglo-Australian or Greek-Australian) and fluency (fluent or stilted) were cued by means of an overheard conversation . It was predicted and found that fluency would have a stronger impact on deference toward the partner than ethnic accent, and fluency affected post-task ratings of confidence . Partner's ethnicity did not affect rates of deference, but did affect post-task ratings of the partner : Anglo-Australians were rated as significantly more educated, quick, intelligent, and confident by both Anglo- and Greek-Australians .

INTRODUCTION When we first encounter people in a work setting or at a more informal occasion, we quickly process a great deal of information from their speech and appearance . Even before they speak, we will have noted such things as gender, race, age, and attractiveness from their physical characteristics, and perhaps socioeconomic status and sexual orientation from their dress . While the content of their initial conversation is very important, so too are the paralinguistic and nonverbal cues which accompany the words. Ethnic and class accent will add important information to our emerging impression of the person . Posture, gaze, and general demeanour will tell us whether the person is comfortable and at ease, or nervous and unconfident . Speech characteristics, such as fluency, loudness, and pitch will either confirm or contradict these initial impressions . As the televised debates of various U .S . presidential candidates showed, these latter cues can be decisive in the judgments made about the abilities and trustworthiness of well-known figures ; they are likely to be even more important in first impressions of strangers . People are also able to form strong and reasonably accurate impressions of others from hearing them speak. In an early analysis, Allport and Cantrill (1934) cited a report in the New York Times Magazine of June 18, 1933 : "The human voice, when the man [sic] is not making conscious use of it by way of impersonation, does in spite of himself, reflect his mood, temper and personality ." Although the study referred to radio speakers, another early study by Allport and Cantrill showed that listeners were quite prepared to judge both physical traits and personality of ordinary speakers from tape recordings . Further, they were more accurate than chance in judging personality dimensions such as extraversion/introversion and ascendance/submission . These authors noted the importance of stereotypes in the judgment process, with listeners including several related features once they had placed a person into a particular category ; this often reduced accuracy, but increased consensus among listeners as to the meaning of different vocal characteristics (see also Pear 1931) .



Fluency and Ethnic Accent as Cues to Competence in Task Groups

1 05

Subsequent studies of paralinguistic and nonverbal cues have demonstrated clearly that nonverbal cues, including paralanguage, ) have a strong influence on the perception and evaluation of others (Ball, Byrne, Giles, Berechree, Griffiths, Macdonald, and McKendrick 1982 ; Callan, Gallois, and Forbes 1983 ; Giles and Sassoon 1983 ; Ridgeway, Berger, and Smith 1985 ; Sebastian and Ryan 1985 ; Driskell, Olmstead, and Salas 1993 ; Rashotte and Smith-Lovin 1997) . Researchers studying the social impact of speech characteristics have focused on how cues such as fluency, loudness, ethnic accent, and "correctness" of speech affect different dimensions of person perception . For example, Lambert (1967 ; Lambert, Hodgson, Gardner, and Fillenbaum 1960), in his studies of ethnic accent, argued that perceptions of ethnic speech reflect two dimensions : status (competence, dominance, power, status, and achievement) and solidarity or affiliation (friendliness, solidarity, affiliation, and helpfulness) . 2 The accents of higher status groups evoke more favorable reactions from both low and high status raters on the status dimension . Other studies have shown that unfavorable evaluations are given to the speech patterns of regional speakers, ethnic group members, and lower-class populations in terms of status and prestige, from judges who may or may not be standard speakers themselves (Callan et al . 1983 ; Sebastian and Ryan 1985 ; Giles and Coupland 1991 ; Cargile, Giles, Ryan, and Bradac 1994) . Ratings on the solidarity dimension reflect a tendency for people to favor members of one's own group, although evidence for this latter relationship is mixed (Ryan and Carranza 1975 ; Callan et al . 1983 ; Nesdale and Rooney 1996) . The two dimensions, status and solidarity, also appear in ratings of speech qualities such as fluency, speed, and loudness, with higher status and competence being assigned to speakers who are fluent, loud, and fast . Several studies (Sereno and Hawkins 1967 ; Smith, Brown, Strong, and Rencher 1975 ; Brown, Giles, and Thakerer 1985) have shown that rate of speech influences listeners' perceptions, with judgments of competence a linear function of increasing rate . Similarly, fluency has been shown to affect perceptions of competence . For example, Pryor and Buchanan (1984) assessed the impact of nonverbal behavior on jurors' evaluations of the defendant's credibility and guilt and found the defendant received the highest competency ratings in the setting where he was highly fluent, maintained good eye contact, and did not fidget. Norton-Ford and Hogan (1980) demonstrated that fluent speech produced significantly higher ratings than nonfluent speech on adjectives such as assertiveness, competence, and aggressiveness (see also Rashotte and Smith-Lovin 1997) . Solidarity (or affiliation) ratings seem to be a more complex function of speech qualities . Existing evidence suggests that the relationship between rate of speech and solidarity ratings, for example, is sensitive to contextual variables such as the type of audience the speaker is addressing (Brown et al . 1985) . These effects may be best explained in terms of speech accommodation theory (Giles, Mulac, Bradac, and Johnson 1986), which predicts a positive evaluation of people who adjust their speech to suit the needs of the listener . For purposes of this paper, the effects



106

MARGARET FODDY and PHOEBE RICHES

of nonverbal cues on the perception of solidarity are of less interest than their impact on perceived status, because it is argued that, in task groups, competence at the group task is more important to status and status-linked behavior than liking derived from shared group membership .

NONVERBAL CUES AS STATUS CUES More recent research has indicated that perceptions of competence and status reflecting the "status dimension" in nonverbal and paralinguistic cues also affect subsequent behavior in task settings (Tuzlak and Moore 1984 ; Ridgeway 1987 ; Riches and Foddy 1989 ; Sev'er 1989, 1991) . Cues which evoke impressions of greater intelligence, competence, and prestige (Tuzlak 1989 ; Riches and Foddy 1989) evoke greater deference or compliance from listeners . This research arose in part from an interpretation of nonverbal and paralinguistic cues in terms of status characteristics theory (Ridgeway et al . 1985 ; Berger et al . 1986 ; Ridgeway 1987 ; Rashotte and Smith-Lovin 1997) . Status characteristics are attributes that vary among the actors in a group, and are "differentially valued in terms of honour, esteem or desirability" (Berger, Rosenholtz, and Zelditch 1980) . They can be specific to a task (e .g ., high compared with low skill at a statistical task), or diffuse, such that people possessing the positive state of the characteristic are not only believed to have some specific abilities, but are also assumed to be generally better and more competent than those possessing the negative state (e.g ., adults are assumed to be more competent than children) . Gender, race, SES, and physical attractiveness have been identified as diffuse status characteristics, as has ethnicity (Foschi and Takagi 1991 ; Riches and Foddy 1989) . Status characteristics theory claims that members of a task group3 take all the status information about each performer, which has become salient in the situation, and combine it to form an aggregated expectation for performance at a group task.4 According to status characteristics theory, diffuse status characteristics continue to exert an influence even when more specific information about task-relevant ability is available . As noted below, one process by which this occurs is through the impact of diffuse status on nonverbal behaviors . In task groups relative expectations for task performance determine the emergence of a power and prestige order : those for whom higher expectations are held are given more opportunities to contribute, their performances are evaluated more positively, and they exert more influence over group decisions (Berger et al . 1980 ; Berger and Zelditch 1985) . Berger and colleagues (1986) argued convincingly that the theory of status characteristics and expectation states could be used to integrate the seemingly disparate set of findings about nonverbal and paralinguistic cues, by seeing them as part of the process by which relative status is expressed or enacted .



Fluency and Ethnic Accent as Cues to Competence in Task Groups

1 07

Berger and colleagues (1986) outlined a typology of verbal, nonverbal, and paralinguistic cues, providing a link between this typology and the general theory of status characteristics and expectation states . They distinguished task and categorical cues . 5 According to Berger and colleagues (1986), task cues are nonverbal and paralinguistic behaviors "that give information about performances taking place in the immediate interaction situation" (p. 5) . Task cues include such things as verbal fluency, loudness, eye gaze, and various gestures which give an indication of how well the actor is doing, or will do, on the group task . In studies of nonverbal cues people giving off high task cues tend to be those who come to the interaction with high external status, or who come to hold higher status within the group . Higher status people speak more quickly, interrupt more, have fewer speech dysfluencies, and so on, and people exhibiting this profile are seen to have higher status and associated task competence (Dovidio, Brown, Heltman, Ellyson, and Keating 1988) . Greater fluency has been associated with higher ratings of competence and dynamism (Sereno and Hawkins 1967) . Regardless of whether a confident speaker is in fact more competent, actors and perceivers act as if such a correlation exists . In terms of status characteristics theory, task cues are read as evidence for the level of more general problem-solving ability which becomes relevant to the immediate task situation . In other words, task cues affect expectations for good or poor performance on the task at hand . There is little research concerning the relative importance of various task cues for impressions of competence, although fluency and rate of speech are clearly important cues (Sereno and Hawkins 1967 ; Cargile et al . 1994) . Berger and colleagues (1986) distinguish between "speech task cues" and "behavioral task cues," but their central point is that while these are superficially very different, they may have very similar effects . While task cues seem to signal possession of competencies relevant to the task at hand, Berger and colleagues (1986) claim that they are "governed" by relative position on external status characteristics brought to the group . These may be signaled by categorical cues, which are signs that a person is a member of a particular social group, or "category," and include things such as skin color, visible indicators of sex or age (Ridgeway et al . 1985), as well as class and ethnic accent, modes of dress, and other status indicators . In the framework of status characteristics theory, categorical cues such as class and ethnic accent have their effect by establishing the relative external status of speakers and listeners, so that an upperclass accent, for example, cues the listener that the speaker possesses the positive state of a diffuse status characteristic, with the attendant positive expectations for performance on a group task . Riches and Foddy (1989) showed that listeners perceived an ethnic Greek-Australian accent as signaling lower education, intelligence, and competence than an Anglo-Australian accent . According to the account offered by Berger and colleagues (1986), cues will come to coincide with any differentiation established by categorical cues to diffuse and specific status characteristics . This is an important theoretical proposition, which suggests that many of the diverse findings about the nonverbal



108

MARGARET FODDY and PHOEBE RICHES

behavior of men and women, people of different military rank, people of different age, and so on can be explained by a more general proposition about how relative societal statuses external to the group become salient in the group setting, and form the basis for performance expectations . Thus, it is predicted that cues such as confident speech patterns are not stable individual propensities, but rather, expressions of the actors' beliefs about who is likely to know more about the task at hand in any given situation . In turn, these beliefs reflect aggregated status characteristics incorporating both specific and diffuse status characteristics of speaker and listener . For example, a tutor who speaks fluently and with confidence to a group of freshmen students, may make only tentative and stumbling contributions in a staff meeting with senior colleagues . Women may display authoritative task cues on a female task, but be deferential when dealing with tasks more associated with men .

INCONSISTENCY BETWEEN TASK CUES AND CATEGORICAL CUES Less studied, but equally important theoretically, is the possibility that relative expectations for competence based on categorical cues may be altered or reversed if task cues are inconsistent with them . What will happen if a person with the lower state of a diffuse characteristic, for example, race, exhibits a high level of task cues, while the person possessing the higher state of the diffuse characteristic (white) does the reverse? The status dominance principle implies that such a combination is unlikely, but status characteristics theory provides mechanisms for aggregation of both consistent and inconsistent status information . Because of their greater "immediacy," task cues may contribute more to the formation of relative expectations than categorical cues . When status information is being combined, some information is given greater weight : characteristics which are more directly relevant to the task at hand have greater impact on expectations (see Humphreys and Berger 1981 ; Berger et al . 1986) . Connections between characteristics and expectations for task performance are represented as "paths of relevance" of differing length ; characteristics with lengthy paths are weighed less in the aggregation of status information than are those with shorter paths . Berger and colleagues (1986) suggest that task cues should have a greater effect than categorical cues on expectations for task performance, other things being equal, although there is no evidence available about this "strength of relevance" prediction for cues (p . 14) . Aside from intuitions that task cues are more "immediate," the relatively greater impact of task cues may rest on a belief that such cues are determined by one's diffuse and specific status characteristics . Such a belief implies that incompetent people cannot act confidently, and that because people with low diffuse status (women, blacks in mixed-sex or mixed-race groups) will feel relatively incompetent, they will be unable to control



Fluency and

Ethnic Accent as Cues to Competence in Task Groups

109

the expression of these self-evaluations . Unfortunately, the research literature provides only a small number of studies of the effect of task cues on perceptions of status and competence (e .g ., Ridgeway 1987 ; Tuzlak and Moore 1984 ; Sev'er 1989) . Sev'er's work compared the relative impact of a large number of task cues (demeanor, gaze, verbal fluency) and a diffuse status characterisic, race . She found that while demeanor could overcome the negative effect of race on rates of influence in a mock jury setting, both demonstrated confidence and race affected ratings on a range of attributes, such as confidence and task capacity (but not intelligence) . What are the possible mechanisms for such differential impact? First, task cues may have a shorter path of relevance to task outcomes than do categorical cues . Expressed nonverbal confidence may lead an observer/listener to infer higher levels of intelligence and general ability in the actor . The numerous studies of speech characteristics and speaker evaluations (e .g., Cargile et al . 1994) provide evidence for a direct link between paralinguistic cues such as rate of speech, loudness, hesitations (confidence), and perceived competence/intelligence . Many of these studies were designed to assess the correlation between speech characteristics and "personality," and so the measured variables may skew the results in favor of inferences to internal traits . Nevertheless, it is plausible that an observer/listener may make a link between task cues and generalized competence . Berger and colleagues (1986) make the explicit claim that "the path of relevance linking task cues to abstract task ability to task outcomes is shorter than that linking status characteristics to task outcomes" (p . 14) . They argue that task cues such as fluency, speech rate, tone of voice, and so on activate states of more general problem-solving ability, and that these states are applicable to the immediate situation . Confidence displayed in a task situation through expressive task cues, for example, may be seen to be "possessed" by an individual, and to induce a link directly through abstract task ability, to the specific task characteristic C*, or instrumental ability to produce a positive task outcome, T(+) . As noted above, such confidence is normally encountered in those with an expectation advantage, that is, those who also possess more positive states of diffuse and specific status characteristics . However, task cues indicating confidence may come to signal task competence in their own right . In this interpretation, when task cues are the only information that a person has about another actor, higher levels of expressed confidence will create relatively high expectations for a positive outcome on the task . This process of inference is very general : nonverbal confidence (paralinguistic, demeanor, gaze, etc .) in any task setting would generate inferences of higher task ability than unconfident cues . The key point is that confident task cues produce a strong inference to a positive task outcome, in contrast to hesitant cues . This is the first interpretation of the term "immediate ." Because there is no existing convention for the representation of task and categorical cues, we use the graph representations of paths of relevance from Fisek, Berger, and Norman (1991) . These authors introduced the concept of "behavior



110

MARGARET FODDY and PHOEBE RICHES

patterns" which may signal higher or lower status . One person may make a suggestion, which the other accepts ; this differentiates the actors and may become salient, especially if no other status characteristics do . 6 Fisek and colleagues argue that behavior patterns are linked to generalized task ability, and that behavior patterns may therefore link actors to task outcome through a burden of proof process (see Skvoretz and Fararo 1996) . We make a parallel argument that task cues may also enter as behavioral elements in the task situation . A person who makes firm, confident statements is making a claim that she has the ability to make a positive contribution ; a hesitant style cues a willingness to let others take the lead . We may represent this as follows : p b (+) Y (+) C*(+) T(+) o b (-) Y (+) C*(-) T (-) p = focal actor; o = other; b = task cue (behavior pattern) ; Y = abstract task ability ; C* = task characteristic ; T = task outcome . A second interpretation is that task cues directly signal the person's state of the instrumental task characteristic in the immediate situation . In this case, a person who knows that he or she can do the task would be expected to sound confident ; if the person sounds confident, it indicates competence to do the task . On this interpretation, the path of relevance would be even shorter-there would be a direct link from the person to the instrumental task ability, to the task outcome : p b (+) C* (+) T (+) o b (+) C* (-) T (-) p = focal actor ; o = other ; b = task cue (behavior pattern) ; C* = task characteristic ; T = task outcome. Here the impact of task cues would be very strong indeed . However, our interpretation of the literature on the effects of task cues such as fluency and rate of speech, for example, suggest that they are mediated through their effect on evaluations of actors as possessing generalized competence . In addition to the information they provide about confidence in the immediate situation, some task cues may also have links to diffuse and specific status char-



Fluency and Ethnic Accent as Cues to Competence in Task Groups

111

acteristics . For example, fluent speech may also signal high levels of education and dominant social class ; particular intonations may be associated with high status backgrounds as well as immediate confidence . As a result there may be multiple paths of relevance to the task outcome, and these may contribute to the strong impact that task cues have on expectations for ability . Because most of the literature on nonverbal cues is not organized in terms of their effect on status organizing processes, it is not possible to make a strong case that task cues have more paths of relevance than do categorical cues ; this is an empirical question which should be investigated . In contrast to task cues, categorical cues may have a relatively long path of relevance to task outcome (unless they signal a very specific relevant status characteristic (e .g ., a physician's diploma on the wall in a medical setting) . Status characteristics theory predicts that they will be given less weight in the formation of expectations than will task cues . Skin color and ethnic accent provide information about which states of a diffuse status characteristic a person possesses . These are argued by Berger and colleagues to be less immediate and to have lower impact on expectations for competence than cues such as demeanor, gaze, speech fluency, loudness, and so on. A graphical representation might be : p categorical cue (+) D (+) I' (+) o categorical cue (-) D (-) F (-)

C* (+) C* (-)

T (+) T (-)

p = focal actor ; o = other ; D = diffuse status characteristic ; F = generalized expectation state ; C* = task characteristic ; T = task outcome . Even if length of paths of relevance are the same for task and categorical cues, the former may be in some sense "stronger," reflecting Berger and colleague's (1986) notion of greater "immediacy ." Foddy and Smithson (1996) outlined three different ways in which paths of equivalent length might produce differing strength of paths of relevance ; one of these was that different indicators of abilities might vary in their "diagnosticity" for task performance, producing more certain inferences about competence . An analogous idea may be applied to cues : some cues, such as task cues, may be regarded as more diagnostic or predictive of performance in the immediate situation, and be treated as "strong cues," while category memberships may be "weaker ." Thus, whether task cues and categorical cues have close or distant relationships to task performance, task cues may have a greater impact on aggregated expectation states . There is currently no way to represent the idea of "stronger paths" in the graph theoretical framework for status



112

MARGARET FODDY and PHOEBE RICHES

characteristics theory (Foddy and Smithson 1996) ; this possibility is not pursued in the current study, but left to future research . The relative impact of task and categorical cues is difficult to detect when they are consistent, but a clearer picture may be gained when they lead to inconsistent expectations . Thus, while we have noted that task cues may usually be governed by categorical cues, there are situations in which they may provide contradictory and inconsistent information . This is the subject of the research reported below . We now state these ideas in the form of general hypotheses, and then outline a more specific test situation . There we examined the relative influence of ethnic accent, a categorical cue, and fluency, a task cue, on evaluations of speakers and on acceptance of influence in a dyadic decision-making task . We take as background the propositions of Berger and colleagues (1986), which claim that : (1) in groups which are initially equal in external status, task cues will determine the power and prestige order which emerges ; and (2) in groups where external status is not equal, aggregated external status will govern task cues in interaction, and that these will contribute to the emergence of the power and prestige order . Differences in the resulting power and prestige order will reflect status differences ; people with higher aggregated status will receive and accept more action opportunities, receive higher evaluations, and exert more influence because they will have a higher net expectation advantage compared to those with lower status . Because our research also investigated perceivers' evaluations of speech, our hypotheses contain reference to evaluations .

HYPOTHESES Hypothesis 1 . When actors are differentiated on categorical cues to status characteristics, their relative expectations for task competence will be consistent with relative evaluations of these status characteristics . Hypothesis 2 . Relative expectations for task competence will be consistent with task cues exhibited by group members . Hypothesis 3 . Combination of task and categorical cues : Task cues will have a greater impact on relative evaluations and expectations for competence than will categorical cues . When a task cue is inconsistent with a categorical cue, net evaluations and expectations of actors will be consistent with the task cue . When multiple task and categorical cues are present, expectations will be formed according to the aggregation principles of status characteristics theory (Humphreys and Berger 1981) . To provide an initial test of these hypotheses we apply them to the case of ethnic accent and speech fluency . We argue that these paralinguistic cues are systematically interpreted as cues to diffuse status characteristics and general task



Fluency and Ethnic Accent as Cues to Competence in Task Groups

113

ability, respectively . Ethnic accent is a categorical cue that a person is a member of a particular ethnic group ; ethnicity is in turn a diffuse status characteristic, when it has associated with it differential evaluations and performance expectations . When people have to make decisions about who is likely to be competent in a novel task setting, they will generalize from their evaluations and expectations about ethnic groups, to consistent beliefs about relative competence of members of the ethnic groups on various group tasks . Ethnic accent has its effects on status/ competence beliefs, by acting as a cue that a given speaker is a member of a particular category (see Riches and Foddy 1989 for a fuller discussion) . As outlined above, fluency of speech is a task cue-people who speak without hesitation and relatively quickly are considered to be more confident and competent than those who do not . While higher status people may normally speak more fluently, we examine the case in which these two cues are deliberately made independent, in order to assess whether expectations evoked by external category membership may be changed or canceled by a more immediate task cue . In the research reported below, we tested the prediction that a task cue (fluency) would have greater impact on evaluations of speakers than would a status cue (ethnic accent) . We also predicted that expectations for performance, and thus rates of rejection/acceptance of influence would also show these effects . An equally important result will be if we find that the two sorts of cues have equal impact. This will suggest that they should be represented as having paths of relevance of equal length and strength . Aside from their importance to the theoretical framework outlined above, studies of the relative impact of different sorts of nonverbal and paralinguistic cues are needed . Most research to date has focused on single cues, or has sought simply to establish that a number of cues has an effect . There is no systematic evidence, for example, concerning which of several fluency cues (pauses, disfluencies, speed, etc .) has the greatest impact on ratings of fluency, or on impressions of competence . Nor is it known how fluency cues interact systematically with other speech characteristics to create overall impressions . The status characteristics framework provides a basis from which investigations of the combinations of cues and their relative impact can proceed . We report the results of two experiments . The first assessed the relative impact of fluency and ethnic accent on perceptions of competence/status and solidarity on ratings made of Anglo-Australian and Greek-Australian speakers . In an earlier study, Riches and Foddy (1989) showed that a Greek-accented speaker received lower ratings on competence-related dimensions (industrious, powerful, educated, intelligent, and competent) than did an Australian-accented speaker .' The first study extends this work to examine how fluency of speech combines with ethnic accent . The second study makes use of the now familiar "dyadic decisionmaking task" developed for tests of status characteristics theory (Berger, Fisek, Norman, and Zelditch 1977, chap . 5), to investigate how fluency and ethnicity jointly affect the acceptance of influence in a group task .



114

MARGARET FODDY and PHOEBE RICHES

An important feature of our research strategy was the measurement of perceivers' impressions of a speaker based on paralinguistic cues . We wanted to discover whether these impressions would differentially reflect task and categorical cues . More specifically, we tested whether these two different cues would have effects of different magnitude, and whether each would have an impact on different dimensions (perceived confidence versus diffuse status) . If this turns out to be the case, it provides a tool for testing the processes by which expectation for competence arise .

STUDY 1 : RATER EVALUATIONS OF RECORDED SPEECH Method In the first study four female speakers were presented to the subjects, three Greek-accented and one Anglo-Australian . The age range of the speakers was 21 to 28 years . A total of 89 first-year female psychology students (Anglo-Australians) listened to one of eight audio-tapes, and made ratings of the speaker (described below) . Subjects were tested in small laboratory classroom groups . Two of the Greek-accented speakers were actors and one was a native Greek speaker ; the Australian speaker was not an actor, and English was her native language. Each confederate read the same passage twice, once fluently and once in a stilted manner. The fluent passage was recorded with no pauses, pacing was maintained at a constant level, and no "ums" and "uhs" were present . The same passage was used for the stilted reading with pauses, and "ums" and "uhs" included at predetermined positions . 9 The experimenter introduced the study by explaining to the subjects that people often form impressions of others by listening to voices on the telephone or radio . Subjects were instructed to form an impression of the speaker by listening to a recording of her reading a passage out of a novel . They were then given a questionnaire which contained : (1) 14 six-point semantic differential adjective scales for rating impressions of the speaker (e .g ., intelligent, helpful ) ; (2) five six-point adjective scales for rating speech quality (loud, fast, natural, high-pitched, and fluent) ; (3) questions about the speaker's demographic background ; and (4) questions about the demographic background of the subject judging the tape . Most of the subjects were Anglo-Australian, and the ratings of those from Greek and Italian students were not included in the analysis . In terms of relative diffuse status characteristics, the listener-judges were equated with the Anglo-Australian speaker (same gender, same ethnicity), and higher status than the Greek-Australian speaker. No measures were made of the subjects' fluency . A further object of the first study was to identify one Anglo- and one Greek-accented speaker whose taped voices would be clearly identifiable as having one of these ethnic accents (the diffuse status characteristic), and whose stilted



Fluency and

Ethnic Accent as Cues to Competence in Task Groups

115

Table 1 . Means and (Standard Deviations) for Subjects' Ratings of Fluency for the Anglo - and Greek-Accented Speakers : Study 1 Versions of Speech Ethnic Accent of Speaker Anglo-Australian Greek-Australian Total

Fluent

Stilted

Total

4 .67 (0 .97) 3 .10(l .07) 3 .86(l .23)

2 .76(l .2) 2 .00 (0 .67) 2 .48(l .08)

3 .740-44) 2 .67(l .10)

and fluent readings of the passage would be perceived as discriminably different on this speech characteristic . Surprisingly, the native Greek speaker was judged least reliably as having a Greek accent . The Greek-accented actor whose tape produced most reliable identification of ethnicity was chosen for use in Study 2 . 10 It was hypothesized that there would be main effects for fluency and ethnic accent on ratings of competence and intelligence (hypotheses 1 and 2), and that the effect size for fluency would be greater than that for ethnicity (hypothesis 3) . Support for hypothesis 3 would also be gained if fluency influenced judgments of the speakers, but ethnic accent did not . 11 Results : Study 1 Manipulation Check : Ratings of Fluency

The fluency ratings for the Greek-accented speaker and the Anglo-Australian speaker are presented in Table 1 . A 2 x 2 ANOVA (ethnicity by fluency) produced a significant main effect for fluency of speaker (F=28 .23, d .f. 1, 62, p < .001), and a main effect for ethnicity (F = 21 .70, d.f. 1, 62, p < .001) ; there was no interaction . The effect size for fluency (d = 1 .5) was greater than for ethnicity (d = 1 .2), but both are reasonably large (Cohen 1977) . Although care was taken to ensure that the fluency of the Greek and Australian speakers was equal in the fluent conditions, and equally stilted in the stilted conditions (same number of pauses, same number of disfluencies, similar length, similar loudness, etc .) rated fluency reflected both speech qualities and ethnicity of the speakers . As we discuss below, expectations based on categorical cues may lead listeners to "hear" the speaker's fluency as consistent with these expectations . Ratings of Speaker Attributes

A factor analysis of the ratings of the two speakers in both the stilted and fluent versions was conducted in order to collapse the 19 variables (14 traits and five speech qualities) on the personality questionnaire . The aim was to examine



116

MARGARET FODDY and PHOEBE RICHES

'table 2. Variables Good Dependable Helpful Pleasant Natural Industrious Competent Active Powerful Confident Loud Fast

Factor Loadings of the Rating Scales onto Three Factors : Study 1 a Factor I

Factor 2

.792 .648 .494 .526 .433 .580 .712 .567 .365 - .383

- .385

.650

Fluent Big Educated Intelligent Note:

Factor

.410 .639

- .667 .308 -.750 - .749

a . Variables with loadings less than .3 are not presented .

whether factors would emerge which coincided with those found in investigations using the matched guise technique, namely, status and solidarity/affiliation dimensions, and whether the ethnic accent and fluency would affect these differently. Principal components analysis and varimax rotation revealed three factors with eigenvalues over two which accounted for 45 percent of the variance . As presented in Table 2, the first factor represents a cluster similar to that described previously as a solidarity/affiliation factor (good, dependable, helpful, pleasant, natural). The second factor may be argued to represent a confidence/ competence dimension (fast, industrious, powerful, confident, active, competent) ; and the third factor an intelligence/education factor . The term "competent" also loaded on the third factor, but to a lesser degree than intelligence and education . One may argue that "competent" is not all that different from "intelligent" in relation to "general task ability" . We comment on this in the general discussion . Of the speech dimensions, "fast" and "loud" loaded on the second factor, and fluency loaded on factor 3 . Factors 2 and 3 are both similar in many ways to the "status" factor of language attitude studies . A two-way MANOVA (ethnicity by fluency) on the three factors as dependent variables produced a significant effect for ethnicity of speaker on the intelligence dimension (F = 29 .806, d .f. 1, 62, p < 0 .001) . The Anglo-Australian speaker was rated significantly higher on the intelligence dimension (factor 3) than the Greek-Australian. The effect of fluency was found to be significant for both dimensions two and three (F= 8 .11, d .f., 1,62, p < 0.01) and (F = 39 .79, d .f. 1,62, p < 0.001) ; the fluent speaker was rated more positively than the stilted speaker on



Fluency and Ethnic Accent as Cues to Competence in Task Groups Table 3 .

117

Mean Ratingsa of Speakers' Attributes by Fluency and

Ethnicity of Speaker (standard . deviations in parentheses) : Study 1 Speaker Attribute Confident Competent Intelligent Educated Note:

a

Anglo-Fluent 3 .38(l .24) 4 .22 (1 .17) 4 .89 ( .583) 5 .38 ( .608)

Greek-Fluent 4 .4 4 .4 4 .0 4 .2

( .681) ( .821) ( .973) ( .894)

Anglo-Stilted 3 .00 3 .41 4 .12 4 .24

( .935) ( .795) ( .60) ( .903)

Greek-Stilted 2 .7 ( .949) 3 .2 ( .919) 4 .0043) 3 .2 (1 .033)

Rating scales were from 1-6, with higher ratings indicating higher levels of the attribute .

both . There were no significant interaction effects for any dimension . The subjects rated the Anglo-Australian speaker as intelligent 35 .5 percent more often than they rated the Greek-Australian speaker as intelligent . They rated the fluent versions of the speakers as intelligent 73 percent more often than the subjects rated the stilted versions as intelligent, the lower overlap indicating a stronger effect for fluency (the task cue) . The responses to the solidarity dimension were not influenced by either of the independent variables . Looking at the effect sizes for the separate attributes of confidence and competence, only fluency had a sizeable effect size (ds 1 .07 and 1 .08, respectively [Cohen 1977]) ; for ethnicity the effect sizes were close to zero (see Table 3 for means and standard deviations) . However, on the attributes intelligent and educated, both fluency and ethnicity had moderate to large effects (for intelligence, .49 for fluency, .65 for ethnicity; for education, 1 .24 for fluency, 1 .28 for ethnicity) . The estimates of the speakers' age and education level showed some indication that accent and fluency act as status cues . Both the Australian and Greek speaker were judged to be in the same age range as the subjects, although the Anglo-Australian was judged somewhat older than the Greek-Australian (26 and 21 years, respectively) . The speakers were in fact 25 and 26 years . A large number of the subjects estimated the Anglo-Australian speaker to have a university degree-17 for the fluent version and 13 for the stilted reading . The fluent Greek speaker was estimated to be a university student by only five of the subjects and a high school graduate by the majority (15) subjects . No subjects estimated the stilted Greek-Australian speaker to be a university student, she was believed to be a high school graduate by seven of subjects and primary level by three . These judgments are consistent with the loadings of education and intelligence on the third factor (educated/intelligent) . While we were not surprised to find that stilted reading produced judgments of lower education levels, it was surprising that this effect was larger for the Greek-Australian speaker . The results of the first study support the argument that a task cue, fluency, has a significant impact on perceptions of generalized task competence and confidence, and on perceptions of the more diffuse characteristic, education/intelli-



118

MARGARET FODDY and PHOEBE RICHES

gence ; the latter presumably also has a path of relevance to expectations for task performance . Ethnicity, a categorical status cue, influenced judgments of the more diffuse status characteristic (education) and a general judgment of ability (intelligence), but did not appear to influence directly judgments of competence and confidence in the situation . These results are interesting in their own right, for they show that paralinguistic task cues make salient status characteristics that attract differential evaluations . While we do not wish to make too much of the different patterns of effect for fluency and ethnicity, they do suggest that future research should include measures of impressions made by different status cues . This would allow an assessment of whether behavioral cues produce qualitatively and quantitatively different effects than do categorical cues (i .e., through different sorts of paths of relevance) . The findings from the first experiment also provided the materials needed to construct an experimental situation in which fluency and ethnicity could be varied independently, so that their relative effects on expectations and influence could be examined . We needed to identify paralinguistic cues which signaled category membership (ethnicity), and levels of task cues (fluency) ; the first experiment established that Greek-Australian accented speakers and Anglo-Australian speakers would be reliably identified as such, and that listeners could also distinguish between fluent and non-fluent speech cues . While the first experiment's results also provide evidence of differential evaluation of the speakers' status characteristics, it remains to be demonstrated that these evaluations translate into differences in influence in a task group .

STUDY 2 : RELATIVE IMPACT OF STATUS CUES ON INFLUENCE IN TASK GROUPS Overview In the second study naive Anglo-Australian and Greek-Australian female subjects participated in the standard expectation states influence task, in which they were led to believe they would be making decisions in a two-person group . In the course of performing the task, the subject found herself in disagreement with her (simulated) partner, and had to resolve the disagreements by staying with her own choice (rejecting influence), or changing to the partner's choice (accepting influence) . The design makes it possible to test the impact of ethnicity and fluency of partner's speech as cues to competence, by assessing their impact on acceptance of influence . Levels of ethnic accent (Anglo or Greek) and fluency (fluent, or disfluent) were introduced via overheard conversations between the experimenter and the bogus partner . Post-experimental impressions of the partner's attributes were obtained in the form of rated "personality impressions ." The experimental paradigm emphasizes the importance of the task, and encourages a collective ori-



Fluency and Ethnic Accent as Cues to Competence in Task Groups

119

entation ; both of these are important scope conditions of status characteristics theory (Berger et al . 1977) . Subjects Female university students who volunteered to participate in the experiment completed an information sheet . Included on the sheet were questions about the ethnic background of the subjects and languages spoken at home . Second generation Anglo- and Greek-Australians (n = 95) were initially selected to participate in the study . Subjects were categorized as second-generation Greek-Australian if at least one of their parents were born in Greece and Greek was spoken at home, and as Anglo-Australian if both their parents were born in Australia . Subjects were fluent speakers of English . The average age of the subjects was 21 .79 years. Of the initial pool of 95 subjects, 80 were retained for analysis, (40 Anglo-Australians and 40 Greek-Australians) . Of the 15 excluded, five expressed clear suspicion of the study, two failed to meet scope conditions, and eight failed to detect the accent of the speaker . Procedure Female participants worked on the task in same-sex dyads and were randomly assigned to one of eight conditions in a 2x2x2 factorial design (fluency of confederate, ethnicity of confederate and ethnicity of subject) . Subjects were told they would be working with a partner on a task that required them to judge complex patterns presented to them on a computer screen . They were told that the task involved working with a "partner" positioned in an adjacent room . Subjects were told that the "partner" was simultaneously completing the first stage of the study . The computer-controlled task 12 presented pairs of stimuli composed of black and green rectangles ; the subject's job was to decide which of the pair had more black in it, and to type their response ("left" or "right") . Subjects were told that successful performance of the task required "contrast sensitivity," an important and well-researched ability . In fact, the patterns were generated by a program to ensure that there were equal amounts of black and green displayed on the screen, in scattered blocks . Proportions of choices of "left" or "right" were virtually equal on practice trials, supporting the conclusion that the task was ambiguous . The study was presented to subjects as an investigation of workplace settingsone in which they worked alone and one in which they worked with a partner . The rationale given to subjects for including a "partner" was that people's ability to make the correct decision might be improved with the additional information provided by the "partner's" response . It was explained that responses would be scored on the basis of the number of correct choices made by the dyad, and so it was legitimate to seek and give advice during the second phase of the experiment . The notion was that subjects would change their answer to agree with the "part-



120

MARGARET FODDY and PHOEBE RICHES

ner's" response if they believed that the "partner" was more competent than themselves and more likely to give a correct response (see Berger et al . 1977 for details) . The subjects' willingness to defer to the partner thus reflects their expectations about their partners' ability ; the central dependent variable in the experiment is the proportion of times a subject rejects influence, that is, refuses to defer to the partner .

First Phase: (no partner) Subjects were given five practice trials, then completed 10 trials, without feedback about their performance . This was done to familiarize them with the task . Once the response was recorded the next set of patterns were presented . At the completion of the first phase the experimenter asked if there were any questions, then went to see if the "partner" was ready .

Second Phase: (partner) Manipulation of accent and fluency of partner . In the adjacent room the experimenter controlled an intercom and tape recorder so that subjects heard a prerecorded conversation between one of the confederates and the experimenter . The conversation involved a discussion on the malfunction of the computer and the "partner's" readiness to complete the second phase of the procedure . Efforts were made to avoid reference to status cues other than fluency and accent. Subjects were led to believe that they were eavesdropping on the conversation, as no mention was made by the experimenter of the conversation on her return to the testing room . The second phase task consisted of 20 patterns . After making each choice in phase two, the subject was told she would receive "feedback" from the "partner ." Once the subject received the "partner's" choice the computer asked for a second choice from the subject, after which the next set of patterns were presented. The subjects were not given information about the partner's second choice, and had to decide whether the partner was more or less likely to be correct when their answers disagreed . The crucial dependent variable is the P(s) or the proportion of critical trials on which the subjects resolved disagreement by staying with their own choices . On the assumption that higher rates of rejection of influence would reflect higher levels of perceived ability relative to the partner, it was hypothesized that subjects with fluent partners would reject influence less than subjects with less fluent partners ; that Anglo and Greek subjects would reject influence more from Greek than from Anglo partners, and that the effect of fluency would be stronger than the effect of ethnic accent . More specifically, we predicted the following patterns of means for P(s) :



Fluency and Ethnic Accent as Cues to Competence in Task Groups

121

Empirical Predictions

For Anglo-Australian and Greek-Australian subjects (focal persons), the order of P(s) from highest to lowest for the four types of partner is predicted to be : Greek-stilted > Anglo-stilted > Greek-fluent > Anglo-fluent If fluency completely overcomes the effect of ethnicity, the first and the last two means should be equal . A comment is needed on our reasons for predicting that Greek-Australian subjects would reject influence more from Greek than Anglo partners . The theory of status characteristics suggests that beliefs related to social categories, stereotypes, will be shared among members of the society, but it is not obvious that the Greek-Australian subjects would necessarily share beliefs about the competence of other Greeks relative to self (but see Ellemers and van Knippenberg 1997) . Further, these subjects were equated on ethnicity with the Greek-Australian subjects (so that only fluency should affect their expectations, just as with Anglo-Anglo pairs) . But, relative to Anglo-Australians, they possess the lower state of the characteristic, ethnicity . Thus, while it is possible that their rates of rejection of influence will be affected only by fluency and not accent for partners of the same ethnic group, ethnicity should be activated with the Anglo partner, and be reflected in lower P(s) values . Thus, the same ordering of means for P(s) are predicted for Anglo and Greek subjects . We did not measure the fluency of the naive subjects ; it was assumed that they were more fluent than the stilted "partner," and equally or less fluent than the "fluent" partner . At the conclusion of the task, the experimenter interviewed each subject, following a protocol of questions concerning task and collective orientation, credibility of procedures, and level of interest . A post-task questionnaire was also completed to check whether subjects met the scope conditions and whether the accent and fluency variables had been activated . Subjects completed a semantic differential type adjective scale of 11 pairs to check their impressions of the "partner" ; these included all but the speech and physical characteristics ratings used in the pre-test . 13 Results : Study 2 Manipulation Checks

Manipulation checks were conducted prior to the analysis of the dependent variables to ensure that the scope conditions of the theory had been established for an adequate test of the hypothesis . Detection of accent . All subjects paired with the Anglo-Australian confederate correctly recognized her as Anglo-Australian . Nine of the 20 Anglo-Austra-



122

MARGARET FODDY and PHOEBE RICHES

lian subjects paired with the Greek-Australian confederate classified the speaker as Greek, while seven recognized her as Southern European and four believed her to be Italian . The Greek-Australian subjects classified the Greek-Australian "partner" as either being Greek (14 out of 20), or Southern European (six) . Because the evaluation of Southern Europeans tends to be similar (McAllister and Moore 1991), all of these subjects were retained for analysis . Detection of fluency . Subjects were asked for a free description of the "partner's" speech on the post-task questionnaire . It was expected that subjects would recognize the paralinguistic cue and classify it with adjectives commonly used to describe fluent and stilted speech . Eight of the 40 subjects classified the fluent speaker as confident, 15 classified her as fluent, 10 regarded her as speaking normally, and seven classified her speech as "proper ." Of the 40 subjects whose partner was stilted, 15 classified the speakers as hesitant, 16 as slow, and nine as timid . Rejection of influence (P(s)) A three-way ANOVA was conducted to investigate the differences among the means . The only significant difference was found for the comparison of the fluent and hesitant versions of the speech (F = 7 .45, p < 0 .01, d .f. 2, 72) . As indicated by the means, the level of rejection of influence was greater for the stilted versions than for the fluent versions . The effect of ethnicity of partner's accent was not significant (F = 2.12, p = 0 .15, in the direction of greater rejection of influence from Greek partners) . There was no interaction effect for subject's ethnicity and partner's ethnicity . While the interaction between fluency and ethnicity of subject was not significant (p = 0 .10), a trend was apparent : Greek-Australian subjects were more influenced by the "partner's" fluency than were Anglo-Australian subjects . The highest rate of rejection of influence was found in the condition where the Anglo subject's "partner" was a Greek-accented stilted speaker (see Table 3) ; the lowest P(s) occurred in the condition where the Greek-Australian subject had a fluent Anglo-Australian partner . These means are consistent with predictions, but otherwise the predicted order of means was not found for the Anglo subjects : levels of P(s) were similar for all cells except when the partner had a stilted Greek-Australian accent . In the other conditions, the levels of P(s) were consistent with the subject having formed equal, or slightly more positive expectations for self than for other . While fluency might "save" the Greek-fluent partner from the effects of ethnicity, the similar level of P(s) for the stilted Anglo partner does not support the claim that fluency will have a greater impact on expectations than ethnicity . For Greek-Australian subjects, the pattern of results was somewhat different, although not sufficiently different to have produced any significant interaction



Fluency and Ethnic Accent as Cues to Competence in Task Groups

1 23

Table 4 . Mean P(s), Means and Standard Deviations of Stay Responses, by Ethnicity of Subjects, and Ethnicity and Fluency of Partners : Study 2 Partner's Ethnic Background Anglo

Anglo subjects Fluent partner Stilted partner Greek subjects Fluent partner Stilted partner

Greek

P(s)a

M

s.d.

.68 .69

10 .3 10 .4

2 .06 2 .12

.55 .77

8 .3 11 .5

3 .71 2 .12

P(s)

M

s .d .

.71 .78

10 .6 11 .7

2 .59 2 .41

.66 .77

9 .9 11 .5

2 .56 1 .51

Note : P(s) is the proportion of stay responses over 15 critical (disagreement) trials .

effects with ethnicity of subject . The order of means of the subject's P(s) with the four partners was : Greek-stilted = Anglo-stilted > Greek-fluent > Anglo-fluent . Here, fluency of partner seems to have been a more important factor ; the Greek subjects rejected influence from both Anglo and Greek partners when they were hesitant ; the levels of P(s) were close to those of the Anglo subjects paired with a stilted Greek partner. The lowest rates of rejection of influence were, as predicted, when the partner was both Anglo and fluent (P(s) = .55) . Since we did not measure the fluency of the subjects' speech, we are not able to establish the extent to which our results depended on relative expectations formed on the basis of this cue . Personality impressions . Analysis of the series of personality impressions of the "partner" by the subjects in experiment 2 was conducted to investigate the consistency between performance outputs and perceptions . A MANOVA (ethnicity of partner, ethnicity of subject, and fluency) was conducted on the 11 variables, and the multivariate effect was significant for fluency (F = 2 .03, d .f. 11, 62, p < .05) and ethnicity (F = 2 .80, d .f. 11, 62, p < .01) . There were no significant interaction effects . Examining the univariate effects, a main effect for fluency was found only on the confidence variable (F = 6 .61, d .f. 1,72, p < 0.05) . The fluent speaker was rated as being more confident than the stilted speaker (fluent : M = 4 .2, stilted : M = 3 .5) . The univariate effect for partner's ethnicity was significant on the following ratings (in all cases, d .f . = 1, 72) : educated (F = 17 .8, p < .0001), quick (F = 9 .2, p < .01), intelligent (F = 5 .6, p < .02) and confident (F = 8 .6, p < .01) . The Anglo-Australian partner was rated significantly higher than the Greek-Australian partner . (The ratings were generally less variable than in Study 1, and sub-



124

MARGARET FODDY and PHOEBE RICHES

jects tended to check between three and four on the six-point scales) . Effect sizes (ds) ranged from .64 for "confident," to .91 for "educated." Given the very minimal exposure that subjects had to their partners' voice over the intercom, the effects are surprisingly large . The post-experimental rating results, on which ethnicity had a more pervasive influence than fluency, seem at first to be inconsistent with the other results which suggest that fluency is a much stronger cue to competence (but see Tuzlak 1988 ; Sev'er 1989) . However, if task cues such as fluency have their effect through their immediacy, that is, by cueing competence in the immediate task setting, then there is no reason to expect this effect to last or to generalize . The naive subject heard only two short verbal interchanges in which the ethnicity and fluency of the confederate were conveyed, and it is possible that the impression of fluency would dissipate more quickly than would the ethnic identification of the partner . Tuzlak (1988) has provided a similar explanation for the relative decay of demeanor effects over time, while the effects of a diffuse status characteristic (race, in her study), increased . Such an interpretation would suggest that while high task cues will be more salient determinants of influence in task-oriented interaction, the experience of having one's category-based stereotypes contradicted will not necessarily produce a lasting effect on those stereotypes . Nevertheless, one should not lose sight of the result, that despite the greater number of effects of ethnicity on post-task ratings relative to fluency, it was still the case that fluency alone had a significant effect on rates of influence . General Discussion

We set out to provide an assessment of the claim that task cues would have a larger impact on evaluations, expectations, and influence than categorical cues to status, and more generally, to investigate the status characteristics theory interpretation of status cues . We offered possible interpretations of the claim that task cues are more "immediate" than categorical cues, and provided suggestions for graph representations of these interpretations . These graphs reflect the argument that the path of relevance from task cues to task outcomes may not only be shorter than that for categorical cues, but there may also be more paths . We adopted an empirical strategy to directly measure whether the two cues appear to "activate" different impressions of the speaker, as would be expected by the definitions of task and categorical cues (or elements, in the graph) . Having obtained some supportive evidence for this latter claim from listeners who judged the impressions of speakers who differed in ethnicity and fluency, we used an experimental design to assess whether the two types of cue could predict levels of influence accepted (a behavioral outcome) . Here we found that only fluency of speaker had a systematic effect, and ratings of the partners after the task showed that fluency affected perceived confidence of the partner . While ethnicity appeared to influence post-ses-



Fluency and Ethnic Accent as Cues to Competence in Task Groups

1 25

sion ratings on several attributes, these had not translated into sufficiently differential expectations to have a systematic effect on P(s) . While we are not the first to use rated impressions of the partner in empirical tests of status characteristics theory (Tuzlak 1989 ; Rashotte and Smith-Lovin 1997), these have not been used in previous work to investigate the process of inference from status cues to the emergence of status hierarchies in groups . One empirical study is not enough to establish that these direct measures will bear a stable and consistent relationship to status characteristics, but the method is promising and can be employed in the study of a range of behaviors thought to constitute behavior elements (Fisek et al . 1991), or behavioral enactments (Rashotte and Smith-Lovin 1997) . It is important to note that we are not claiming that people can report directly on the process of generalization from status characteristics to beliefs about task ability and task outcome, but rather, that we can measure their impressions of others' characteristics which the theorist may then employ in explanations of status emergence . This raises a further theoretical question concerning status cues themselves . Status cues are not status characteristics-they are visible and audible indicators of status, which operate by activating relationships of possession and relevance . While the status characteristics theoretical formulation of the operation of nonverbal cues provides a useful framework for understanding their effect on impression formation and interaction, there is much to be learned about how they exert this effect . Fisek, Berger, and Norman (1991), and Skvoretz and Fararo (1996) provide possible representations of how behavior patterns may induce differential expectation states, through links to abstract ability . Our research suggests that these links may be more complex-task cues such as fluency may trigger beliefs associated with diffuse and specific status characteristics, as well as abstract task ability . Categorical cues may activate expectations for competence, and thus biased perceptions of performance and evaluations of task cues . To date, status cues themselves are not represented in the graph representations of the theory ; in this paper we have shown them as "behavioral cues," but much work needs to be done to establish just how cues activate status characteristics . In the second study fluency had a greater impact on acceptance of influence (P(s)) than did ethnic accent, but the ratings of the partner after the experimental session suggest that the effect of ethnicity will not invariably be weaker. According to the status characteristics theory interpretation of nonverbal and paralinguistic cues, diffuse status characteristics "govern" task cues, making it difficult if not impossible for lower status participants to display confident cues . It might thus be argued that the combinations of fluency and ethnic status in this experiment are unnatural, since it is unlikely that a person with Greek-accented English would speak more fluently than would an Anglo-Australian . However, such "unnatural" combinations do occur, and the reactions of subjects to them shed light on the relative importance given to, task (fluency) and categorical (ethnic accent) cues . It may well be that a perceiver assumes consistency between specific (fluency-cued)



126

MARGARET FODDY and PHOEBE RICHES

and diffuse (ethnicity) status characteristics, but inconsistency is also interpretable . In the case studied in this experiment, task cues inconsistent with ethnicity seemed to be interpreted as evidence that the speaker was competent . It is important to note that there was, however, no evidence for an "augmentation" effect, as proposed by attribution theorists (Jones and Davis 1965 ; Hansen and O'Leary 1985) for cases in which enhanced ability is attributed to the member of a category who is expected to do poorly, but in fact performs well (see also Biernat and Kobrynowicz 1997) . If anything, the Greek-Australians were judged by a stricter standard for ability (Foschi and Foddy 1988 ; Foddy and Smithson 1989) when they spoke fluently (high performance), and a more lenient standard for lack of ability when they spoke in a stilted fashion (low performance) . This interpretation is consistent with the ratings given in the first study ; although the tapes of the fluent Greek-Australian were equivalent to those of the Anglo-Australian on length, number of pauses, number of disfluencies (that is, fluency), the Greek-Australian was judged to be less fluent . This only makes sense if the Greek-Australian was being judged by a stricter standard for fluency . It would seem necessary to make the Greek-accented speaker much more fluent on objective measures, before she was judged to be equally fluent, in comparison to the Anglo-Australian . This points to the need for further definition of the concept of fluency ; ethnic accent may be part of perceivers' understanding of what it is to be fluent . While a central aim of the current study was to assess the relative impact of a task and categorical status cues on expectations for task competence and related influence processes, the generality of the finding of greater impact for the task cue can only be established by empirical research using a variety of task and status cues . As Nemeth (1983) has noted, decisive tests will be difficult to construct, because of the difficulty of equating strength of the task and categorical cue stimuli . More important, it may not always be the case that perceivers attend more to "immediate" cues ; in some circumstances, categorical cues may provide better prediction of task competence, or may be believed to provide better prediction . For example, negative stereotypes about Aboriginal Australians are often so pronounced that a high level of task cues might well have minimal impact on beliefs about the actor's task competence . When a diffuse status characteristic has multiple short paths of relevance to task competence, it may well be a more important determinant of expectations . The theory of status characteristics makes explicit provision for the variability in beliefs about and evaluation of the characteristics of the members of different social categories but does not at present incorporate the concept of magnitude of different status characteristics which may operate in a continuous rather than a discrete manner (Foddy and Smithson 1996, 1999) . There is evidence from language attitudes research which is also relevant . Brennan and Brennan (1981) showed that the degree of accentedness of the speech of Mexican Americans predicted the extremity of status ratings . Ryan, Carranza, and Moffie (1977) found that Spanish-American accents were rated negatively, and the more extreme the accent, the more negative the stereotype . These findings



Fluency and Ethnic Accent as Cues to Competence in Task Groups

127

suggest that the magnitude of status differences suggested by different status cues needs to be taken into account in predicting their net impact . Elaboration of the theory to allow for "degree of difference" on status characteristic may provide an explanation of those cases in which categorical cues outweigh more immediate task cues . In addition to such an elaboration, recent developments in status characteristics theory provide potential accounts of the conditions under which task cues might not be as strong as expected . Ridgeway and Berger (1988) have suggested that the interpretation of task cues may be qualitatively different for high and low status actors, especially in legitimated status orders (i .e., the same task cue may indicate that an actor is competent or "uppity") . Similarly, Foddy and Smithson (1989) argued that the same nonverbal behaviors may be evaluated against different standards, depending on the status of the actor (p . 85 .) These possibilities merit exploration . As this study, and a previous one (Riches and Foddy 1989) have shown, our understanding of nonverbal behavior will be enhanced by exploiting a systematic theoretical framework that provides a link between the perception of these cues, and subsequent behavior .

ACKNOLWEDGMENTS We thank Ben Ong for programming the task and for statistical advice, and Joseph Berger for comments on an earlier version of the manuscript . The comments of two anonymous reviewers were also valuable .

NOTES 1 . Following De Paulo and Friedman (1998, p. 3), we define "nonverbal" as "not words ." They note that vocal cues such as voice tones and inflections are nonverbal in that they are not words, and include them with facial expression, touch, gaze, and gestures as "nonverbal communication ." 2. The term "affiliation" will be used here to refer to this dimension, although it is more frequently labeled "solidarity" in studies of speech perception. 3 . The scope of the theory is explicitly limited to groups in which members want to solve or perform well on a task requiring an important ability, but for which there is no objectively correct answer . While this limits the range of the theory, there are many settings in which the scope conditions are met, particularly those where attributions of competence must be made. 4 . Expectations are always held relative to others in the group . 5 . They also distinguished expressive and indicative cues ; this distinction will not be discussed in this paper. We deal only with two sorts of expressive cues . 6 . Fisek and colleagues (1991) argue that behavior patterns activate "status typification states," or "abstract conceptions of typical high-status and low-status behaviors" (p . 1373) ; these then become relevant to like-signed abstract task ability . For our purposes, we do not need to include the additional step of status typification states . 7 . Several studies have shown that unfavorable evaluations are given to the speech patterns of regional speakers, ethnic group members, and lower-class populations in terms of status and prestige, from judges who may or may not be standard speakers themselves (Ryan and Sebastian 1985 ; Callan, Gallois, and Forbes 1983) .



128

MARGARET FODDY and PHOEBE RICHES

8 . In Melbourne, as in other parts of Australia, Greeks form a sizeable minority group, many of them first-generation migrants to the country. Studies of social distance indicate that Anglo-Australians evaluate Greeks below people from Northern Europe and North America, but higher than a number of other groups from Southeast Asia and the Pacific . 9 . The duration of the fluent tape was two minutes 20 seconds, and the stilted tape was slightly longer, due to the insertion of pauses and "umm's" and "uh's" (two minutes 45 seconds (Australian accent), and three minutes 10 seconds (for the longest Greek-accented passage) . Subjects responded to only one version of the passage. 10. For the fluent version of the Greek accented speaker, there were 24 raters ; 13 judged her to be Greek, three Italian, four Mediterranean, and four Yugoslavian . For the stilted version there were 12 raters ; five judged her to be Greek, two Italian, three Mediterranean, and two Indian . The degree of inaccuracy was not of concern since previous research had shown that perceptions of Greeks, Italians, and Southern Europeans were similar with respect to competence. All raters judged the Anglo-Australian speaker's ethnicity accurately in both the stilted and fluent versions . 11 . It is always possible that the particular operationalizations of task and categorical cues maybe relatively more or less successful, and that any differences in impact may reflect such differences . While we endeavored to make both cues equally salient, we recognize that this calibration problem is a persistent one. 12. Barry Markovsky provided the first version of the task which was then modified by Ben Ong to suit the experiment. A copy of the modified program is available from the first author . 13 . The rating dimensions were: industrious, educated, powerful, pleasant, competent, interested, quick, considerate, intelligent, helpful, and confident .

REFERENCES Allport, G . W., and H . Cantrill . 1934 . "Judging Personality from Voice ." Journal of Social Psychology 5 : 37-55 . Ball, P., J. Byrne, H . Giles, P . Berechree, J. Griffiths, H . Macdonald, and I . McKendrick. 1982 . "The Retrospective Speech Halo Effect : Some Australian Data." Language and Communication 2 : 277-284. Berger, J ., M . H . Fisek, R . Z. Norman, and M . Zelditch. 1977 . Status Characteristics and Social Interaction. An Expectation States Approach . New York: Elsevier . Berger, J., S . J . Rosenholtz, and M . Zelditch . 1980 . "Status Organising Processes." Annual Review of Sociology 6 : 479-508 . Berger, J., M . Webster, C . Ridgeway, S . J. Rosenholtz. 1986 . "Status Cues, Expectations and Behaviors." Pp. 1-22 in Advances in Group Processes, Vol . 3, edited by E . Lawler . Greenwich, CT : JAI Press . Brennan, E. M ., and J . S . Brennan . 1981 . "Accent Scaling and Language Attitudes : Reactions to Mexican American English Speech." Language and Speech 24 : 207-221 . Brown, B . L., H. Giles, and J. N . Thakerar . 1985 . "Speaker Evaluations as a Function of Speech Rate, Accent and Context." Language and Communication 5 : 207-220. Callan, V., C . Gallois, and P. A . Forbes . 1983 . "Evaluative Reactions to Accented Speech : Ethnicity, Sex Role and Context ." Journal of Cross-Cultural Psychology 14 : 407-426 . Cargile, A . C ., H . Giles, E . B . Ryan, and J . J . Bradac . 1994 . "Language Attitudes as a Social Process : a Conceptual Model and New Directions ." Language and Communication 14 : 211-236 . Cohen, J . 1977 . Statistical Power Analysis for the Behavioral Sciences . New York : Academic Press. Depaulo, B . M ., and H. S . Friedman . 1998 . "Nonverbal communication ." Pp . 3-41 in The Handbook of Social Psychology (4th ed .), edited by D. T . Gilbert, S . T. Fiske, and G. Lindzey . Boston : McGraw-Hill .



Fluency and

Ethnic Accent as Cues to Competence in Task Groups

129

Dovidio, J. F ., C . E . Brown, K. Heltman, S . L . Ellyson, and C . F . Keating . 1988 . "Power Displays Between Women and Men in Discussions of Gender-linked Tasks : a Multichannel Study ." Journal of Personality and Social Psychology 55 : 580-587 . Driskell, J. E., B . Olmstead, and E. Salas . 1993 . "Task Cues, Dominance Cues, and Influence in Task Groups ." Journal of Applied Psychology 78 : 51-60. Ellemers, N., and A . van Knippenberg . 1997 . "Stereotyping in Social Context ." Pp. 208-235 in The Social Psychology of Stereotyping and Group Life, edited by R . Spears, P . J . Oakes, N . Ellemers, and S . A . Haslam. Oxford : Blackwell . Fisek, H., J. Berger, and R. Norman. 1991 . "Participation in Heterogeneous and Homogeneous Groups : a Theoretical Integration ." American Journal of Sociology 97 : 114-142. Foddy, M ., and M . Smithson . 1989 . "Fuzzy Sets and Double Standards : Modelling the Process of Ability Inference ." In Sociological Theories in Progress : New Formulations, edited by J . Berger, M . Zelditch, and B . Anderson . Newbury Park, CA : Sage. Foddy, M ., and M. Smithson . 1996 . "Relative Ability, Paths of Relevance and Influence in Task-oriented Groups ." Social Psychology Quarterly 59: 140-153 . Foddy, M ., and M . Smithson . 1999 . "Can Gender Inequalities Be Eliminated?" Social Psychology Quarterly 62 : 302-324 . Foschi, M ., and M . Foddy. 1988 . "Standards, Performances and the Formation of Self-other Expectations." Pp . 248-260 in Status Generalization : New Theory and Research, edited by M . Webster and M . Foschi . Stanford, CA : Stanford University Press . Foschi, M ., and J. Tagaki . 1991 . "Ethnicity, Task Outcome and Attributeions : A Theoretical Review and Assessment ." Pp. 177-203 in Advances in Group Processes, Vol . 8, edited by B . Markovsky and E. Lawler. Greenwich, CT : JAI Press. Giles, H., and N . Coupland. 1991 . Language : Contexts and Consequences . Milton Keynes : Open University Press . Giles, H ., A. Mulac, J . J . Bradac, and P . Johnson. 1986 . Communication Yearbook 10 : 13-48 . Giles, H ., and R . N . St . Clair . 1985 . Recent Advances in Language, Communication, and Social Psychology . London : Lawrence Erlbaum . Giles, H., and C . Sassoon. 1983 . "The Effect of Speakers Accent, Social Class Background, and Message Style on British Listeners' Social Judgments ." Language and Communication 3 : 305-313 . Hansen, R . D ., and V . E . O'Leary . 1985 . "Sex-determined Attributions ." In Women, Gender and Social Psychology, edited by V . E . O'Leary, R . K . Unger, and B . S . Wallston . Hillsdale, NJ : Lawrence Erlbaum . Humphreys, P., and J . Berger. 1981 . "Theoretical Consequences of the Status Characteristics Formulation ." American Journal of Sociology 86 : 953-983 . Jones, E . E ., and K . E . Davis . 1965 . "From Acts to Dispositions : The Attribution Process in Person Perception ." In Advances in Experimental Social Psychology, Vol . 2, edited by L . Berkowitz. New York: Academic Press . Lambert, W . E. 1967 . "The Social Psychology of Bilingualism ." Journal of Social Issues 23 : 91-109 . Lambert, W . E ., R . C . Hodgson, R. C . Gardner, and S . Fillenbaum . 1960. "Evaluational Reactions to Spoken Languages ." Journal of Abnormal and Social Psychology 60 : 44-51 . McAllister, I., and R. Moore . 1991 . "Social Distance among Australian Ethnic Groups ." Social Science Research 75 : 95-100. Nemeth, C . J . 1983 . "Reflections on the Dialogue Between Status and Style : Influence Processes of Social Control and Social Change." Social Psychology Quarterly 46 : 70-74 . Nesdale, D ., and R . Rooney . 1996 . "Evaluations and Stereotyping of Accented Speakers by Pre-adolescent Children ." Journal of Language and Social Psychology 15 : 133-154 . Norton-Ford, J . L., and D. R. Hogan . 1980. "Role of Nonverbal Behaviors in Social Judgments of Peers' Assertiveness ." Psychological Reports 46 : 1085-1086 . Pear, T . H . 1931 . Voice and Personality . London : Wiley .



130

MARGARET FODDY and PHOEBE RICHES

Pryor, B ., and R. W . Buchanan. 1984 . "The Effects of a Defendant's Demeanor on Juror Perceptions of Credibility and Guilt." Journal of Communication 34 : 92-99 . Rashotte, L . S ., and L. Smith-Lovin . 1997 . "Who Benefits from Being Bold? The Interactive Effects of Task Cues and Status Characteristics on Influence in Mock Jury Groups." Pp . 235-255 in Advances in Group Processes, Vol . 14, edited by E . Lawler. Greenwich, CT : JAI Press. Riches, P., and M . Foddy . 1989 . "Ethnic Accent as a Status Cue." Social Psychology Quarterly 52: 197-206. Ridgeway, C . L. 1987 . "Nonverbal Behavior, Dominance, and the Basis of Status in Task Groups ." American Sociological Review 52: 683-694. Ridgeway, C . L ., and J. Berger. 1988 . "The Legitimation of Power and Prestige Orders in Task Groups ." Pp . 207-231 in Status Generalization : New Theory and Research, edited by M. Webster and M . Foschi . Stanford, CA : Stanford University Press . Ridgeway, C . L ., J . Berger, and L. Smith . 1985 . "Nonverbal Cues and Status : An Expectation States Approach ." American Journal of Sociology 90 : 955-978 . Ryan, E . B ., and M . A. Carranza . 1975 . "Evaluative Reactions of Adolescents Toward Speakers of Standard English and Mexican American Accented English ." Journal of Personality and Social Psychology 31 : 855-863 . Ryan, E . B ., M. A . Carranza, and R. W. Moffie . 1977 . "Reactions Towards Varying Degrees of Accentedness in the Speech of Spanish-English Bilinguals ." Language and Speech 20 : 267-273 . Sebastian, R . J ., and E. B . Ryan . 1985 . "Speech Cues and Social Evaluation : Markers of Ethnicity, Social Class, and Age ." Pp . 112-143 in Recent Advances in Language, Communication, and Social Psychology, edited by H . Giles and R . N . St. Clair. London : Lawrence Erlbaum . Sereno, K. K., and G . J . Hawkins . 1967. "The Effects of Variations in Speakers' Nonfluency upon Audience Ratings of Attitude Toward the Speech Topic and Speakers' Credibility ." Speech Monographs. 34 : 58-64. Sev'er, A . 1989 . "Simultaneous Effects of Status and and Task Cues : Combining, Eliminating, or Buffering?" Social Psychology Quarterly 52: 327-335 . Sev'er, A . 1991 . "Women, Men, and Dominance in Small Groups : A Social Roles Assessment ." Canadian Journal of Sociology 16 : 265-280 . Skvoretz, J ., and T . J. Fararo. 1996 . "Status and Participation in Task Groups : A Dynamic Network Model." American Journal of Sociology 101 : 1366-1414 . Smith, B. L., B . L. Brown, W. J . Strong, and A. C . Rencher. 1975 . "Effects of Speech Rate on Personality Perception ." Language and Speech 18 : 145-152. Tuzlak, A . 1988. "Boomerang Effects : Status and Demeanor over Time ." Pp. 261-274 in Status Generalization : New Theory and Research, edited by M . Webster and M . Foschi . Stanford : Stanford University Press. Tuzlak, A . S. 1989 . "Joint Effects of Race and Confidence on Perceptions and Influence : Implications for Blacks in Decision-making Positions ." Canadian Ethnic Studies 21 : 120-126 . Tuzlak, A ., and J . C . Moore . 1984. "Status, Demeanor and Influence : an Empirical Reassessment ." Social Psychology Quarterly 47 :178-183 .

STRATEGIC DECISION MAKING

Frans N . Stokman, Marcel A .L.M . van Assen, Jelle van der Knoop, and Reinier C .H . van Oosten

ABSTRACT This paper introduces a methodology for strategic intervention in collective decision making . The methodology is based on (1) a decomposition of the problem into a few main controversial issues, (2) systematic interviews of subject area specialists to obtain a specification of the decision setting, consisting of a list of stakeholders with their capabilities, positions, and salience on each of the issues ; (3) computer simulation . The computer simulation models incorporate only the main processes through which differences in positions and salience are accommodated in binding decisions : management of meaning through the provision of convincing information, challenges, and exchanges . The methodology generates insights into the likely outcomes of the process, the amount of conflict involved, and the stability of the outcomes . These insights and the investigation of the effects of strategic moves provide major strategic advantages to the user . This is likely to lead to a better representation of the user's own position in the decision outcome and the creation of a broader political and social support behind the decision outcome .

Advances in Group Processes, Volume 17, pages 131-153 . Copyright m 2000 by JAI Press Inc . All rights of reproduction in any form reserved . ISBN : 0-7623-0651-3 131



132

FRANS N. STOKMAN ET AL . INTRODUCTION'

Collective decision making is one of the most fundamental processes in society . People's long-term common interests often require binding decisions for their successful accomplishment . Whenever all people strive toward the same outcome there is no problem . They can simply turn the preferred outcome into a decision and act accordingly . Collective decision making becomes difficult when stakeholders take different positions and express different preferences with respect to the outcome . Then, the different positions have to be accommodated in one way or the other . The dynamics in the decision-making process result from the fact that each stakeholder, with different intensity and potential, attempts to realize his 2 position whereas only one outcome can be chosen . Small informal groups can often make decisions on the basis of informal rules and agreements . As soon as common interests become more complex and groups larger, collective decision making has to be institutionalized through the elaboration of generally accepted decision-making procedures . Such formal procedures can be found at all levels of society . They specify, among other things, which bodies have the authority to make the final decisions, how these bodies are composed and their members selected, and which stakeholders at what moment have to be heard or are otherwise involved in the decision making . In addition, rules specify procedures for appeal against decisions that are harmful or not made according the required procedures . In the social sciences, models for collective decision making have been developed since the early eighties that provide far-reaching insights into collective decision-making situations (among others, Bueno de Mesquita et al . 1985 ; Coleman 1990 ; Laumann et al . 1987 ; Stokman and Van den Bos 1992 ; Stokman and Van Oosten 1994 ; Torenvlied 1996 ; Konig 1997 ; Pappi and Henning 1998) . Game theoretical insights are combined with mathematical models and computer simulation to model essential collective decision-making processes . The models represent three processes that precede formal decision making and that affect the final positions of the stakeholders involved . First, position changes are sometimes . induced because stakeholders receive or provide new convincing information rather to take other positions (management of meaning) . Second, in other situations stakeholders feel challenged by others to change their positions because the losses incurred by losing such a challenge are larger than the losses incurred by defending their position . Finally, stakeholders may change their positions when they see possibilities to create win-win situations through exchanging (logrolling) . All models require a limited set of data obtained through interviews with subject area specialists . The models generate information on the outcome, the type of process, and the amount of conflict to be expected . The aim of the models is twofold . First, the models are used for the analysis of decision-making processes and for the prediction of decision outcomes without intervention . On the basis of a



Strategic Decision Making

133

large number of evaluation studies of mostly political decisions at the international, national, and local level (Bueno de Mesquita and Stokman 1994 ; Ray and Russett 1996 ; Stokman and Berveling 1998 ; Rojer 1999 ; Payne 1999), it can be concluded that the models are quite accurate . 3 Second, the models are particularly useful for intervention in decision making in order to optimize the utility of one stakeholder, to create sufficient support for a decision, or to arrive at a fair decision on the basis of mediation . These strategically oriented applications are mainly applications for large organizations and governmental agencies . They prove the usefulness of the models for obtaining direct strategic insights and for intervention in ongoing decision-making processes . The models depart from a systematic analysis of the decision-making situation which is useful in itself. Such a systematic approach involves at first a decomposition of the problem resulting in the specification of a limited number of main controversial issues to be solved . Subsequently, an analysis of all stakeholders is required, including an overview of their positions, capabilities, and salience on each of the issues . Next, different strategies have to be considered. Computer simulation models are particularly useful for evaluating these strategies . We will elaborate each of these elements .

THE DECOMPOSITION OF THE PROBLEM INTO CONTROVERSIAL ISSUES The first essential step in a strategic decision-making analysis is the specification of the problem at stake in terms of a limited number of issues on which decisions have to be made . Each issue can be seen as a major controversial point for decision . Decisions of the stakeholders on the specified set of issues should determine the contours of the chosen solution . Usually, one to five issues are sufficient, even in complicated situations, but incidentally up to 20 issues have been used in some applications . The requirement of the specification of a limited number of issues is in its own right a good exercise to distinguish main from subordinate points . Ill-defined specification at this stage implies a possible solution to something that has little impact on the actual problem . Misspecification of the issues is the main cause of the failures in model predictions . An important aid in the specification of issues is the requirement that at least the two extreme positions on each issue should be identified . Identification of the two extreme positions actually implies that an issue can be seen as unidimensional, a straight line on which different positions can be located . 4 Intermediate positions indicate more moderate positions, but also possible compromise outcomes . If extreme and possible intermediate positions cannot be ordered, the specification of the issues is wrong, for example, because more underlying issues are involved . In such cases it is recommended to go one step deeper and to identify the underlying dimensions .



134

FRANS N. STOKMAN ET AL .

If we, for example, aim to analyze the merger of two companies, typical issues are the premium to be paid for the other company above the market or shares value, the desired degree of integration of the two companies, the initial number of representatives of each company in the integrated board, the timing of the merger, and which of the two presidents will be nominated as the president of the merged company . The last issue is an example of a dichotomous issue without intermediate outcomes . Often, such issues are most difficult to solve, as no compromises are possible . Whenever possible, it is therefore advisable to integrate such issues with other issues, facilitating, for example, a solution through an exchange with other issues (e .g., the initial number of representatives in the board) . A decomposition of the problem also implies at least a provisional specification of the main stakeholders . Without such a specification it will be difficult to specify extreme and intermediate positions . Stakeholders can be individuals or organizations . Individuals usually represent organizations or hold important positions in organizations .

CAPABILITIES, POSITIONS, AND SALIENCE After specifying the issues, the stakeholders, and the different possible positions on the issues, the next step in a strategic decision analysis consists of the specification of three characteristics of each stakeholder in relation to each issue . These data are usually obtained in interviews with one or two experts who make estimates relating to all stakeholders . The experts should have a good overview of the whole decision-making setting . Numerical specifications are required in the models . An example of such a numerical specification is given in Table 1 . The example originates in a project carried out by a transition manager from the Dutch company W&S Transition and Interim Management . W&S was asked to reorganize a company in the Amsterdam harbor that was facing serious economic problems . The company, given the acronym ABC, had a strong ideological orientation and had to accept more projects on economic grounds in order to survive . Maintaining the present ideological orientation was highly controversial . Some stakeholders wanted it to be reduced, whereas others wanted its continuation or even strengthening . We specified the central issue of the ideological orientation as the percentage of projects that the company would be allowed to accept on ideological grounds after the reorganization . The second column of Table 1 gives the positions of the stakeholders on the issue . The range is between 0 percent for competitors to 100 percent for the trade unions . The transition manager, the chairman of the supervisory board, and the economically oriented faction in the board took the rather radical position of 10 percent . Several other intermediate positions were also taken .



Strategic Decision

Making

135

Table 1 . Example of a Specification of an Issue and the Capabilities, Positions and Saliences of the Stakeholders Issue : Ideological Orientation Capabilities

Position

Salience

0 .15 0 .15 0 .30 0 .30 0 .15 0 .15 0 .30 0 .20 0 .70 0 .40

70 10 40 80 80 90 30 70 100

0 .2 0 .3 0 .3 0 .4 0 .2 0 .3 0 .3 0 .4 0 .1 0 .1 0 .4

Ideological faction in the Supervisory Board Economically oriented faction in the Supervisory Board Managing director Deputy director Administrator Project leaders Managers of commercially oriented departments Managers of ideologically oriented departments Trade unions Municipality of Amsterdam Employee Board

0 .9 0 .2 0 .9 0 .2

Transition manager Accountant Chairman of the Supervisory Board Competitors

0 .60 0 .80 0 .10 1 .00 0 .10 Note:

80 50 10 50 10 0

Position : Percentage of projects with an ideological orientation .

The interests of the stakeholders are not only represented in the positions they take, but also in the degree to which they are interested in the issue . We denote this as the salience of the issue for the stakeholder . The salience indicates the willingness of the stakeholder to push his position in the decision making . If the issue is hardly related to essential interests of a stakeholder, he will hardly fight for an outcome close to his position and he is to be expected to compromise quickly . If the issues are related closely to his main interests, a stakeholder will put all available clout behind his position and he is unlikely to compromise quickly . In the table, salience is scored on a scale from 0 (not related to own interests) to 1 (the stakeholder is willing to go to the limit) . The transition manager was hired to transform the company into an economically viable one . As the success of his task depended strongly on his success on this issue, the issue was very salient to him . In other words, he had a strong interest in the outcome . The chairman of the supervisory board hired the transition manager and had committed himself to make the company economically viable . For him too, success on this issue was highly important . Failing would probably have cost him his chairmanship . The other stakeholders took other positions but were also less committed to them . In the literature and in practice it is often forgotten that the combination of position and salience determines the behavior of stakeholders . These combinations are generated by the incentive structures of stakeholders . If one of the two is overlooked (in practice often salience), serious miscalculations are inevitable . The third element is the capabilities of the stakeholders to co-determine the outcome. This refers to a potential, not the actual mobilization of the capabilities .



1 36

FRANS N . STOKMAN ET AL .

Formal discretional or voting power is an important capability to co-determine the outcome of decisions . Nevertheless, people with the authority to make the final decision do not do so in isolation without taking into consideration the interests of the other stakeholders . If they were to neglect the interests of others, their decisions would quickly provoke societal and political opposition, resulting in the undermining of their own authority and legitimacy . Apart from that, most decision makers lack the required expertise to decide themselves and have to rely heavily on experts and advisors who are often strongly committed themselves . In addition, decision makers often need the cooperation of others for the implementation of their decisions, for example for the required financial resources or for loyal implementation . Other capabilities, therefore, result from exclusive expertise ahd information, financial resources or the capability to mobilize a large number of people. Scaling capabilities therefore requires the addition of things that are difficult to compare . Nevertheless, experts are often able to provide such a scale . During the extraction procedure, a "1" is often assigned to the stakeholder with the largest capabilities . Subsequently, the capabilities of the others are related to the one with the highest capabilities and to the capabilities of each other . Additivity is assumed here . If, for example, two stakeholders form a coalition against a third, their joint capabilities are compared with those of the third to determine which of them is stronger . When a subject area specialist tries to construct such a scale, it is important to make many such comparisons, that is, to compare many opposing coalitions of different compositions and sizes . Capabilities and salience together determine the effective power of stakeholders . Salience can be interpreted as the fraction of capabilities that a stakeholder is

2.5 a

2 m

•

1

5

• capabilities effective power

1

05 a • u

0 O Position Figure 1 .

Capabilities and Effective Power on the Position Scale



Strategic Decision

Making

137

willing to mobilize to effectuate his position in the outcome of the decision . A graphic showing the distribution of capabilities and effective power over the positions often generates much insight in the decision-making setting . Figure 1 contains an example of such a graphic for the data in Table 1 . Two aspects become clear immediately : a large amount of capabilities was concentrated behind the 10 percent position and almost all of that potential was mobilized . The latter was not true on the other side of the continuum . There, only a small fraction of the capabilities was mobilized . This immediately leads to the conclusion that the transition manager (who had the position of 10 percent) had to avoid a strongly confrontational strategy because this would have had the danger of mobilizing a countervailing power . If that had happened in this example, it would also have resulted in a bipolar distribution of effective power over the continuum . Such issues are often difficult to resolve in contrast to issues where the power distribution is unimodal .

PROCESSES AND STRATEGIES The dynamics of decision making result mainly from the efforts of stakeholders to realize an outcome of the decision that is as close as possible to their own position . The first and most obvious target for these efforts are the positions of the other stakeholders : is it possible to move their positions toward one's own, thus building an effective coalition behind one's own position? Building such a coalition makes it more likely that final decision makers will shift their positions, resulting in a decision closer to one's own position . A characteristic of collective decision-making processes is that not only one stakeholder attempts to build support for his own position, but that they all do, thus provoking countervailing forces with the possibility that the coalition behind one's own position is not strengthened but weakened . An analysis of a particular strategy might even result in the insight that one's own position cannot be maintained whereas the aim was to build a strong coalition behind it . There are three main processes through which a stakeholder changes his position : he receives convincing information implying that another position reflects his incentive structure better (management of meaning) ; he feels more or less forced to change his position because others challenge his position ; or he is prepared to take another position in exchange for a favorable (for him) move by another stakeholder on another issue . In management of meaning and challenge processes, in principle at least, only one issue is involved, whereas in exchange more issues are always involved . We therefore consider management of meaning and challenge processes first . The three processes are also points of departure for individual strategies and we will therefore also consider them from this perspective in the next sections .



1 38

FRANS N . STOKMAN ET AL . Management of Meaning Processes and Strategies

In management of meaning processes convincing information plays a dominant role (Weick 1979 ; O'Reilly 1983 ; Bouwen 1993) . As stated above, the position and salience of a stakeholder are connected to his incentive structure . The more directly an issue is connected to the central higher ordered objectives of a stakeholder and the more an issue is seen as an important condition for their realization, the more salient the issue is to him . His position on the issue corresponds to the outcome of the decision that he sees as optimal for the realization of his objectives . It does not matter what his higher ordered objectives are or where they come from . They may be related to the goals of the organization he represents or they may be related to his personal goals, such as status and behavioral confirmation, or to a combination of these . His own perception of the salience of the issue and the best outcome (position) in terms of these higher ordered objectives is essential . New information can drastically change this perception of salience and best position only if the new information is seen as reliable and is accepted as relevant . Two conditions substantially increase the likelihood that this will happen and, consequently, determine the success of convincing information strategies . First, convincing new information is generally more easily accepted in earlier stages of the decision making than in the later ones . During the decision-making process, positions taken by the stakeholders tend to freeze . In the beginning, stakeholders often do not have very crystallized ideas about the connection between the issue and their objectives . They are relatively open to information that clarifies that connection . Second, a substantial amount of trust in the provider of the information increases the likelihood that information is accepted as relevant and reliable . Trust will be greater if the provider has proven to be reliable in the past and if the provider can be expected to experience future negative consequences of providing distorted or incomplete information . This "shadow of the future" (Axelrod 1984) is more effective if the provider of distorted information loses his reputation not only with respect to the recipient stakeholder but also with respect to others (Raub and Weesie 1990; Buskens 1999) . Trust will also be greater if the information is less related to central interests of the provider . These conditions for trust emerge more readily among like-minded stakeholders and among stakeholders who also meet each other in other contexts, than among stakeholders with conflicting interests . Stakeholders also tend to assign more weight to the opinion of powerful stakeholders, whereas powerful stakeholders tend to listen more to one another than to less powerful ones (Molm 1997 ; Stokman and Zeggelink 1996) . 5 In general, convincing information strategies do not work well or sufficiently in relation to strongly polarized issues (see Figure 1) and in situations with huge power differences . Misuse of trust strongly reduces the effectiveness of convincing information strategies in the future . In decision-making processes this should not only be a guideline for one's own behavior, but can be made unequivocally clear to



Strategic Decision Making

139

others, thus reducing their inclination to provide distorted or incomplete information . In the context of convincing information strategies it is important to take account of the fact that people tend to be myopic in two respects . First, short-term benefits and costs weigh stronger than long-term ones, otherwise nobody would smoke, nor would obligatory retirement arrangements be necessary (Lindenberg 1998) . Second, stakeholders tend to see small differences around their own position much sharper than differences among remote positions (Bauer at al . 1963) . This may result in strong trust crises in one's own coalition about differences of opinion that outsiders would classify as being of minor importance. Challenge Processes and Strategies

In challenge processes differences in power and salience play an important role . To the degree that the salience of the issue to oneself is lower and the power dominance of stakeholders who support other position is larger, one will be inclined to give up his own position . Unnecessary costs are saved by conceding a position on an issue that is only marginally related to one's own interests . In challenge processes power dominance relations are more important than arguments . The process is therefore represented in the form of a noncooperative game (Bueno de Mesquita et al . 1985 ; Bueno de Mesquita 1994) . A challenge is more likely to be successful to the degree that the salience of the other stakeholder is lower and the support for the other's position is lower relative to the support for one's own position. In the computer simulation these two aspects dominate the determination of which stakeholder will challenge which other stakeholder (see Figure 2) . Each stakeholder makes this choice in relation to each of the other stakeholders . Based on the challenges made, each stakeholder has a set of cards in his hands that represent the challenges made and received . If he received challenges, he has to draw the one that is best for him . The result is either conflict (if he made a challenge to the other as well) or a forced position change . These position changes create a new decision-making setting (iteration in the computer simulation model) . In that new setting the stakeholders repeat the choice process . This continues until none of the stakeholders move anymore (substantially) or until all stakeholders take the same position . The latter does not occur often but it does in the example of Table 1 . Table 2 shows how the decision making evolves on the issue of the ideological orientation of the company . The first column gives the original positions of the stakeholders as the subject area specialist provided them (see Table 1) . The second column shows that (in the computer simulation) some stakeholders had to adapt their position because of the challenges obtained . Almost all movements are in the direction of the 10 percent position . After the first round or iteration the support for that position increases substantially, resulting in a still stronger movement in that direction . In iteration four all stakeholders take that position . In reality,



1 40

FRANS N . STOKMAN ET AL .

consensus emerged around that position, notwithstanding the very controversial nature of the issue at the beginning . For challenge strategies to be successful at least three elements have to be considered explicitly . The first involves the choice of one's own position at the beginning of the decision-making process . Will a more extreme position result in a better outcome (closer to one's own position or an outcome with a larger amount of political and social support) or is it better to choose for a more moderate position? The same holds for the leverage a stakeholder shows to others . 6 Is it better to push with all available clout or will a less involved stance to better results? If decision-making processes were to evolve monotonically, a more extreme position and a higher leverage would always lead to a better result, but this is not the case . The main reason is that individuals behave differently in loss situations than in gain situations, both in individual decision making (e .g ., Kahneman and Tversky 1979) and in collective action in social dilemma situations (Raub and Snijders 1997 ; Van Assen 1998) . In loss situations individuals are more willing to accept risks, whereas they tend to be more risk averse in gain situations . If a stakeholder confronts his adversaries with a strong loss situation by taking an extreme position and showing a high leverage, his adversaries tend to become more risk seeking and resisting . The consequence may well be that he loses support for his position and that he has to accept a worse outcome . These effects are also incorporated in the simulation model and can be evaluated systematically . If, for example, the transition manager in the example of Table I had chosen for the position

Challenge Existing Policy

Do Not Challenge Existing Polic

No Change Policy

Source :

Policy Changes

Opponent Gives in

Bueno de Mesquita et al . 1985

Figure 2 .

The Challenge Process

Opponent Does Not Give in

141

Strategic Decision Making Table 2.

Example of Evolution of Challenge Process (Computer Simulation) Positions in iterations Iteration Iteration Iteration Iteration

Initial

1

2

3

4

70 10

10 16

40 80 80 90 30 70 100

10 80 35 90 10 70 100

10 10 10

10 10 10

10 10 10

80 50 10 50 10 0

35 10 10

10 10 77 10 10 100 10 10 10

10 10 10 10 10 40 10 10 10

10 10 10 10 10 12 10 10 10

10 10 2

10 10 10

10 10 10

10 10 10

Ideological faction in the Supervisory Board Economically oriented faction Supervisory Board Managing director Deputy director Administrator Project leaders Managers of commercially oriented depts . Managers of ideologically oriented depts . Trade unions Municipality of Amsterdam Employee Board Transition manager Accountant Chairman of the Supervisory Board Competitors

of 0 percent, the predicted outcome would still be 10 percent . However, unanimity would then only have been reached after six iterations with a larger amount of conflict . This would have also resulted in a lower prestige for the transition manager, as the computer simulation shows he would have had to move from 0 to 10 percent in the fifth iteration . In addition to an explicit reflection on one's own position and leverage, an explicit evaluation of the likelihood of success of challenges is recommended . If a stakeholder takes an extreme minority position, his own chances of successful challenges are likely underestimated (resulting in no or less challenges) for the same reason as above : there is a tendency to overestimate the risks stakeholders close to the expected outcome are likely to take . As they are in a gain situation, stakeholders with positions close to the expected outcome are reluctant to make challenges and are an easier target than stakeholders with extreme minority positions expect . The simulation model provides insight into such likely misperceptions of stakeholders and their effect on the outcome of the decision . Repeatedly, these misperceptions have been shown to exist in reality and their analysis has increased success rates considerably (Ray and Russett 1996) . Exchange Processes and Strategies In exchange processes contrasting interests play a dominant role . Contrasting interests manifest themselves in a combination of fundamental differences in positions and salience over different issues . Whereas management of meaning



1 42

FRANS N . STOKMAN ET AL . Table 3 .

Position Exchange Possibilities Issue 2 Left

Issue 1

Right

Left

Right

strategies are effective among like-minded stakeholders, exchanges of positions are most profitable between stakeholders who have (1) opposing positions on the issues and (2) very different patterns of salience . The latter means that one stakeholder is mainly interested in one issue, whereas the other issue is strongly related to higher objectives of the other stakeholder . The result of a position exchange between two stakeholders is that they shift their voting position in the direction of the other on the issue that is relatively more salient for the other. In this way they create a win-win situation . Table 3 describes exchange possibilities in terms of positions held by the stakeholders . Exchange possibilities exist between stakeholders who hold opposing positions on both issues, that is, are located at different sides of the expected outcome on both issues. From this perspective we distinguish four groups of stakeholders . Stakeholders in group A are located on the left side of the expected outcome on both issues, stakeholders in group D on the right side . They have opposing interests on both issues and are therefore important potential exchange partners . The same holds for the groups B and C . As A and B have the same position on the first issue, they cannot exchange . A and C cannot exchange as they have the same position on the second issue . Similar arguments hold for the groups C and D, and B and D . Stakeholders from two groups with opposing positions can profit from position exchange if the relative salience of the two issues for each of them is different (Stokman and Van Oosten 1994) . A position exchange is then profitable for both, but has also important side effects for others . This can be seen clearly in Figure 3 . Assume a stakeholder from group D attaches relatively more salience to issue 1 than to issue 2 if we compare his saliences with those of a stakeholder in group A. Then issue 1 is D's demand issue and A's supply issue . Position exchange between A and D implies that A is willing to shift his position on issue 1 in the direction of D, whereas D does the same on issue 2 . If they do, they both shift away from B in the direction of C on both issues . B is thus punished doubly and C rewarded doubly, while none of the two is engaged in the exchange . If issue 1 would have been A's demand issue, C would have been punished twice and B rewarded twice .



Strategic Decision Making

143

Exchanges can also be threatening for internal group cohesion . Threats emerge when stakeholders in a group differ in their prioritization of the issues . We consider again two potential exchange groups, for example A and D . Assume that there are two stakeholders in group A, A 1 and A2 (see Figure 4) . Assume also that the salience of issue 2 for A 1 relative to issue 1 is higher than that for the stakeholders in group D, while the reverse is true for A 2 . Then D can split group A . For A 1 , issue 1 is the supply issue on which he is prepared to shift his position in the direction of D . In contrast, A 2 wants to maintain his position on issue 1 and is prepared to shift his position on issue 2 in the direction of D . If D were to exchange with both, the effect would be that D shifts on both issues in the direction of A, and that A 1 and A2 differ in their positions on both issues (see Figure 4) . It can be shown that D creates a prisoner's dilemma for A 1 and A2 where both A 1 and A2 are better off not exchanging at all (Van Assen et al . forthcoming) . The representation of exchanges used in our models is fundamentally different from the representation used in most social exchange theory literature . The representation of exchanges we use is similar to that used by Coleman (e .g ., 1972, 1990) who also represents both actors' interests in goods and the goods actors possess . In network exchange theory the division of a common resource pool (e.g ., Bienenstock and Bonacich 1992 ; Cook and Yamagishi 1992 ; Friedkin 1992 ; Markovsky, Willer, and Patton 1988 ; Skvoretz and Willer 1993) is used to represent exchanges . Van Assen (forthcoming) demonstrates that the often-used division of a common resource pool is generally not a theoretically valid representation of exchanges . All solutions proposed for the exchange rate in the literature can be applied to both our representation and the representation in terms of the division of a common resource pool . Stokman and Van Oosten (1994) use an equal utility gain for both exchange partners . This has the advantage that exchanges have the same utility for both partners and can be handled in order of attractiveness to both . The disadvantage is that it involves an intersubjective comparison of utility, which is theoretically unattractive . For our representation of exchanges two solutions of the exchange rate have been derived that are independent of the utility scale : the Nash solution (Achterkamp 1999 ; Van Assen forthcoming), and the Raiffa-Kalai-Smorodinsky (RKS) solution (Van Assen forthcoming) . There is a strong link between our work and network exchange theory because exchange-resistance solution used in network exchange theory (e .g ., Willer, Markovsky, and Patton 1988 ; Skvoretz and Willer 1993 ; Willer and Szmatka 1995) is also derived from the game theoretical RKS solution (Heckathorn 1980) . Van Oosten (forthcoming) bases the exchange rate on satisficing adaptation of random rates . A difference between our work and work in network exchange theory is that the latter until now only deals with exchanges of private goods . This is also the starting point for Coleman's models and his generalizations to public goods are not straightforward (Stokman and Van Oosten 1994) . We, however, investigate



1 44

FRANS N . STOKMAN ET AL .

exchanges of voting positions with externalities for other actors . A change in voting position affects all stakeholders in collective decision making . Interestingly, network effects on exchanges and their rates in social exchange theory are completely different from these effects in exchanges of goods with externalities to other actors (Van Assen et al . forthcoming). In exchange processes, three elements have to be considered explicitly when formulating one's own stance . The first concerns the selection of the issues one wants to include in the exchange process . Of course, one cannot always determine this alone, but there are many situations in which this is possible, for example, through the order in which different issues are negotiated . The second is the change one incorporates deliberately into one's own positions . Change will in general be incorporated in the less salient issues and is designed to facilitate concessions during an exchange . One should of course realize that extreme positions could provoke extremity on the other side, which easily nullifies the positive effect by resulting in more conflict and tension . The third element concerns one's prioritization of the issues, the distribution of the leverage over the issues that one shows . Here too a certain balance is needed . If one shows too much leverage on highly salient issues, one has to give in a lot on the other issues to obtain a position shift by the other stakeholders (it is generally known that in social relations, the person most interested in the relation tends to be exploited (Blau 1964) . If one does not show enough leverage on highly salient issues, one's own position and salience may well not be discerned by others, giving them an opportunity to move

Issue I

A}

rte

C D

B 01

Issue 2 A

D

-4

B

C 02

Figure 3 .

Effects of Exchanges for Third Stakeholders

I



145

Strategic Decision Making

Issue 1 (q1

~

01

Issue 2 A1



D

-.4

D

-0 1

1

02

After two exchanges both A t and A 2 have changed their position . Figure 4 .

Effects of Differences in Prioritization within Groups

the issue aside and to maintain the status quo . If one shows too quickly that one is not really interested in less salient issues, the proposal to shift one's own position on these issues may well not be seen as a serious concession that justifies a move on another issue by the other stakeholder . An Example of a Combined Strategy

Table 4 contains an example of how different strategies can be combined to determine which issues should be included in the exchange process in order to obtain an optimal outcome . The example is derived from an application in the context of Dutch labor negotiations . The main issue concerned a fundamental change in social benefit rights . The data were collected just before the start of the negotiations . In the row "Position" the positions of the client at the start of the negotiations are given . The client aimed at an obligatory start of the new system for all employees younger than a certain age (position 100) . The other stakeholders aimed at an obligatory start only for new employees (position 70) or no new system at all (position 0 ; status quo) . The positions 70 or 0 would imply considerable delay or no change at all . A second important issue in the new regulation concerned a certain degree of freedom for the employees to adapt the new system to their own wishes (in Table 4 denoted "Introduction of Choice Elements") . The client was for a large amount of freedom (position 100), while the other parties

146

FRANS N . STOKMAN ET AL .

Table 4.

Effects of Different Strategies in Labor Negotiations

Position Outcome before Exchange, No Optimization Challenge at Main Issue, Rest through Exchange Best Exchange on First Three Issues Exchange without Alternative Exchange Issue Exchange without Choice All Issues Included in Exchange Notes:

Alternative Introduction Exchange Of Choice Issue Elements 0 100

Main Issue 100

Guarantee 60

Exchange Issue 50

40

75

70

100

70

70

70,75

70

50

70,100

100

63,75

70

Excluded

100

100 100

68,75 70,75

70 70

Excluded 40

81,100 100

100

70,75

70

40

86

Not included in exchange solution . Italics : Acceptance by unanimity . Normal : majority, including the largest stakeholder . Double number : Majority for first, excluding the largest stakeholder . Bold :

wanted the freedom to be limited strongly (position 0) . A third important issue concerned guarantees for a minimum level of the benefits (the higher the position, the higher the guarantee) . A fourth issue concerned the level of another social benefit, in the table denoted "Exchange Issue ." The client introduced this issue explicitly as his supply issue and chose to adopt a low position initially that could be changed during the negotiations . An external advisor from an organization with a much experience in this subject area advised to include another issue in the exchange process, denoted "Alternative Exchange Issue" in Table 4 . This would have been very risky, as any concession on this issue would have been very expensive for the client . Moreover, the issue is normally dealt with in another negotiation context . The client's position was 0 (no change) against position 100 supported by all other stakeholders (resulting in extra benefits) . The client aimed at an outcome of the negotiation process that was not only close to his positions but was broadly supported as well . The row "Outcome Before Exchange, No Optimization" contains the outcomes predicted by the computer simulations when all issues are dealt with in isolation . These are the predicted outcomes on the basis of the aforementioned challenge strategy . They are obtained in different iterations without optimization aiming at outcomes closer to the positions of the client or at broader support . The predicted outcome on the main issue implies no change in the system or only in the long run . The outcomes on the other issues are also far from the preferred positions of the client. Under this scenario, the best course of action for the client would be to call a halt to the decision-making process immediately .



Strategic Decision Making

147

The row "Challenge at Main Issue, Rest through Exchange" gives the outcomes with an optimal challenge strategy on the main issue combined with an exchange process over all remaining issues . With optimization we were able to arrive at a computer prediction of 70 for the main issue, corresponding to implementation for all new employees . The optimization consisted of a carefully selected leverage the client should show in combination with certain challenges the client would likely miss (and were indeed surprising for him) . The exchange outcomes on the other four issues are not very promising . None of the predictions corresponds with his position . He has to give up a lot on the other issues in return for a better result on the Alternative Exchange Issue that has a high salience for him because of the large costs involved . Even then, he has to make a huge concession on that issue . The next row, "Best Exchange on First Three Issues," shows the results of a strategy that produces broadly supported outcomes close to the positions of the client . In this solution the Alternative Exchange Issue is entirely left out of the decision-making process . In isolation from the other issues, the issue cannot be solved . The best solution would be to exclude the issue entirely from the decision making . The main argument for its exclusion is the other negotiation platform to which it belongs . Our analysis therefore showed that the advice of the external advisor to include this issue was really very disadvantageous for the client . We advised that the decision on the issue Choice Elements should be made after negotiations on the first three issues had taken place . Choice Elements is the only nonfinancial issue . Moreover, the outcomes on the first three issues have to be known before one can make sensible decisions on this issue . Using an optimized challenge strategy the issue could be resolved according to the position of the client. By using an exchange strategy, broadly supported outcomes close to the positions of the client were obtained on the first three issues . In the last rows of the table the results are shown of alternative strategies in which more issues are included in the exchange . All these strategies produce worse results than the strategy "Best Exchange on First Three Issues ." Therefore, we recommended that strategy and the client implemented it successfully . This example demonstrates the power and fruitfulness of our approach .

CHANGING THE DECISION SETTING The management of meaning, challenge, and exchange strategies take a decision setting as given . The setting corresponds to the specification of the issue(s) and the stakeholders with their capabilities, positions, and salience . If none of the strategies is effective and the issues are essential to obtain higher ordered objectives, a stakeholder can deliberate on whether the setting itself could be changed for better goal attainment . Stakeholders are inclined to pose this question more forcefully when they, time and time again, do not succeed in obtaining their objectives or when the decision-making process is blocked repeatedly by funda-



148

FRANS N. STOKMAN ET AL .

mental controversies . The approach presented above focuses directly on how to realize such a change in the decision setting . As the stakeholders with their power, positions, and salience determine the decision-making setting, strategies to change the decision setting can best be evaluated on their likely effects on the distribution of these three variables . The combination of position and salience of the stakeholders is connected with the incentive structures of the stakeholders . Changing the incentive structures is consequently the most important point of departure to obtaining fundamental changes in position and salience of certain stakeholders . Assume, for example, that the cooperation of farmers is essential for the success of an environmental policy concerning the disposal of manure, and that they are not prepared to cooperate freely or to accept certain policy measures . Their position and salience can then be influenced by a tax system that shifts their short-term economic interests in the direction of a more environment-friendly way of farming . This will lead to another stance of the farmers on issues related to manure policy (Baarda 1999) . An important part of the power distribution between stakeholders is related to institutional arrangements . Long-lasting deprivation of certain stakeholders often results in a demand for institutional change, for example, through changes in the composition of bodies and advisory committees, changes in competencies between bodies, extension or reduction of possibilities for appeal . Changes in institutional arrangements and customs are therefore the most important ways to change the power distribution among stakeholders . In addition, other, more informal capabilities can be increased or diminished . One can think of alliances, takeovers, or even mergers that can shift the power balance considerably .

CONCLUSIONS We have demonstrated that far-reaching insights into the decision-making process can be obtained from a correct problem decomposition, a decomposition resulting in a few main controversial issues, and a proper specification of the decision setting (stakeholders with their capabilities, positions, and salience) . These insights concern the processes and the likely outcomes of the process, but also the amount of conflict involved and the stability of the outcomes . These insights are obtained by combining a systematic decomposition of the problem into main issues with systematic interviews of subject area specialists and computer simulation . In these computer simulation models not all aspects of the decision-making process are represented, but only the main processes through which differences in positions and saliences are accommodated in binding decisions . In our view, three main processes are at stake : management of meaning through the provision of convincing information, challenges, and exchanges . In the provision of convincing information, trust and credibility play a dominant role . These conditions prevail among stakeholders with rather similar positions, between stakeholders with bridgeable



Strategic Decision Making

149

power differences, and in situations where the positions are not yet fixed and still have to be formed . In contrast, in challenge strategies stakeholders purposively exploit power differences to build coalitions . In exchange processes large differences in interests prevail, not only in terms of extreme opposing positions but also in terms of the priority rankings of the issues . Important strategic advantages can be gained from the insights generated by such an analysis . They can be used to obtain a better representation of one's own position in the decision outcome, and also to broaden the political and social support behind the decision outcome . With respect to challenge strategies, the effects of increasing or reducing leverage and of taking a more extreme or moderate position can be investigated . Moreover, warnings against often occurring misperceptions are generated, misperceptions resulting from differences between stakeholders' attitudes toward risks that are often neglected . For exchange strategies, the effects of linking or excluding issues, of differences in prioritization, and of built-in change can be investigated . These are fundamental new insights in collective decision-making processes that have been generated by the social sciences since the beginning of the eighties . This has resulted in a shift of attention from questions about power differences to differences in the effects of power differences . These effects are context dependent . It is the combination of the power distribution over positions and the differences in leverage of the stakeholders that co-determines the setting . It is this combination that conditions the effects of power differences (Mokken et al . 1998) . The approach presented above has passed the experimental phase . The team of scholars around Bueno de Mesquita and the Dutch group analyzed more than 6,000 issues . In the United States the first applications were related to American foreign policy, but gradually the emphasis has switched toward business applications . In the Netherlands the method has been applied for several ministries, large municipalities, and large companies for the analysis of complicated and important issues . Time and time again the clients indicated that they obtained many new insights through the approach and that they were able to realize their objectives more than they expected to do . Many clients returned for advice on other problems . Important scientific questions remain . Within the domain of collective decision making a further analysis of the three processes continues . The conditions under which each process is dominant is one of the main questions to be answered (see also Achterkamp 1999) . This is of particular importance because some of these processes are based on cooperative strategies, whereas others are based on conflictual strategies . Under what conditions do we need to be prepared for conflict strategies and what macro effects do such conflictual strategies have? Do they result in other and worse outcomes from the perspective of social welfare? Do noncooperative strategies lead to longer and more difficult decision-making processes? Is it possible and desirable to add complexity to the models? How do these models relate to other models, such as coalition and pure game theoretic models?



150

FRANS N . STOKMAN ET AL .

From the perspective of societal relevance, the conditions for broadly supported outcomes are of central importance . The approach described above enables much more sophisticated research to be conducted on this question, thus contributing to research with a view to the enhancement of social cohesion and solidary behavior . In addition, central questions remain concerning the relation between collective decision making and other processes . The models make intensive use of insights from game theory, social psychology, and economics . Nevertheless, further investigation of the relationship between these areas in general, and individual and collective decision making in particular could well result in important new insights . Such research could help to derive positions and salience on issues from the incentive structures of stakeholders . Research on the weight of the different power components in the overall power or capability index of a stakeholder could provide fundamental insights to be used in the design of institutional arrangements in society . All these questions belong to the central questions of the social sciences and cross several disciplinary boundaries . Answers to these questions will likely have huge societal impact and utility .

NOTES 1 . This paper is based on insights gained during the long-term development of the methodology described here, and applications of collective decision-making models . The roots of this development can be found in Mokken and Stokman (1976) . Part of the methodology is based on a model developed by Bruce Bueno de Mesquita of Stanford University . The collaboration with him has been very inspiring and helped in developing our own contributions . Many insights originate from scientific applications, mainly within the context of PhD projects in the Dutch Interuniversity Research Institute and Graduate School ICS . Commercial applications of the methodology by DECIDE by helped significantly in the systematic analysis and testing of the effectiveness of strategic moves. DECIDE by is a company associated with the University of Groningen and the ICS . We thank Robert Thomson for correcting the English . 2 . Wherever we use the male form, we also imply the female one . 3 . A good illustration of the accuracy of the models was given in 1996 . At the start of the labor negotiations in the Dutch metal industry (July 1996) Maurice Rojer and the first author collected data on 16 controversial issues and publicly announced that we had given our predictions in a sealed envelope to a notary . At the end of the negotiations in September we opened the envelope at a press conference. One issue was held in study . Of the other 15 issues we predicted 13 correctly according to sharp criteria (87%) . Among these 13 were exact predictions of the salary increase and the changes in the retirement arrangements, two of the most controversial issues. 4 . All models discussed in this paper assume that the issues are unidimensional and that the stakeholders have single-peaked preference functions . 5 . In our management of meaning model, three parameters are estimated for policy networks . The first parameter is related to the overall density of influence relations, the second to the difficulty to surmount power differences in establishing influence relations, and the third to the importance of similarity of positions on issues in establishing influence relations (Van Assen and Stokman forthcoming) . If no network data are available (only the capability, position, and salience scores of the stakeholders on the issues), the model can be used to estimate the network effects and the effects on the outcomes of the issues of different values of the parameters.



151

Strategic Decision Making

6. The term leverage is used to refer to the strategically chosen salience in decision-making processes. For a model that optimizes the leverage over a set of issues, see Stokman and Stokman (1995) . 7 . The exchange process as described and modeled is an example of a cooperative game model. The analysis of win-sets is an example of a noncooperative game through which stakeholders can create win-win situations . Win-sets are solely based on an analysis of the positions of the stakeholders . They do not take salience into account .

REFERENCES Achterkamp, M. 1999 . Influence Strategies and Models in Collective decision Making . A Comparison of Two Models. Amsterdam : Thela Thesis . Axelrod, R. 1984 . The Evolution of Cooperation. New York: Basic Books . Baarda, C . 1999 . Politieke Besluiten en Boerenbeslissingen . Her draagvlak van het Mestbeleid tot 2000. Amsterdam : Thela Thesis . Bauer, R . A ., I . de Sola Pool, and L. A . Dexter. 1963 . American Business & Public Policy . The Politics of Foreign Trade. Chicago : Aldine Atherton . Blau, P. M . 1964 . Exchange and Power in Social Life. New Brunswick, NJ : Transaction Publishers . Bienenstock, E . J ., and P . Bonachich. 1992 . "The Core as a Solution to Exclusionary Networks ." Social Networks 14 : 231-243. Bouwen, R . 1993 . "Organizational Innovation as a Social Construction : Managing Meaning in Multiple Realities ." Pp . 133-49 in Interdisciplinary Perspectives on Organization Studies, edited by S . Lindenberg and H . Schreuder . Oxford : Pergamon Press. Bueno de Mesquita, B . 1994 . "Political Forecasting : An Expected Utility Method ." Pp . 71-104 in European Community Decision Making . Models, Comparisons, and Applications, edited by B . Bueno de Mesquita and F .N. Stokman . New Haven, CT: Yale University Press . Bueno de Mesquita, B ., D . Newman, and A. Rabushka. 1985 . Forecasting Political Events, The Future of Hong Kong. New Haven, CT : Yale University Press . Bueno de Mesquita, B ., and F . N. Stokman, eds . 1994. European Community Decision Making. Models, Comparisons, and Applications. New Haven, CT : Yale University Press . Buskens, V . 1999 . Social Networks and Trust. Amsterdam : Thela Thesis . Coleman, J . S . 1972 . "Systems of Social Exchange ." Journal of Mathematical Sociology 2 : 145-163 . Coleman, J. S . 1990. Foundations of Social Theory. Cambridge, MA : The Belknap Press of Harvard University Press . Cook, K. S ., and T . Yamagishi. 1992 . "Power in Exchange Networks : A Power-dependence Formulation ." Social Networks 14 : 245-265 . Friedkin, N . E . 1992. "An Expected Value Model of Social Power : Predictions for Selected Exchange Networks." Social Networks 14 : 213-229. Heckathorn, D . D . 1980 . "A Unified Model for Bargaining and Conflict ." Behavioral Science 25 : 261-284 . Kahneman, D ., and A . Tversky. 1979 . "Prospect Theory : An Analysis of Decision Under Risk." Econometrica 47 : 263-291 . Konig, T. 1997 . Europa auf dent Weg zum Mehrheitssystem . Griinde and Konsequenzen nationaler and parlamentarischer Integration . Opladen: Westdeutscher Verlag . Laumann, E . 0., D. Knoke, and Y .H. Kim . 1987 . "Event Outcomes ." Pp. 343-373 . in The Organizational State . Social Choice in National Policy Domains, edited by E. O . Laumann and D . Knoke. Madison : The university of Wisconsin Press . Lindenberg, S . 1998 . "Solidarity : Its Microfoundations and Macrodependence . A Framing Approach." Pp . 61-112 in The Problem of Solidarity. Theories and Models, edited by P. Doreian and T . Fararo . Gordon and Breach Publishers.



152

FRANS N. STOKMAN ET AL .

Markovksy, B ., D. Willer, and T . Patton 1988 . "Power Relations in Exchange Networks ." American Sociological Review 53 : 220-236. Mokken, R . J ., and F . N. Stokman . 1976 . "Power and Influence as Political Phenomena." Pp . 33-54 in Power and Political Theory, edited by B . Barry . London : John Wiley . Mokken, R . J ., D . Payne, F. N . Stokman, and F . W. Wasseur . 1998 . "Praferenzaggregation and Effektivmacht. Die Reform der EG-Strukturfonds fur Irland ." Pp . 344-362 in Europa der Burger? Voraussetzungen, Alternativen, Konsequenzen, edited by T. KSnig, E . Rieger, and H . Schmitt . Frankfurt : Campus Verlag . Molm, L . 1997 . Coercive Power in Social Exchange . Cambridge: Cambridge University Press . O'Reilly, C . A . 1983 . "The Use of Information in Organizational Decision Making : A Model and Some Propositions ." Research in Organizational Behavior 5 : 103-139 . Pappi, F . U ., and C . H. C . A . Henning . 1998 . "Policy Networks : More than a Metaphor?" Journal of Theoretical Politics 10 : 553-575 . Payne, D . 1999 . Policy Making in the European Union . An Analysis of the Impact of the Reform of the Structural Funds in Ireland . Amsterdam: Thela Thesis . Raub, W., and J. Weesie . 1990. "Reputation and Efficiency in Social Interactions : An Example of Network Effects." American Journal of Sociology 96: 626-654. Raub, W., and C . Snijders . 1997 . "Gains, Losses, and Cooperation in Social Dilemmas and Collective Action: The Effects of Risk Preferences ." Journal of Mathematical Sociology 22: 263-302. Ray, J ., and B . Russett . 1996 . "The Future as Arbiter of Theoretical Controversies : Predictions, Explanations and the End of the Cold War ." British Journal of Political Science 26 : 441-470 . Rojer, M . 1999 . "Collective Decision making Models Applied to Labour Negotiations in the Netherlands : A Comparison between an Exchange Model and a Conflict Model ." Rationality and Society 11 : 207-235 . Skvoretz, J ., and D . Willer 1993 . "Exclusion and Power : A Test of Four Theories of Power in Exchange Networks ." American Sociological Review 58: 801-818 . Stokman, F . N ., and J . M . M . Van den Bos . 1992 . "A Twostage Model of Policy Making : With an Empirical Test in the U.S . Energy Policy Domain ." Pp . 219-253 in The Political Consequences of Social Networks, Volume 4 of Research and Society, edited by G . Moore and J . A . Whitt. Greenwich, CT : JAI Press . Stokman, F . N ., and R . Van Oosten. 1994 . "The Exchange of Voting Positions : An Object-Oriented Model of Policy Networks ." Pp. 105-127 in European Community Decision Making . Models, Comparisons, and Applications, edited by B . Bueno de Mesquita and F . N. Stokman. New Haven, CT : Yale University Press . Stokman, F . N., and J. V . Stokman . 1995 . "Strategic Control and Interests . Its Effects on Decision Outcomes ." Journal of Mathematical Sociology 20: 289-317 . Stokman, F . N., and E . P . H . Zeggelink. 1996. "Is Politics Power or Policy Oriented? A Comparative Analysis of Dynamic Access Models in Policy Networks ." Journal of Mathematical Sociology 21 : 77-111 . Stokman, F . N., and J . Berveling . 1998. "Dynamic Modeling of Policy Networks in Amsterdam ." Journal of Theoretical Politics 10 : 577-601 . Torenvlied, R. 1996 . "Political Control of Implementation Agencies . Effects of Political Consensus on Agency Compliance" Rationality and Society 8 : 25-56 . Van Assen, M. A. L . M . 1998 . "Effects of Individual Decision Theory Assumptions on Predictions of Cooperation in Social Dilemmas ." Journal of Mathematical Sociology 23 : 143-153 . Van Assen, M . A. L . M . Forthcoming. "Representation of Exchanges in Exchange Theory: Review and Critique ." Internal manuscript. Van Assen, M . A . L . M ., and F . N . Stokman . Forthcoming . "Assessing Power and Policy Orientation in Policy Networks ." Internal manuscript . Van Assen, M . A . L . M., F. N . Stokman, and D . Willer . Forthcoming . "Social Exchange Theory Applied to Exchanges of Goods with Externalities ." Internal manuscript.



Strategic Decision

Making

153

Weick, K. E. 1979 . The Social Psychology of Organizing . Reading, MA : Addison-Wesley . Willer, D ., and J . Szmatka 1993 . "Cross-national Experimental Investigations of Elementary Theory : Implications for the Generality of the Theory and the Autonomy of Social Structure ." Pp. 3781 in Advances in Group Processes, Vol . 10, edited by B . Markovsky and E. Lawler . Greenwich, CT : JAI Press .

THE NORMATIVE REGULATION OF POWER

Morris Zelditch and Henry A . Walker

ABSTRACT In most theories of authority, normative regulation is the price that power pays for legitimacy . But what are the mechanisms by which norms regulate power and under what conditions do they in fact constrain its use? In an experimental study of a three-level hierarchy of power and authority, we found that the internal constraint of conscience alone was not sufficient to constrain the abuse of power by authority : Where authority had something to gain, one in four Ss abused power out of selfinterest; where it had nothing to gain, one in three followed their own conscience rather than the situation's norms . But internal constraints are not the only constraints on the exercise of power by authority . The mechanisms of the normative regulation of power depend on the fact that the exercise of authority, at least in organizations, is a collective phenomenon . It is not simply a matter of a superior, A, exercising authority over a subordinate, B, but of multiple actors executing the directives of A in such a way that the behavior of B is in fact directed . The normative regulation of power depends on how norms regulate these other agents of power and how dependence on these other agents, in turn, controls the behavior of A .

Advances in Group Processes, Volume 17, pages 155-178 . Copyright m 2000 by JAI Press Inc . All rights of reproduction in any form reserved . ISBN : 0-7623-0651-3 155



1 56

MORRIS ZELDITCH and HENRY A . WALKER

In the present paper we report the results of nine experiments, five of which investigate the use of power by A and four of which investigate conditions under which that behavior is or is not supported by superiors and peers of A and peers of B . The driving force of the normative regulation of power is the validity of the norms-where by validity we mean a collective consensus that observably governs, and is binding on, the behavior of the members of a group . In the first instance, the effect of validity is on authorization of the behavior of A, its support by superiors and peers of A, and its endorsement, its support by the peers of B . Expectations of authorization and endorsement, in turn, together with the direct effect of validity (i .e., of conformity) and validity's (social influence) effect on propriety, on A's own sense of what is right, are the mechanisms that constrain the abuse of power by A. To the extent that validity is what drives the superiors and peers of A and the peers of B, norms regulate the exercise of power . But we found that external constraints on the abuse of power by authority were themselves easily subject to abuse, especially by eccentric consciences . Under what conditions do norms actually regulate the use of power by authority? Our theory implies that (1) there must be substantial consensus with respect to the norms ; (2) A must actually be dependent on others for his or her power ; (3) the ties of these others to A must be dense ; and, especially important, (4) the ties among these others must themselves be dense. Only to the extent that agents of social control are themselves densely interconnected are the eccentricities of any one agent of control constrained and norms do in fact regulate the abuse of power by authority .

PROBLEM The Normative Regulation of Power

In Barnard (1938) and Weber ([1918] 1968), normative constraint is the price of legitimate power . In both theories, authority is a relation between A and B such that B is obligated to comply with A's directives . A's power to exact B's compliance is said to be "legitimate" if it is in accord with rule and felt to be right . In both Barnard and Weber it is legitimacy that empowers A : Without it, A is unable to mobilize the resources that provide authority with its power ; and, in any case, A has no warrant to expect B's compliance . With it, A expects B to voluntarily comply with A's directives, independent of B's preferences ; and A has the resources required to negatively sanction Bs who fail to comply . But legitimacy imposes constraints on the exercise of power : A has no right to expect B's compliance or to negatively sanction B's noncompliance with directives outside the scope of A's authority . This has always been a puzzling conception of authority because empowering A gives A control over resources that, once available, are available whether or not B's noncompliance is within the scope of A's authority . In Barnard, furthermore, A's judgment is the only judgment that really counts in deciding whether non-



The Normative Regulation of Power

1 57

compliance is or is not within the scope of A's authority . It is true that B also has some power over A : A depends on B for performance of a task valued by A . But B's power is necessarily less than A's . If the only judgment that matters is therefore A's, the only constraint on the illegitimate use of power by A is voluntary and is entirely internal to A . The Frailty of Internal Constraints on the Use of Power by Authority

But internal constraint is a frail protection against the illegitimate use of power by authority : In the first instance, self-interest may override it . In the first experiment we describe a subject (S) that controlled the agenda of a group's decision about whether or not, and how, to change the structure of a centralized network that unfairly under-rewarded everyone in the group except S . S was at the center of the network and was unfairly over-rewarded . The group had a legitimate right to change the structure of the network if a majority agreed, but almost half the Ss at the center of the network made illegitimate use of their control over the agenda to delay or prevent the majority from voting to change the structure of the network. Of this half, about half redressed inequity by other means . But half simply took the money and ran . Thus, one in four Ss allowed self-interest to simply override internal constraints on the abuse of their power . But even if self-interest does not directly override internal constraint, the individual beliefs on which internal constraint depends may be self-serving . In the experiment just described, many Ss resisted change simply because a centralized network was the most efficient way to perform a valued, collective task and its effectiveness made it legitimate (cf. Lipset 1959) . But post-session questionnaires found that belief in the legitimacy of the centralized network by Ss at its center increased as Ss' gain from it increased . Some part of the effect of the legitimacy of the wheel was therefore self-serving . Even if the beliefs on which internal constraint depends are not self-serving, they may differ from the accepted norms of the group, either because of dissensus or because norms, values, and beliefs are open to interpretation even if there is consensus . In a later experiment described below, S, assisting the experimenter (E) in E's office, was responsible for final decisions about how to divide team earnings . Where individual beliefs in equity were inconsistent with an equality norm backed by E, fully a third of the Ss followed their own conscience rather than the norms of the experiment (Walker et al . 1997) . How and Under What Conditions do Norms Regulate Power?

Thus, internal constraint is not sufficient to constrain the illegitimate exercise of power by authority . In Weber, the solution to this problem lies in the fact that legitimacy is external as well as internal to A . To the extent that a moral order is "valid" it is not only "exemplary," it is "binding" (Weber [1918]1968, p . 31) . It is



1 58

MORRIS ZELDITCH and HENRY A. WALKER

made "binding" on A by embedding it in a system of social controls which makes compliance expedient even if A does not share the system's norms . But even in Weber the mechanisms that actually constrain the use of power by authority are unclear . The present paper therefore addresses two questions : First, if there is normative regulation of power, what are its mechanisms? Second, equally important, under what conditions do they constrain the illegitimate use of power by authority?

THEORY

Scope of the Theory

With some elaboration, Dornbusch and Scott's theory of authority (1975) is sufficient to answer the first question (see also Walker and Zelditch 1993 ; Zelditch and Walker 1984) . This theory applies to multilevel groups oriented to goals that can be transformed into divided tasks . These tasks are divided among positions that can be evaluated for performance . In particular, in this theory there is a division of labor in authority : Rights to make final decisions, allocate tasks, direct performance, assess performance, allocate sanctions, and execute sanctions are divided among different positions . Hence, the exercise of authority is a collective rather than individual act . Because of the division of labor in authority, authority depends on the cooperation of others both for resources and for support . A control attempt or an attempt to positively or negatively sanction a subordinate is powerless if these other positions do not support it by carrying out their tasks at authority's command . The theory is driven by evaluation processes : In particular, because authority depends on the cooperation of others, it is accountable to others (cf. Tetlock 1992) . Accountability is effective to the extent that (1) members of the group care about its positive and negative sanctions, (2) the sanctions depend on evaluations, and (3) evaluations depend on performance . In Dornbusch and Scott, task structures are therefore also power structures . Positions are given differential access to the sanctions that control the performance of other positions . But, because of the division of labor in authority, positions of authority are as accountable as any other positions in an organization . It is fundamental to the theory's model of authority that some control others but are themselves subjected to control . Basic Concepts of the Theory

Dornbusch and Scott's theory makes three distinctions that will matter to our analysis of the normative regulation of power . The first is between individual and collective sources of legitimacy, hence between propriety, which is a matter of



The Normative Regulation of Power

1 59

individual belief, and validity, which is a matter of collective consensus . A norm, value, belief, practice, or procedure is "proper" if individual p believes it is right . It is "valid" if it observably governs the behavior of and is binding on members of a group . Thus, from the point of view of an individual actor p, validity in Dornbusch and Scott, following Weber, is an object of orientation ; it is how others will act if p does or does not violate the norms, values, beliefs, practices, or procedures of the group . Validity is difficult to operationalize except in rational legal systems with written statutes made by established procedures . Thus, equality of outcomes is made valid in one experiment described below by incorporating it into written rules to be followed by Ss, playing the role of E's assistant, who are responsible for final decisions allocating rewards . Written codes aside, validity is essentially actual, or from the perspective of any individual actor, perceived consensus . But what is important is the distinction the theory makes between individual belief and an externally given normative order, An important property of the normative order this presupposes, and hence validity, is that it is enduring : In any particular encounter it is pre-given ; between encounters it is trans-situational . A second distinction made by Dornbusch and Scott is between sources of legitimacy and sources of support for it-where "support" includes both execution of directives and positive sanctions for compliance or negative sanctions for noncompliance . Given the second distinction, the third simply stratifies sources of support : Support by the peers and/or superiors of A is authorization ; support by peers of B is endorsement. (We will indicate peers or superiors of A by A' and peers of B by B' .) For example, a superior of A, A', authorizes a directive of A if A' sanctions B for noncompliance with it ; a peer of B, B', dis-endorses it if B' allies with B in resisting it. Mechanisms Constraining the Use of Power by Authority Thus, the theory has four mechanisms that regulate the exercise of power by authority-propriety, validity, authorization, and endorsement-three of which are external to the actor . The mechanism that drives the theory is validity . The theory describes the normative regulation of power simply by assuming that, first, independent of either individual preferences or propriety, validity causes authorization by A' and endorsement by B' of acts by A that are valid, but de-authorization by A' and dis-endorsement by B' of acts by A that are outside the scope of authority's power. Second, again independent of individual preferences or propriety, it assumes that embedding validity in a system of social controls external to A causes A to exercise power when it is valid but deters A from exercising power when it is not . The effect of validity, in turn, is driven by the presupposition of consensus . Following Weber, the theory assumes that a valid social order is (or is nearly) consensual (Zelditch and Floyd 1998) . Hence, individual actors presuppose



1 60

MORRIS ZELDITCH and HENRY A . WALKER

consensus with respect to it in any particular situation : If an actor p (p = A, B, A', or B') accepts a norm, value, belief, practice, or procedure e l as valid in the situation, p expects any other actor in the situation to accept e as valid unless and until p learns otherwise . Given the presupposition of consensus, the validity of norms, values, beliefs, practices, and procedures has four kinds of effect on the conduct of A (and, in particular, on A's use of power) . It has a direct effect, either because custom and law are respected or because "that's the way things are ." But it also has three indirect effects . One indirect effect is through propriety . Perceived consensus has a social-influence effect on individual beliefs : If p believes others believe e, p is more likely to believe e . Individual beliefs cause compliance with the norms regulating the use of power by authority . But propriety and validity also interact in such a way that the effect of propriety on A's conduct is stronger when it is backed by validity. A is more likely to act on his or her conscience when it is backed by the norms of the group . This follows from the scope of the theory : Decisions that affect group outcomes, commit group resources, or enforce group norms have consequences for others. A can be held to account for them, hence owes others a justification for them (Tetlock 1992 ; Zelditch and Floyd 1998). Where a valid norm, value, belief, practice, or procedure exists, behavior in accord with it justifies A's actions . But the presupposition of consensus also creates expectations of authorization, which creates a second indirect effect of validity on the conduct of p . As in Weber, a valid normative order is prescriptive . 2 If an actor p presupposes consensus with respect to e, p not only expects others to accept e, p also expects them to support e . If p acts in accord with e, p expects either positive sanctions for compliance, or at least that others do nothing that contradicts e . If p does not act in accord with e, p expects negative sanctions for noncompliance . To the extent that A is accountable to others, A is therefore constrained to comply with e by expectations of support for e whether or not A personally believes in it . Hence expectations of authorization will constrain the use of power by A . The final indirect effect of validity is through expectations of endorsement . There will be some difference between the effects of the two kinds of support, because they differ both in source and object . The difference in source of support (A' vs . B') is a difference in power . The difference in object of support (A vs . B) is that peers of A directly sanction A, while peers of B support A by negatively sanctioning noncompliance by B, which from the point of view of A is a less direct effect . Together, these differences make authorization a more powerful constraint than endorsement . Nevertheless, to the extent that A depends on B for performance of a task on which A's own evaluations depend, expectations of endorsement will also constrain the use of power by A .



The Normative Regulation of Power

1 61

Normative Regulation of the Mechanisms Constraining the Use of Power by Authority

Thus, propriety is neither necessary nor sufficient for validity to constrain the behavior of A . The theory embeds norms in a system of social controls that it assumes is, in turn, driven by validity . Propriety, of course, may also affect authorization and endorsement . But the theory assumes that validity affects authorization and endorsement independent of propriety (and, of course, of preferences) . It is the fact that validity drives the system of social controls that explains the normative regulation of power . In the first instance, independent of either preferences or propriety, validity causes authorization of acts that are valid, de-authorization of acts that are not . The peers and superiors of A are more likely to support a directive by A if it is valid, less likely to support it if it is invalid . Furthermore, the theory's conception of social control implies that validity and propriety again interact in such a way that the effect of propriety on authorization is stronger when validity is present than when it is absent-that is, that A' is more likely to actually sanction A and/ or execute the directives of A if the personal beliefs of A' are backed by group consensus . Similarly, independent of either preferences or propriety, validity causes endorsement of acts that are valid, dis-endorsement of acts that are not . The peers of B are more likely to support a directive by A if it is valid ; B' either withholds support of B if B does not comply or actually sanctions B negatively for noncompliance . If the conduct of A is invalid, B' is less likely to support A and more likely to support B ; B' may positively sanction B, or even ally with B, if B does not comply. Furthermore, validity and propriety will interact such that the effect of propriety on endorsement is stronger if validity is present than if it is absentB', like A', is more likely to act on individual beliefs if they are backed by consensus .

METHOD Standardized Experimental Setting

Eight of the nine experiments described in the present paper were carried out in a standardized experimental setting. Task

Five participants solved a series of 10 problems, each of which required them to construct a five-point, multi-arc graph . Each participant had some but no-one had all of the information required to solve the problem . Each trial of the experi-



1 62

MORRIS ZELDITCH and HENRY A . WALKER

ment therefore required participants to coordinate information distributed to them at the beginning of each problem (from Bavelas 1950) . A trial ended when each participant submitted a solution to E . Interaction Conditions Participants worked in separate rooms, communicating by written messages only . Communication was further restricted in such a way that messages could pass from a participant only to those members of the team linked to the participant by open channels of communication and all networks were initially centralized, forming a "wheel" : A central position (C) could communicate directly with all others but the four peripheral positions (P) could communicate only with or through C (also from Bavelas 1950) . Incentives Teams earned a reward for each correct solution. Team earnings were divided equally among members . However, on each trial of the experiment there was also a bonus for the person who first submitted a correct solution . The value of the bonus on each trial was equal to the total value of a team's earnings on any given trial had each member of the team obtained the correct solution . In a centralized network, only C ever won the bonus . This was uniformly seen by all participants as unfair and the inequity was typically blamed on centralization of the network . To underline the inequity, participants were told that all of them were alike and had been allocated to positions in the network purely by chance . Dependent Variable Participants were allowed to change the structure of the network at any trial of the experiment by adding or subtracting channels of communication if a majority agreed (MacKenzie 1976) . An all-all network, for example, would equalize opportunity to win the bonus . There was a small fee for adding channels . Therefore, to alter channels of communication, a member of the group had to make a motion specifying the proposed changes and the motion had to be seconded by another member of the group . If the motion were seconded, an election would be held to determine if a majority agreed . E never actually held an election : In previously published experiments in this setting all Ss were in peripheral positions of the network and the dependent variable of the experiment was a change-response by an S in the P position, that is, a motion by P, addressed to P', to change the structure of the network . Ss were stopped, interviewed, and debriefed at any trial at which they made a change-response or, if they did not, if they completed 10 trials of the experiment (for reviews see Walker et al . 1991 ; Walker and Zelditch 1993 ; Zelditch et al .



The Normative Regulation of Power

1 63

1983 ; Zelditch and Walker 1984 ; for specific experiments, see Ford and Zelditch 1988 ; Thomas et al . 1986 ; Walker et al . 1986 ; Walker et al . 1988 ; Zelditch and Ford 1994) . But in the experiments reported in the present paper, Ss were either in a C position or in a position to support C . The dependent variable was either (1) resistance by an S at the C position to a change-response made by a confederate at a P position; or (2) support of a change-response (made by a confederate at either C or P) by an S who was a peer of C (C') ; or (3) support of a change-response (made by P) by a peer of P (P') . In all three cases, a confederate made an attempt to change the structure of the network at the end of the second or third trial of the experiment . When S is at the C position, S effectively controls the agenda of any decision by the group because all messages pass through C . A change-response by C was therefore measured by the rate at which C forwarded change-responses made by and addressed to others . (Thus, the forwarding rate, P(F) = df the probability that C forwards a change-response .) Use of the agenda power to delay or prevent change is an abuse of C's power unless authorized by E . The use of C's power to delay or prevent change was measured by the rate at which C does not forward the confederate's message to its addressee (1-P(F)) . Support was measured by (1) the execution of directives (from C or P) by C' ; (2) the endorsement of them by P' ; or (3) positive or negative sanctions by C' or P' for compliance or noncompliance with either rules or directives (from C or P) . To create a peer of C, E assigned S to a role as an assistant to E in E's office and support was measured by the rate at which S authorized change (or resistance to change) by a confederate at either C or P . (This measured the probability of authorization, P(A) .) To create peers of P, a confederate at a P position attempted change and support was measured by the rate at which Ss at other P positions endorsed it . (This measured the probability of endorsement, P(E) .) Sanctions by C' were measured by re-assignment of leaders in the two-team variant of the standardized experimental setting described below . Sanctions by P' were measured by acceptance or rejection of P as a co-worker at the end of the standard experiment .

A Variant of the Standardized Experimental Setting One of the experiments we describe in the present paper (from Walker et al . 1997) was set in a variant of the standardized experimental setting in which C' assisted E by monitoring performance (at the same task) by two teams, half of which were centralized and half decentralized . There was no bonus in this setting, no change procedure, and all Ss (at C') completed five trials of the experiment . Instead of a bonus, the procedure creating inequity was that team leaders of each team made recommendations to C' about how team earnings should be divided, half of which were equal and half unequal . Centralization made unequal divisions equitable while decentralization made them inequitable .



1 64

MORRIS ZELDITCH and HENRY A. WALKER

An important difference between the standardized and variant settings was that the variant gave C' more autonomy than the standardized setting . C was responsible not only for monitoring team performance (e .g ., calculating the number of correct solutions), calculating team earnings, and communicating results, but also for making final decisions about how to divide team earnings and reassigning team leaders at what Ss at C' thought was mid-experiment, that is, at the end of the fifth trial of the experiment . Overview

The nine experiments described in the present paper divide into five in which S is at the C position and four in which S is either C' or P' . The five in which S is at the C position test hypotheses about the effect of internal and/or external constraints on the use of the agenda power by C . The four in which S is either C' or P' test hypotheses about the effects of valid norms, values, beliefs, practices, and procedures on authorization or endorsement .

MECHANISMS A Baseline Experiment

If S is at the C position, a bonus for the first correct solution of the task has two sides to it : It is both a gain to C and an inequity for which C is responsible . A baseline experiment showed that the two effects tend to cancel each other . It also showed considerable abuse of the agenda power even when there was no bonus . The purpose of the baseline experiment was to provide a standard against which to measure all other results of experiments with S at the C position . In the control group of the baseline experiment (which was carried out in the standardized experimental setting), no bonus was awarded, the agenda power was neither authorized nor de-authorized, and the wheel was neither endorsed nor dis-endorsed . Under these conditions, 71 percent of Ss at the C position forwarded a change-response made by a confederate at a P position . But even without the incentive of a bonus, 29 percent of Ss in the C position delayed or prevented change in the wheel . A post-session measure of the efficiency and propriety of the wheel showed that, other things being equal, Cs in the control group believed the wheel was the most efficient, and therefore the "right," way to organize for performance of the task, which in turn justified attempts to delay or prevent change (cf . Lipset's "effectiveness" hypothesis [1959] ; see also Linz 1978) . 3 In two treatment conditions a bonus of either $0 .75 or $1 .50 was awarded for the first correct solution of the task . The size of the bonus did not significantly affect the forwarding rate . At $0 .75, P(F) =.52 ; at $1 .50, P(F) = .60. (The likeli-



The Normative Regulation of Power

1 65

hood ratio chi square L2 =2.10, P = n . s .) . The most likely explanation for this is that the increase in gain was offset by the increase in inequity . There was clearly a "gain effect," because 25 percent of Cs in the two bonus conditions neither forwarded P's change-proposal nor redistributed the bonus by other means . But, as gain increased so did the off-setting effect of inequity . Hence, the size of the gain did not affect the forwarding rate . However, gain had more, hence inequity arguably less, effect on post-session measures of the efficiency and propriety of the wheel . Without a bonus, Cs in this setting uniformly regarded the wheel, because it was efficient, as the proper way to organize for performance of the task (see note 3) . It was the bonus that de-legitimated the wheel : A bonus of $0 .75 significantly decreased both its efficiency and propriety . However, at a bonus of $1 .50 both the efficiency and propriety of the wheel increased again and the deviation from linearity was in both cases statistically significant . (For the post-session measure of efficiency, F for the deviation from linearity was 4 .58, P = .004 . For the measure of propriety, F = 5 .10, P = .03) . 5 Mechanisms Constraining the Use of Power by Authority Propriety

The effect of the bonus on measured propriety is an important factor to take into account in interpreting the behavior of C because propriety, in general, had a significant effect on the forwarding rate : Aggregating over five experiments with S at the C position, P(F) was inversely proportional to the propriety of the wheel (see Table 1, row 2) . If the wheel was proper, C felt empowered to delay or prevent change ; if the wheel was improper, C felt constrained to forward a change-proposal . However, although gain affected propriety, the effect of propriety on P(F) was not due simply to gain : P(F) was correlated with propriety even in those experiments in which gain was a constant (Phi= - .16, P < .001) . Thus, propriety affected P(F) even when it was not self-serving .

Table 1 .

Source of or Type of Support for Legitimacy Validity Propriety Validity x Propriety Authorization Endorsement

Odds of Not Forwarding a Change-Proposal by Source of or Type of Support for Legitimacy

N 88 307 88 40 41

Odds

Ratio

4 .50 4 .50 8 .68 4 .56 1 .08

L2 7 .50 12 .22 2 .74 8 .60 1 .50

df

P

1 1 1 1 1

< .001 < .10 < .01 n .s .

< .01



1 66

MORRIS ZELDITCH and HENRY A . WALKER

Validity

In the Dornbusch-Scott theory of authority, validity, independent of propriety, regulates the use of power by authority . In order to test this hypothesis, we made the bonus, and hence the wheel, proper by emphasizing the effort required by C's role ; or improper, by saying nothing to justify the bonus . We made the wheel valid by making change invalid, instructing Cs that change damaged E's objectives in this particular experiment ; or valid, by saying nothing about the validity of change-Cs were left to infer that change was valid from the existence of a change procedure in the experiment . The agenda power was neither authorized nor de-authorized, the wheel neither endorsed nor dis-endorsed . But in this particular experiment, the confederate at P sent not one but four change-proposals to S at the C position, E stopping S the first time S forwarded a change-proposal . Thus, the measure of the forwarding rate in this experiment is the probability of ever forwarding a change-proposal . We found that independent of the effect of propriety, there was a direct effect of validity on P(F) (Table 1, row 1) : If change was invalid, C felt empowered to delay or prevent change by not forwarding a change-proposal . If change was valid, C was constrained from using the agenda power to delay or prevent change . Furthermore, validity interacted with propriety (though the significance level was marginal) in such a way that Cs were more willing to follow their conscience if it was backed by validity (Table 1, row 3) . Authorization

In addition to the direct effect of validity on P(F), validity should have an indirect effect through its effect on expected authorization . We did not test the hypothesis that validity creates expectations of authorization, but we did test the effect that authorization, as a proxy for expected authorization, has on P(F) . In order to test the hypothesis, E both authorized and de-authorized use of the agenda power by C . The wheel was neither endorsed nor dis-endorsed and the bonus was $1 .50 in both conditions . Authorization of the agenda power had the largest effect found in any experiment in which S was at the center of the wheel (Table 1, row 4) . If the agenda power was authorized, C felt empowered to not forward change-proposals and only 35 percent of Cs forwarded them . If the agenda power was de-authorized, C felt constrained to forward change-proposals and 80 percent of Cs forwarded them . Endorsement

In order to test the expected endorsement hypothesis, E manipulated the outcome of an election that Ss at the C position knew would not determine the structure of the network but that did express the majority's preferences for or against



1 67

The Normative Regulation of Power

the wheel . Half the Cs were informed that a majority favored the wheel, half that it favored an all-all network . The agenda power in this experiment was neither authorized nor de-authorized and again the bonus was $1 .50 . The expected endorsement hypothesis was not confirmed by this experiment . There was in fact a direct effect of endorsement on propriety (The odds ratio=8 .18, L2 = 7 .52, P < .01) and P(F) was inversely proportional to propriety (Table 1, row 2), but there was no direct effect of endorsement on P(F) when propriety was controlled (Table 1, row 5) . 6 That endorsement had no direct effect on P(F) would be easy to explain if peripheral positions had no power, but in fact the task in this experimental setting is such that any individual in a P position has the power to prevent anyone else from ever getting the correct solution simply by withholding the information that she or he controls . But it is possible that Ss in the C position were not conscious of P's power or, even if they were, simply took it for granted that P would without question provide others with the information on which the solution of the task depended . We therefore tested the expected endorsement hypothesis again by making C more aware of the power of P and manipulating the expected probability that P would use the position's potential power . In order to magnify the effect of this manipulation we authorized the use of the agenda power by C and dis-endorsed the wheel (Zelditch 1999) . Unfortunately, we failed to anticipate the effect that the illegitimacy of the use of P's potential power would have on C : We did find that P(F) was directly proportional to the cost to C of the use of power by P (Table 2, column 2) . But the effect was curvilinear, because P(F) was also inversely proportional to the legitimacy of the use of power by P . Withholding task information violated task norms, making use of it as power by P illegitimate . At a low probability of retaliation, Cs who were made aware of P's power decreased P(F) because illegitimacy of such a use of power by P offset the costs to C of its use . At a high probability of retaliation, the use of power by P was even less legitimate, as is indicated by the significant decrease in the probability of redistribution by other means (Table 2, column 3) . But the negative effect of legitimacy was offset by the increase in expected costs to C of the use by P of P's power . (The chi square for the deviation from lin-

Table 2 .

Odds of Authorizing or Endorsing a Change-Proposal by Source of and Type of Support for Legitimacy

Source of Legitimacy

Type of Support

Validity Interest Validity Propriety Endorsement

Authorization Authorization Endorsement Endorsement Endorsement

N

Odds Ratio

L2

df

P

83 83 88 168 80

8 .90 3 .09 2 .50 5 .18 4 .00

21 .89 6 .70 4 .60 25 .70 9 .50

1 1 1 1 1

< .001 < .05 < .05 < .001 < .01

168

MORRIS ZELDITCH and HENRY A . WALKER

Table 3 .

Odds of Sanctioning the Source of a Change-Proposal by Source of and Type of Support for Legitimacy

Source of Legitimacy

Type of Support

N

Odds Ratio

L2

df

Validity Propriety Validity Propriety

Authorization Authorization Endorsement Endorsement

83 83 88 88

0 .28 0 .08 0.38 0 .65

9.14 24 .00 4 .42 1 .30

1 1 1 1

P < .01 < .001 < .05 n .s.

earity is significant [Table 2, rows 6 and 7 of column 2] .) Thus, given power at the periphery, the effect of expected endorsement is to empower C to delay or prevent change if the wheel is endorsed but to constrain C if the wheel is dis-endorsed (for details see Zelditch 1999) . Normative Regulation of Constraints On the Use of Power by Authority Authorization

Other things being equal, almost all Ss at C' authorized an election to change the structure of the wheel in the standardized experimental setting . In an experiment in which C' assisted E by evaluating team performance (against an answer key), calculating and allocating team earnings, and communicating results to team members, C' also had the agenda power. When the wheel was made improper by a bonus of $1 .25 and change was valid, 95 percent of Ss at C' authorized an election (proposed by a confederate at the P position) that would change the structure of the wheel (Zelditch, Gilliland, and Thomas 1984) . If change was invalid, 39 percent of Ss at C' approved an election to change the wheel-a large number, certainly, but a significant decrease due to validity (Table 3, row 1) . Furthermore, the effect of validity was independent both of preferences, also manipulated in the experiment, and of propriety . Gain was manipulated in this experiment by paying C' either a flat fee for participation or an amount equal to the highest amount paid to any member of the team C' supervised . A vested interest in the status quo had a significant independent effect on authorization (Table 3, row 2), but the effect of validity was independent of gain and the interaction of the two was not statistically significant . Propriety should have little or no effect on authorization . But propriety did not differ by condition in the experiment just described . It was equally low in all conditions. To test the propriety hypothesis, it was necessary to re-design the standardized experimental setting, which was the reason for a two-team variant of it . In a two-team setting C' again assisted E by evaluating team answers (from an answer key), calculating team earnings, and communicating results to each team, but in addition made final decisions about the distribution of team earnings rec-



1 69

The Normative Regulation of Power

ommended by C in each team . C' was also charged with responsibility for reassigning C at mid-experiment if C' believed it was required . An important feature of the experiment-one we underestimated-was the much greater structural autonomy of C' in this setting . In E's instructions, emphasis was placed on the fact that the decisions of C' were final . To create propriety we varied the structure of the teams instead of introducing a bonus . Absent a bonus, which creates a very large inequity, a centralized network does legitimate some inequality (Walker et al . 1997) . Therefore, we used a wheel to legitimate recommendations by C of a moderately unequal division of team earnings and used an all-all network to de-legitimate them . ("C" in an all-all network was arbitrarily assigned .) Half of the confederates at the C position recommended an equal division, half recommended an unequal division of team earnings . Validity was also manipulated in this experiment : To create validity, E either instructed C (overheard by S at C') that team earnings should be divided equally or made no rule about how to divide team earnings . Because both the structure of the network and the distribution rule had an effect on the post-session measure of propriety, this experiment has to be thought of in terms of three rather than two effects on authorization : An effect of network structure (manipulated propriety), validity (whether or not their is a rule), and measured propriety (a co-variate due to both structure and rule) . And in fact, contrary to hypothesis, acceptance of unequal pay recommendations by C' was directly proportional to the propriety of inequality (Table 4, row 3 of the right-hand panel of the table) . In fact, where validity and propriety were inconsistent, 33 percent of C's violated the rule in order to act on their conscience . Even worse for our hypothesis, sanctions were also directly proportional to mea-

Table 4.

Mean Earnings of C Authorized by Subjects Assisting E, by Source of Legitimacy with Selected Statistics of Repeated Measures Analysis of Variance

Source of Legitimacy: Inequality is

Repeated Measures Analysis of Variancea Mean Earnings of C

Proper and Valid

$3 .14

Proper and Invalid

$2 .94

Improper and Valid

$2 .38

Improper and Invalid

$2 .49

Note:

a. b.

Source of Variation Validity Structure Propriety b VX S V x P S X P V x S x P

Sum of Squares

df

F

P

.03 .08 7 .31 .75 .04 .24 1 .21

1 1 1 1 1 1 1

.16 .42 37 .49 3 .83 .21 1 .25 6 .21

n .s . n .s . < .001 .05 n .s . n .s . .02

Omitting constant, error, and within S's source of variation . No within-S's factor was statistically significant. Both the strcuture of the communication network and the rule affected propriety as measured in the post-session questionnaire . The means are for the types of structure that make inequality proper or improper . The RMANOVA distinguishes these structures from measured propriety .

170 Table 5.

MORRIS ZELDITCH and HENRY A . WALKER Odds of Sanctioning the Source of a Change-Proposal by Source of and Type of Support for Legitimacy

Source of Legitimacy

Type of Support

Validity Propriety

Authorization Authorization

Validity Propriety

Endorsement Endorsement

Odds

L2

df

3 .62

9 .14

1

< .01

0 .08 2 .62 1 .55

24 .00 4 .42 1 .30

1 1 1

< .001 < .05

N

Ratio

83 83 88 88

P

n .s .

sured propriety, independent of the effect of either structure or validity (Table 5, row 2) . Not only did propriety have a main effect on both the division of team earnings and sanctions, but validity, though it did have a direct effect on sanctions (Table 5, row 1), had only an indirect effect, through propriety, on the division of team earnings (Table 4, row 1 of the right-hand panel) . Validity did interact with propriety to increase the effect of propriety on the division of team earnings : The effect of manipulated propriety was stronger if it was backed by validity and the effect of measured propriety was stronger if both structure and validity backed it (Table 4, rows 4 and 7 of the right-hand panel) . But in the case of sanctions, manipulated propriety (Table 5, row 2), measured propriety (not shown in table 5), and validity (Table 5, row 1) all had direct effects on sanctions and none of the interactions were statistically significant . All this, we believe, was due to the structural autonomy of C' in the two-team setting . In the standardized experimental setting, authorization was directly proportional to validity (Table 3, row 1) . There, C' was responding to a change-proposal by P, not C, which should have increased the effect of personal inclinations, but was less autonomous than C' in the two-team setting . Aside from the difference in the confederate's position, the critical difference between the two experiments is the freedom of C' from constraint in the two-team setting . Endorsement To study the effect of validity and propriety on endorsement, we varied both the validity of change and the propriety of the wheel in the standardized experimental setting with both S and the confederate in P positions . A confederate at P sent a change-proposal to S, also at P, who either did or did not second ("endorse") it . On the post-session questionnaire, sanctions were also measured . The Ss (as in most of the nine experiments) were male . Propriety was manipulated by awarding a bonus for the fastest correct solution in half the groups but no bonus in the other half. Validity was manipulated by making change invalid (because it damaged the objectives of E) in half the groups and valid in the other half .



The Normative Regulation of Power

1 71

Endorsement by P' of change proposals by P was inversely proportional to both the propriety and validity of the wheel . (See Table 3, rows 3-4 . Row 4 aggregates the results of two experiments .) Their interaction was not significant . Sanctions by P' were directly proportional to the validity of the wheel but propriety had no effect on them (Table 5, rows 3-4) . On the other hand, a separate experiment that directly manipulated endorsement of the wheel by others, though it found that-independent of proprietyendorsement of the wheel by others decreased endorsement of change by P' (Table 3, row 5), found too few sanctions to test the hypothesis that sanctions were proportional to endorsement . The difference in the probability of sanctions between the two studies of endorsement is too large to be due simply to chance (57% versus 5%) . The Ss in the second of these experiments were female and the gender of the Ss is of course a possible explanation of the differences in the rate of sanctions . But this explanation is inconsistent both with other evidence in these two experiments (see Walker et al . 1990) and evidence in other experiments in the program (see Walker and Smith-Donals 1985) . A more likely explanation is the difference in the source of legitimacy between the two experiments . Validity (the source of legitimacy in the first case) is a normative effect of consensus and is clearly enforceable . On the other hand, the effect of endorsement (the source of legitimacy in the second case) is equivocal: If one member of the group proposes change, she or he is not necessarily violating a norm because she or he is also changing the state of consensus in the group and decreasing consensus weakens backing for sanctions (Walker et al . 1990) . Frailties in the Normative Regulation of External Constraints on the Use of Power by Authority

An important finding of experiments on the normative regulation of external constraints on the use of power by authority is that the same frailties found in internal constraints on the use of power are found in its external constraints . Self-interest, self-serving propriety, and eccentric propriety are all frailties in the internal constraints on the use of power by authority . But self-interest and eccentric propriety also affected authorization (see IVC(1)) . Our interpretation of the latter finding-that it depended on structural autonomy-raises an important and more general question : Under what conditions do norms actually constrain the use of power by authority? The same question arises from other anomalies in the findings : That endorsement had no effect on an S at the center of the wheel until S was made aware of the potential power of the periphery-implying that the effect of external constraints depends on power-dependence (IVB(4)) ; that decreasing endorsement by others decreased the likelihood of sanctions by P'-implying the dependence of sanctions on consensus (IVC(2)) . These anomalies cast doubt on the initial conditions of our experiments, not the general conceptual and theoreti-



1 72

MORRIS ZELDITCH and HENRY A . WALKER

cal ideas of the theory . They raise questions not about the mechanisms of the process but about the conditions under which norms regulate power . Therefore, we now turn to the question of what our theory and results imply about the conditions under which norms regulate power . We do not ask the conditions under which there are frailties ; what we ask is the conditions under which, if there are frailties, norms regulate them . And what we offer are not findings, but untested hypotheses : the deductions of the conditions of the normative regulation of power from our theory of authority were not tested by the experiments we have reported in the present paper and the interpretation of anomalies in our results are ex post facto, hence also untested .

CONDITIONS The Dornbusch-Scott theory of authority implies four conditions on which normative regulation of the use of power by an authority A depends : consensus ; density of ties to A ; dependence of A on peers and superiors, A', and/ or subordinates, B and B' ; and density of ties to A' . Consensus The entire structure of the Dornbusch-Scott theory of authority depends on validity . But validity presupposes consensus (Zelditch and Floyd 1998) . Consensus is therefore an important condition under which norms regulate the abuse of power. In the first instance, actual consensus conditions the effect that validity has on authorization by A' and endorsement by 13' . In addition, the perception of consensus by A conditions the effect that validity has on A's expectations of authorization by A' and endorsement by B' . Actual dissensus decreases validity . A decrease in validity has five effects : First, it decreases the frequency of sanctions by A' and B' . Second, it decreases the association between validity and authorization by A' . Third, it decreases the association between validity and endorsement by B' . Furthermore, it also decreases the effect that validity has on the strength of the association between propriety and authorization by A' and, fifth, the effect that validity has on the strength of the association between propriety and endorsement by B'-as consensus decreases, both A' and/or B' are more likely to act on their own sense of what is proper, whether or not backed by validity . A decrease in the effect of validity on authorization and endorsement is one way in which actual dissensus undermines the presupposition of consensus by A . Whether by this or other means, the perception of dissensus by A decreases the effect that validity has on expectations by A of authorization by A' and/or endorsement by B' . The effect of perceived dissensus on the strength of the association between validity and expectations of authorization and endorsement, in



The Normative Regulation of Power

1 73

turn, decreases the effect of expected authorization and endorsement on the use of power by A . It also decreases the effect that validity has on the association between A's sense of propriety and the use of power by A . If A believes the norms, values, beliefs, practices, and procedures supposedly governing A's actions are not shared, A is more likely to act on his or her own sense of what is proper even if it is not backed by validity . Density of Structural Constraints on A

But even if there is consensus, the scope of the Dornbusch-Scott theory of authority presupposes accountability of A to others : In organizations, to boards of directors, to peers who execute directives and sanction noncompliance, to subordinates on whose performance A depends for outcomes valued by the organization ; in communities, by extension, to those on whom selection of authorities, mobilization of resources, and enforcement of authority depend . Anything that weakens accountability increases the structural autonomy of A . Structural autonomy of A decreases external constraints on A (cf . Burt 1992 ; Markovsky, Willer, and Patton 1988) . Given the frailty of internal constraints, the abuse of power is therefore proportional to the structural autonomy of A . The structural autonomy of A decreases the frequency of sanctions by A' . But it also decreases the association between validity and expected authorization (or nonauthorization) by A' . It therefore decreases the association between validity and the use of power by A . It also attenuates the effect validity has on the association between A's sense of propriety and the use of power by A . Because redundancy of the network of ties to A decreases structural autonomy (Burt 1992 ; Markovsky, Willer, and Patton 1988), external constraints on A therefore depend on the density of the network of ties in which A is embedded . The greater the redundancy of these ties, the greater the association between validity and expected authorization by A', the greater the association between validity and the use of power by A, and the greater the effect of validity on the strength of the association between propriety and the use of power by A . The normative regulation of power is therefore proportional to network density at A . Power Dependence

But even if A is not structurally autonomous, the effect of external constraints on A depends on the extent to which A is dependent on others for resources valued by A (Emerson 1962, 1972) . If density is held constant, the strength of the effect of sanctions by A' and B' increases as A's dependence on them increases . As A's dependence on them decreases, external constraints are enfeebled and neither validity nor expected support constrains the use of power by A . The greater the dependence of A on A', the greater the effect of expected authorization . The greater the dependence of A on B', the greater the effect of expected endorsement .



1 74

MORRIS ZELDITCH and HENRY A . WALKER

The greater the effect of either expected authorization, expected endorsement, or both, the greater the association between validity and the use of power by A and the greater the effect of validity on the strength of the association between A's sense of propriety and the use of power by A . Structural Constraints at A' But who guards the guardians? If the possibility of an independent counsel constrains the president, what constrains the independent counsel? Even if A is constrained by dependence on A' and/or B', the frailties in constraint regress if A' or B' is autonomous. (This seems less likely in the case of B' who we neglect for the rest of this section .) Where there is only one A' on whom constraint of A depends, any eccentricity of A' is itself unconstrained . 8 The normative regulation of power therefore depends as much on density at A' and the dependence of A' on others as it does on density at and dependence of A . Redundancy of ties to A', and dependence by A' on peers and/or superiors other than A, increases the association of validity with authorization by A' and the effect validity has on the strength of the association between what A' believes is proper and authorization by A' . To the extent that agents of social control are densely interconnected, the eccentricities of any one agent's propriety is constrained and normative consensus in fact constrains the abuse of power by authority.

CONCLUSIONS The first conclusion we came to was that internal constraints are not sufficient to normatively regulate the use of power by authority : Some uses of power by authority remain self-interested ; even if they are proper, propriety may be self-serving ; even if propriety is not self-serving, it may be idiosyncratic-not only is propriety not always the solution, it is often the problem . External constraints are therefore a necessary condition of the normative regulation of power . The second conclusion we came to was that an elaboration of the Dornbusch-Scott theory of authority is a satisfactory theory of the mechanisms of the normative regulation of power. The evidence supported all the hypotheses of the theory concerning the effects of external constraints on A . Validity had a direct effect on the use of power by A . It had indirect effects not only through propriety but also through expectations of authorization by peers or superiors of A and endorsement by subordinates of A-though in the case of expected endorsement, only after strengthening the power of the subordinate, B . Furthermore, the effect of propriety on the use of power by A was stronger if it was backed by validity . But the most important hypothesis of the theory is that it is validity that drives external constraints on the use of power by A. And support of this hypothesis was mixed . In support of the theory, independent of propriety, authorization by A' of



The Normative Regulation of Power

1 75

change by B was directly proportional to validity . The effect of propriety on the authorization by A' of division of team earnings by A was stronger when backed by validity . Endorsement by B' of change by B was also directly proportional to validity, as well as to endorsement by others . And validity had a direct effect on sanctions by either A' of A or B' of B . But contrary to hypothesis, in one of two experiments on the normative regulation of authorization, validity affected authorization of team earnings only indirectly, through propriety . And in one of two experiments on the normative regulation of endorsement, endorsement by the majority, though it had a direct effect on endorsement of change, had no effect on sanctions . But these anomalies cast doubt on the initial conditions of the experiments, not the basic concepts and laws of the theory : structural constraints on A in the first case, group consensus in the second case . They raised questions not about the mechanisms of the process but about the conditions under which norms regulate power . Similar questions were raised by the third conclusion to which we came : That the frailties we had found in the internal constraints on the use of power by A regress, if not infinitely, at least to the external constraints used by A' and/or B' . Interest in gain had an independent effect on authorization by A' of change by B . In the two-team setting, when validity and propriety were inconsistent, a third of A's peers followed their conscience rather than the rules . And (again in the two-team setting), propriety had its own independent effect on sanctions of A by A' . With respect to the conditions under which norms regulate power, we came to the conclusion that four are necessary : Consensus, a dense network of ties to A, dependence of A on either A', B', or both ; and a dense network of ties among agents of social control . Absent consensus, its presupposition, on which the effects of validity depend, is undermined . Absent a dense network of ties to A, accountability, on which the entire structure of the argument that norms regulate power depends, is undermined . If A does not care about sanctions by A' and/or B', the process of social control in which the norms regulating power are embedded is undermined . Undermining any of these three conditions (1) decreases the effect of validity on actual and expected support by A' and B' and therefore (2) decreases the effect of validity and expected support by A' and B' on the use of power by A . In addition, (3) it decreases the effect of validity on the strength of the association between propriety and support by A' and/or B' . Finally, (4) it decreases the association between A's sense of propriety and the use of power by A . But regress of the frailties in constraint depends also on the density of ties among agents of social control . The sparser the ties, especially to A', the weaker the effect of validity on support and on the strength of the association between propriety and support . It is only to the extent that agents of social control are densely interconnected that the eccentricities of any one agent's propriety is constrained and normative consensus in fact regulates the use of power by authority .



176

MORRIS ZELDITCH and HENRY A . WALKER

ACKNOWLEDGMENT We would like to acknowledge the support of NSF grants SOC 7817434 for research on "Group Determinants of Agenda Setting," SES 8420238 for research on "Legitimacy and the Stability of Authority," and SES 8712097 for research on "Agenda Gatekeeping."

NOTES 1 . This is an intentionally very abstract, hence inclusive, expression for any kind of "element" of the external social order of a group, whether norms, values, beliefs, practices, or procedures . 2 . This is as true of values, beliefs, and practices as it is of norms . Whether or not e is a prescriptive element of the pre-given social framework that structures a particular encounter, if it is valid it is prescriptive, hence "binding," in any particular encounter (Berger and Zelditch 1993) . 3 . In a post-session questionnaire, Ss in the control group rated the mean efficiency of the wheel 1 .57 on a scale from 1 (very efficient) to 5 (very inefficient) . Approval of the wheel was used to measure its propriety . On a scale from 1 (very highly approve) to 5 (very highly disapprove), the mean propriety of the wheel was 2 .33 . 4. Though we did not intend to have a second dependent variable, Cs were quite imaginative about ways to redistribute the bonus . They found ways to split it, despite the fact that E instructed them not to do so, or to rotate it, or to delay their own answer until each other P in turn had had a chance to submit an answer first . About 25 percent of Cs in the baseline experiment redistributed the bonus by these means . Furthermore, in the five experiments in which S is at the center of the wheel, there is a negative correlation between the probability of redistributing by other means and the probability of forwarding a change-proposal (Phi =- .18, L 2=12 .2, P < .001) . But the percent of the variance explained is not large enough to account for the effects of the treatment variables manipulated in these experiments . 5 . In general, the forwarding rate was correlated with measured propriety (see Table 1, row 2) . This result was obtained by aggregating over five experiments, including the baseline experiment . But the strength of the correlation was not strong enough to significantly differentiate the forwarding rate of the two bonus conditions in the baseline experiment. 6 . The odds ratio in row 5 of Table 1 indicates that, despite the indirect effect of endorsement on P(F), there was no zero-order association between endorsement and P(F) in this experiment . This was due to the fact that the effect of endorsement on propriety was positive while the effect of propriety on forwarding was negative . In this experiment, the opposing effects masked each other when the 2 x 2 x 2 table was collapsed across categories of propriety . 7 . We follow Emerson (1962, 1972) in assuming that the dependence of p on o for resource R is directly proportional to the value of R to p but inversely proportional to the number of alternative sources of R available to p . 8 . This paper was written about the time of the investigation of Clinton by Staff and sometimes sounds as if we were actually writing about them-especially, perhaps, Starr-but in fact the original proposal for the research dates from 1987 .

REFERENCES Barnard, C . 1938 . Functions of the Executive . Cambridge, MA: Harvard. Bavelas, A . 1950 . "Communication Patterns in Task-Oriented Groups ." Journal of the Acoustical Society of America 22 : 725-30 . Berger, J ., and M . Zelditch. 1993 . "Notes on a Theory of Institutionalization ." Unpublished manuscript . Stanford, CA : Stanford University .



The Normative Regulation of Power

177

Burt, R . 1992 . Structural Holes . Cambridge, MA : Harvard . Dornbusch, S . M., and W. R. Scott. 1975 . Evaluation and the Exercise of Authority . San Francisco, CA : Jossey-Bass . Emerson, R . 1962 . "Power-Dependence Relations ." American Sociological Review 27 :31- 41 . Emerson, R . 1972 . "Exchange Theory." Pp 38-87 in Sociological Theories in Progress, Vol 2, edited by J . Berger, M . Zelditch, and B . Anderson . Boston, MA : Houghton Mifflin . Ford, J . B ., and M . Zelditch . 1988 . "A Test of the Law of Anticipated Reactions ." Social Psychology Quarterly 51 : 164-71 . Linz, J. 1978 . The Breakdown of Democratic Regimes .' Crisis, Breakdown, and Re-equilibration . Baltimore, MD : Johns Hopkins . Lipset, S . M . 1959 . "Some Social Requisites of Democracy : Economic Development and Political Legitimacy ." American Political Science Review 53 : 69-105 . MacKenzie, K . 1976 . A Theory of Group Structures . New York: Gordon and Breach . Markovsky, B ., D . Willer, and T. Patton . 1988 . "Power Relations in Exchange Networks ." American Sociological Review 53 : 220-36. Tetlock, P. 1992 . "The Impact of Accountability on Judgment and Choice: Toward a Social Contingency Model ." Advances in Experimental Social Psychology 25 : 331-376 . Thomas, G ., H . Walker, and M . Zelditch, 1986 . "Legitimacy and Collective Action ." Social Forces 65 : 378-404 . Walker, H ., L . Rogers, K . Lyman, and M . Zelditch . 1990 . "Legitimacy and the Support of Revolutionary Coalitions." Unpublished manuscript . Ithica, NY : Cornell University . Walker, H ., L . Rogers, G . Thomas, and M . Zelditch. 1991 . "Legitimating Collective Action : Theory and Experimental Results ." Pp 1-25 in Research in Political Sociology, Vol 5, edited by P . Washburn . Greenwich, CT : JAI . Walker, H., L . Rogers, and M . Zelditch . 1988 . "Legitimacy and Collective Action : A Research Note ." Social Forces 67 : 216-228 . Walker, H., and L . Smith-Donals . 1985 . "Gender or Status : The Effects of Differences in Sex on Behavior Under Certain Conditions of Disadvantage ." Working Paper No . 85 .2 . Stanford University, Center For Sociological Research . Walker, H ., G . Thomas, and M . Zelditch, 1986 . "Legitimation, Endorsement, and Stability ." Social Forces 64 : 620-643 . Walker, H., and M . Zelditch . 1993 . "Power, Legitimacy, and the Stability of Authority : A Theoretical Research Program ." Pp 364-381 in Theoretical Research Programs : Studies in the Growth of Theory, edited by J . Berger and M . Zelditch . Stanford, CA : Stanford. Walker, H ., M . Zelditch, P. Taylor, and V . Montoya . 1997 . "Choosing Standards: Validity, Propriety, and Pay Equity ." Unpublished manuscript . Ithica, NY : Cornell University . Weber, M . [1918] 1968 . Economy and Society, edited by G . Roth and C . Wittich and translated by E . Fischoff et al. New York: Bedminster. Zelditch, M . 1999 . "Legitimacy, Expected Value, and the Strength of Peripheral Power ."Paper presented at the West Coast Conference on Group Processes, Stanford University, Stanford, CA . Zelditch, M ., and A . Floyd. 1998 . "Consensus, Dissensus, and Justification ." Pp 339-368 in Status, Power, and Legitimacy, edited by J . Berger and M . Zelditch . Stanford, CA : Stanford . Zelditch, M ., and J . Ford . 1994 . "Uncertainty, Potential Power, and Non-decisions ." Social Psychology Quarterly 57 : 64-73 . Zelditch, M ., E. Gilliland, and G . Thomas . 1984 . "The Legitimacy of Redistributive Agendas ." Working Paper # 84-9 . Center For Sociological Research, Department of Sociology, Stanford, CA, Stanford University. Zelditch, M., W . Harris, G . Thomas, and H . Walker. 1983 . "Decisions, Non-decisions, and Meta-decisions ." Pp 1-32 in Research in Social Movements, Conflict, And Change, Vol 5, edited by L. Kriesberg . Greenwich, CT: JAI .



178

MORRIS ZELDITCH and HENRY A . WALKER

Zelditch, M ., and H. Walker. 1984. "Legitimacy and the Stability of Authority ." Pp 1-25 in Advances in Group Processes, Vol 1, edited by E . Lawler. Greenwich, CT : JAI.

BEYOND THE CONCEPT OF VALUE IN POWER-DEPENDENCE THEORY EXPANDING A MODEL OF THE "WHOLE ACTOR"

John F . Stolte

ABSTRACT Emerson's (1987) theory suggests that some notion of "the whole actor" is axial to a comprehensive understanding of value in social exchange . Building from this notion, the present paper explores four pathways along which the theory might be clarified and extended . First, it explores the intricate connections that exist among need, uncertainty, and value, bringing to bear the operant conditioning and the rational choice perspectives . Second, it examines the complex interplay among extrinsic motivation, intrinsic motivation, and value . Third, it considers the dynamic oscillation that may occur between rational/cognitive and mindless/emotional processes in behavioral decision making . Fourth, it distinguishes socially intrinsic from socially extrinsic environmental value domains . The present analysis provides a basis for continued theoretical refinement and focused empirical research on Emerson's potentially important theory of value .

Advances in Group Processes, Volume 17, pages 179-202 . Copyright © 2000 by JAI Press Inc . All rights of reproduction in any form reserved . ISBN : 0-7623-0651-3 179



180

JOHN F . STOLTE

INTRODUCTION In an unfinished but intriguing formulation, Emerson (1987) advanced the core of a social psychological theory of value . The formulation is ambiguous at points, because it is a preliminary statement, a work in progress, le